Show all abstracts
View Session
- Front Matter: Volume 6510
- Keynote
- Dual Energy
- Performance Assessment
- Innovative Imaging I
- Detector Technology
- Innovative Imaging II
- System Modeling
- Cardiac Imaging
- X-ray Imaging
- Breast Imaging
- Tomosynethsis
- CT Systems
- Signal Corrections
- Cone Beam Reconstruction
- Advanced Reconstruction
- Poster Session: Radiography/Computed Tomography
- Poster Session: Breast Imaging
- Poster Session: Innovative Imaging
- Poster Session: Detectors
- Poster Session: Performance Assessment
- Poster Session: Signal Analysis
- Poster Session: Reconstruction
Front Matter: Volume 6510
Front Matter: Volume 6510
Show abstract
This PDF files contrains the front matter associated with SPIE Proceedings Volume 6510, including the Title Page, Copyright information, Table of Contents, Introduction (if any), and the Conference Committee listing.
Keynote
Statistical characterization of radiological images: basic principles and recent progress
Show abstract
This paper surveys our current understanding of the statistical properties of radiological images and their effect
on image quality. Attention is given to statistical descriptions needed to compute the performance of ideal or
ideal-linear observers on detection and estimation tasks. The effects of measurement noise, random objects and
random imaging system are analyzed by nested conditional averaging, leading to a three-term expansion of the
data covariance matrix. Characteristic functionals are introduced to account for the object statistics, and it is
shown how they can be used to compute the image statistics.
Intensity-modulated fluence patterns for task-specific imaging in cone-beam CT
Show abstract
Cone-beam CT images of a patient with a pre-defined distribution of noise in the image and dose to the patient can be
accomplished through the development of advanced compensation schemes. Such compensation schemes involve
delivery of x-ray fluence patterns that vary in intensity both across a single projection image (u,v) and for different
projection view angles (θ) and provide the ability to perform intensity-modulated cone-beam CT. Implementation of an
intensity-modulated cone-beam CT system for task-specific imaging has potential for tremendous reductions in patient
dose and x-ray scatter reaching the detector. Pursuing this advanced imaging technique requires detailed characterization
of the cone-beam CT platform. Determination of appropriately modulated fluence patterns relies on knowledge of
numerous properties of the imaging system, including the constraints imposed by the modulator, the magnitude of x-ray
scatter under different patient sizes and modulator positions, and properties of the detector. With an estimate of the
patient anatomy and knowledge of the imaging system, an iterative process can be used to determine modulated fluence
patterns corresponding to an image prescribed for the specific task and patient. Delivery of such modulated fluence
patterns provide a CBCT image tailored to a specific patient and imaging task offering the optimum balance between
image quality and patient dose. Specifically, arbitrary regions of interest requiring high image signal-to-noise ratio can
be generated through knowledgeable spatio-angular intensity modulation, allowing image quality to degrade in other
regions in order to minimize x-ray scatter and imaging dose.
Observer evaluation of a method for producing simulated mammograms
Show abstract
We have developed a method for producing simulated mammograms from high fidelity breast specimen radiographs. The method has the advantage of having access to all the truth information for the lesions. By modeling different parts of a screen-film system, we simulated the output of the system, and compared it to the real mammography images from the same samples. In this work we show how our simulation program produces realistic mammography images and also the observer study that tests how well the observers can distinguish the real and simulated images. Preliminary results from the ROC study show that the observers could not distinguish the two types of images very well.
Dual Energy
Dual-energy technique for digital flat-panel detectors without x-ray tube voltage switching
Show abstract
Dual-energy imaging increases the possibility of pulmonary nodule detection by reducing the bone structure
noise. The major problem of the dual-energy acquisition process with digital flat-panel detectors is the interval
of time between low-energy (LE) exposure and high-energy (HE) exposure. Due to misregistration between LE
and HE images, motion artifacts pollute the subtracted image. This paper presents a new acquisition approach
for dual-energy imaging developed in order to reduce this inter-exposure time. The idea is to keep the tube
voltage constant and to just switch a filter in front of the imaged object and thus to modulate the outgoing x-ray
spectrum.
The first part of this study presents how to optimize system parameters for the new acquisition protocol:
source voltage, dynamic filtration before the patient, exposure time for LE and HE acquisition. The tube load
is kept constant to focus the optimization study on the dose and the exposure time. A noise quality factor
(NQF) and a spectral quality factor (SQF) are used as criteria for optimization. The new approach system is
then compared to the state-of-the-art system with voltage switching between low and high energy. A filtering
algorithm of dual energy acquisitions enabling a significant noise reduction is presented. Performance between its
combination with the new acquisition protocol and the reference one are compared. For a limited noise quality
factor, three times faster acquisition time is obtained using the new system. Noise reduction techniques improve
the image SNR by 61% in the new system and only 32% in the reference system, without taking into account
the impact of better registration on the dual-energy image quality.
Development and implementation of a high-performance cardiac-gated dual-energy imaging system
Show abstract
Mounting evidence suggests that the superposition of anatomical clutter in a projection radiograph poses a major
impediment to the detectability of subtle lung nodules. Through decomposition of projections acquired at multiple kVp,
dual-energy (DE) imaging offers to dramatically improve lung nodule detectability and, in part through quantitation of
nodule calcification, increase specificity in nodule characterization. The development of a high-performance DE chest
imaging system is reported, with design and implementation guided by fundamental imaging performance metrics. A
diagnostic chest stand (Kodak RVG 5100 digital radiography system) provided the basic platform, modified to include:
(i) a filter wheel, (ii) a flat-panel detector (Trixell Pixium 4600), (iii) a computer control and monitoring system for
cardiac-gated acquisition, and (iv) DE image decomposition and display. Computational and experimental studies of
imaging performance guided optimization of key acquisition technique parameters, including: x-ray filtration, allocation
of dose between low- and high-energy projections, and kVp selection. A system for cardiac-gated acquisition was
developed, directing x-ray exposures to within the quiescent period of the heart cycle, thereby minimizing anatomical
misregistration. A research protocol including 200 patients imaged following lung nodule biopsy is underway, allowing
preclinical evaluation of DE imaging performance relative to conventional radiography and low-dose CT.
Dual-energy subtraction for contrast-enhanced digital breast tomosynthesis
Show abstract
We have developed a dual-energy subtraction technique for contrast-enhanced breast tomosynthesis. The imaging
system consists of 48 photon-counting linear detectors which are precisely aligned with the focal spot of the x-ray
source. The x-ray source and the digital detectors are translated across the breast in a continuous linear motion; each
linear detector collects an image at a distinct angle. A pre-collimator is positioned above the breast and defines 48 fan-shaped
beams, each aligned with a detector. Low- and high-energy images are acquired in a single scan; half of the
detectors capture a low-energy beam and half capture a high-energy beam, as alternating fan-beams are filtered to
emphasize low and high energies. Imaging was performed with a W-target at 45 and 49 kV. Phantom experiments and
theoretical modeling were conducted. Iodine images were produced with weighted logarithmic subtraction. The
optimal tissue cancellation factor, wt, was determined based on simultaneous preservation of the iodine signal and
suppression of simulated anatomic background. Optimal dose allocation between low- and high-energy images was
investigated. Mean glandular doses were restricted to ensure clinical relevance. Unlike other dual-energy approaches,
both spectra must have the same peak energy in this system design. We have observed that wt is mainly dependent on
filter combination and varies only slightly with kV and breast thickness, thus ensuring a robust clinical implementation.
Optimal performance is obtained when the dose fraction allocated to the high energy images ranges from 0.55 to 0.65.
Using elemental filters, we have been able to effectively suppress the anatomic background.
Image-domain material decomposition using photon-counting CT
Show abstract
Novel CdTe photon counting x-ray detectors (PCXDs) have been developed for very high count rates [1-4] suitable for
x-ray micro computed tomography (μCT) scanners. It counts photons within each of J energy bins. In this study, we
investigate use of the data in these energy bins for material decomposition using an image domain approach. In this
method, one image is reconstructed from projection data of each energy bin; thus, we have J images from J energy bins
that are associated with attenuation coefficients with a narrow energy width. We assume that the spread of energies in
each bin is small and thus that the attenuation can be modeled using an effective energy for each bin. This
approximation allows us to linearize the problem, thus simplify the inversion procedure. We then fit J attenuation
coefficients at each location x
by the energy-attenuation function [5] and obtain either (1) photoelectric and Compton
scattering components or (2) 2 or 3 basis-material components. We used computer simulations to evaluate this approach
generating projection data with three types of acquisition schemes: (A) five monochromatic energies; (B) five energy
bins with PCXD and an 80 kVp polychromatic x-ray spectrum; and (C) two kVp with an intensity integrating detector.
Total attenuation coefficients of reconstructed images and calculated effective atomic numbers were compared with
data published by National Institute of Standards and Technology (NIST). We developed a new materially defined
"SmileyZ" phantom to evaluate the accuracy of the material decomposition methods. Preliminary results showed that
material based 3-basis functions (bone, water and iodine) with PCXD with 5 energy bins was the most promising
approach for material decomposition.
Atomic number resolution for three spectral CT imaging systems
Show abstract
The material specificity of computed tomography is quantified using an experimental benchtop imaging system
and a physics-based system model. The apparatus is operated with different detector and system configurations each
giving X-ray energy spectral information but with different overlap among the energy-bin weightings and noise
statistics. Multislice, computed tomography sinograms are acquired using dual kVp, sequential source filters or a
detector with two scintillator/photodiodes layers. Basis-material and atomic number images are created by first
applying a material decomposition algorithm followed by filtered backprojection. CT imaging of phantom materials
with known elemental composition and density were used for model validation. X-ray scatter levels are measured with a
beam-blocking technique and the impact to material accuracy is quantified. The image noise is related to the intensity
and spectral characteristics of the X-ray source. For optimal energy separation adequate image noise is required. The
system must be optimized to deliver the appropriate high mA at both energies. The dual kVp method supports the
opportunity to separately engineer the photon flux at low and high kvp. As a result, an optimized system can achieve
superior material specificity in a system with limited acquisition time or dose. In contrast, the dual-layer and sequential
acquisition modes rely on a material absorption mechanism that yields weaker energy separation and lower overall
performance.
Investigation of the use of photon counting x-ray detectors with energy discrimination capability for material decomposition in micro-computed tomography
Show abstract
Recently developed solid-state detectors combined with high-speed ASICs that allow individual pixel pulse processing
may prove useful as detectors for small animal micro-computed tomography. One appealing feature of these photon-counting
x-ray detectors (PCXDs) is their ability to discriminate between photons with different energies and count
them in a small number (2-5) of energy windows. The data in these energy windows may be thought of as arising from
multiple simultaneous x-ray beams with individual energy spectra, and could thus potentially be used to perform
material composition analysis. The goal of this paper was to investigate the potential advantages of PCXDs with
multiple energy window counting capability as compared to traditional integrating detectors combined with acquisition
of images using x-ray beams with 2 different kVps. For the PCXDs, we investigated 3 potential sources of crosstalk:
scatter in the object and detector, limited energy resolution, and pulse piluep. Using Monte Carlo simulations, we
showed that scatter in the object and detector results in relatively little crosstalk between the data in the energy
windows. To study the effects of energy resolution and pulse-pileup, we performed simulations evaluating the accuracy
and precision of basis decomposition using a detector with 2 or 5 energy windows and a single kVp compared to an dual
kVp acquisitions with an integrating detector. We found that, for noisy data, the precision of estimating the thickness of
two basis materials for a range of material compositions was better for the single kVp multiple energy window
acquisitions compared to the dual kVp acquisitions with an integrating detector. The advantage of the multi-window
acquisition was somewhat reduced when the energy resolution was reduced to 10 keV and when pulse pileup was
included, but standard deviations of the estimated thicknesses remained better by more than a factor of 2.
Performance Assessment
A method to estimate the point response function of digital x-ray detectors from edge measurements
Show abstract
Currently, the most accurate measurement of the detector point response can be performed with the pinhole method. The small size of the pinhole however, severely reduces the x-ray intensity output, requiring long exposures, something that can potentially reduce the x-ray tube life-cycle. Even though deriving the 1D Line Response Function (LRF)of the detector using the edge method is much more effcient, the measurement process introduces a convolution with a line, in addition to the common pixel sampling, effectively broadening the LRF. We propose a practical method to recover the detector point response function by removing the effects of the line and the pixel from a set of Edge Response Function (ERF) measurements. We use the imaging equation to study the effects of the edge,line and pixel measurements, and derive an analytical formula for the recovered detector point response function based on a gaussian mixture model. The method allows for limited recovery of asymmetries in the detector response function. We verify the method with pinhole and edge measurements of a digital flat panel detector. Monte Carlo simulations are also performed, using the MANTIS x-ray and optical photon and electron transport simulation package, for comparison. We show that the standard LRF underestimates the detector when compared with the recovered response. Our simulation results suggest that both hole methods for estimating the detector response have limitations in that they cannot completely capture rotational asymmetries or other morphological details smaller than the detector pixel size.
The variability of software scoring of the CDMAM phantom associated with a limited number of images
Show abstract
Software scoring approaches provide an attractive alternative to human evaluation of CDMAM images from digital
mammography systems, particularly for annual quality control testing as recommended by the European Protocol for the
Quality Control of the Physical and Technical Aspects of Mammography Screening (EPQCM). Methods for correlating
CDCOM-based results with human observer performance have been proposed. A common feature of all methods is the
use of a small number (at most eight) of CDMAM images to evaluate the system. This study focuses on the potential
variability in the estimated system performance that is associated with these methods.
Sets of 36 CDMAM images were acquired under carefully controlled conditions from three different digital
mammography systems. The threshold visibility thickness (TVT) for each disk diameter was determined using
previously reported post-analysis methods from the CDCOM scorings for a randomly selected group of eight images for
one measurement trial. This random selection process was repeated 3000 times to estimate the variability in the
resulting TVT values for each disk diameter. The results from using different post-analysis methods, different random
selection strategies and different digital systems were compared. Additional variability of the 0.1 mm disk diameter was
explored by comparing the results from two different image data sets acquired under the same conditions from the same
system.
The magnitude and the type of error estimated for experimental data was explained through modeling. The modeled
results also suggest a limitation in the current phantom design for the 0.1 mm diameter disks. Through modeling, it was
also found that, because of the binomial statistic nature of the CDMAM test, the true variability of the test could be
underestimated by the commonly used method of random re-sampling.
Spatial resolution of x-ray tomosynthesis in relation to computed tomography for coronal/sagittal images of the knee
Show abstract
We consider in this report, a tomosynthesis (TS) system utilizing a recently developed Selenium flat panel detector (Shimadzu Safire II) capable of acquiring 30 pulsed frames per second (fps) with precise geometric registration to a moving gantry. The reconstructed TS resolution achieved with this system is reported in relation to the resolution obtained with modern volumetric computed tomography (vCT) systems. Computed tomography resolution is assessed using a sphere phantom. To describe the resolution of coronal/sagittal re-formatted planes, we average the transverse and axial response to express typical vCT performance as 1.00 mm (FWHM PlSF) and 0.83 cycles/mm (10% MTF). Tomosynthesis response is assessed using a wire phantom. To describe the resolution of the coronal/sagittal TS images, we use the FWHM of the LSF and the frequency at 20% of the peak spatial frequency response to express typical TS performance as 2.41 mm (FWHM LSF) and 2.62 cycles/mm (20% Peak). Cadaver images confirm that the TS resolution is about 3 times better than that for vCT systems. On the other hand, the TS slice sensitivity of 3 mm is about 3 times larger than vCT resolution (FWHM LSF).
Development of the 4D Phantom for patient-specific end-to-end radiation therapy QA
Show abstract
In many patients respiratory motion causes motion artifacts in CT images, thereby inhibiting precise treatment planning and lowering the ability to target radiation to tumors. The 4D Phantom, which includes a 3D stage and a 1D stage that each are capable of arbitrary motion and timing, was developed to serve as an end-to-end radiation therapy QA device that could be used throughout CT imaging, radiation therapy treatment planning, and radiation therapy delivery. The dynamic accuracy of the system was measured with a camera system. The positional error was found to be equally likely to occur in the positive and negative directions for each axis, and the stage was within 0.1 mm of the desired position 85% of the time. In an experiment designed to use the 4D Phantom's encoders to measure trial-to-trial precision of the system, the 4D Phantom reproduced the motion during variable bag ventilation of a transponder that had been bronchoscopically implanted in a canine lung. In this case, the encoder readout indicated that the stage was within 10 microns of the sent position 94% of the time and that the RMS error was 7 microns. Motion artifacts were clearly visible in 3D and respiratory-correlated (4D) CT scans of phantoms reproducing tissue motion. In 4D CT scans, apparent volume was found to be directly correlated to instantaneous velocity. The system is capable of reproducing individual patient-specific tissue trajectories with a high degree of accuracy and precision and will be useful for end-to-end radiation therapy QA.
Anisotropic point spread function of cone-beam computed tomography
Show abstract
Cone-beam computed tomography (CBCT) provides a real volume imaging modality, which can reconstruct a
digital volume with an isotropic grid resolution. However, the imaging resolution of a CBCT system is not identical
among spatial orientations, which manifests an anisotropic three-dimensional point spread function (3D PSF), or a PSF
ellipsoid. According to the CBCT imaging procedure, the anisotropic factors include the x-ray cone-beam quality, focal
spot size and shape, divergence projection angle, scanning orbit, and reconstruction algorithms. Since x-rays are
generated by electron bombardment on a slope target anode, the x-ray source is of finite size and non-circle shape,
which can neither be considered as a point nor a planar spot in strict sense. Due to non-uniform penetration depths in
the anode target, the wavefronts of the emanating x-rays assume non-spheric distributions, as described by the heel
effect, which becomes pronounced in a large cone-angle projection. The spatial blurring thereby is rather complicated.
In this paper, we will study the anisotropy of 3D PSF under diverse imaging factors, including the heel effect, the focal
spot size and shape, and the cone angle. Both computer simulation and phantom experiment on a flat-panel-detector
CBCT system will be reported.
Innovative Imaging I
Simultaneous x-rays/optical tomography of small animals
Show abstract
A small animal multimodality tomographer dedicated to the co-registration of fluorescence optical signal and X-rays measurements has been developed in our laboratory. The purpose of such a system is to offer the possibility to get in vivo anatomical and functional information at once. Moreover, anatomical measurements can be used as a regularization factor in order to get the reconstructions of the biodistribution of fluorochromes more accurate and to speed up the treatment.
The optical system is basically composed with a CW laser (Krypton, 752 nm) for an optimal excitation of Alexa-Fluor 750 fluorochromes, and a CCD camera coupled with a combination of filters for the fluorescence detection. The animal is placed inside a transparent tube filled with an index matching fluid. In order to perform multiple views of fluorescence data acquisitions, the cylinder is fixed to a rotating stage. The excitation beam is brought to the cylinder via two mirrors mounted on translation plates allowing a vertical scan. The optical data acquisitions are performed with a high sensitivity CCD camera. The X-ray generator and the X-ray detector have been placed perpendicularly to the optical chain.
A first study on phantoms was conducted to evaluate the feasibility, to test the linearity and the reproducibility, and to fix the parameters for the co-registration. These test experiments were reproduced by considering mice in the oesophagus of which thin glass tubes containing fluorochromes were inserted. Finally, the performance of the system was evaluated in vivo on mice bearing tumours in the lungs, tagged with Transferin-AlexaFluor 750.
Multi-source inverse geometry CT: a new system concept for x-ray computed tomography
Show abstract
Third-generation CT architectures are approaching fundamental limits. Spatial resolution is limited by the focal spot size and the detector cell size. Temporal resolution is limited by mechanical constraints on gantry rotation speed, and alternative geometries such as electron-beam CT and two-tube-two-detector CT come with severe tradeoffs in terms of image quality, dose-efficiency and complexity. Image noise is fundamentally linked to patient dose, and dose-efficiency is limited by finite detector efficiency and by limited spatio-temporal control over the X-ray flux. Finally, volumetric coverage is limited by detector size, scattered radiation, conebeam artifacts, Heel effect, and helical over-scan. We propose a new concept, multi-source inverse geometry CT, which allows CT to break through several of the above limitations. The proposed architecture has several advantages compared to third-generation CT: the detector is small and can have a high detection efficiency, the optical spot size is more consistent throughout the field-of-view, scatter is minimized even when eliminating the anti-scatter grid, the X-ray flux from each source can be modulated independently to achieve an optimal noise-dose tradeoff, and the geometry offers unlimited coverage without cone-beam artifacts. In this work we demonstrate the advantages of multi-source inverse geometry CT using computer simulations.
Compact multi-spectral imaging system for dermatology and neurosurgery
Show abstract
A compact multi-spectral imaging system is presented as diagnostic tool in dermatology and neurosurgery. Using an
electronically tunable filter, a sensitive high resolution digital camera, 140 spectral images from 400 nm up to 720 nm
are acquired in 40 s. Advanced image processing algorithms are used to enable interactive acquisition, viewing, image
registration and image analysis. Experiments in the department of dermatology and neurosurgery show that multispectral
imaging reveals much more detail than conventional medical photography or a surgical microscope, as images
can be reprocessed to enhance the view on e.g. tumor boundaries. Using a hardware-based interactive registration
algorithm, multi-spectral images can be aligned to correct for motion occurred during image acquisition or to compare
acquisitions from different moments in time. The system shows to be a powerful diagnostics tool for medical imaging in
the visual and near IR range.
Demonstration of three dimensional imaging of blood vessel using a no-moving parts electronic lens-based optical confocal microscope
Show abstract
To the best of our knowledge, for the first time, biological Three Dimensional (3-D) imaging has been achieved using an electronically controlled optical lens to accomplish no-moving parts depth section scanning in a modified commercial 3-D confocal microscope. Specifically, full 3-D views of a standard CDC blood vessel (enclosed in a glass slide) have been obtained using the modified confocal microscope operating at the red 633 nm laser wavelength.
Elemental spectrum of a mouse obtained via neutron stimulation
Show abstract
Several studies have shown that the concentration of certain elements may be a disease indicator. We are developing a spectroscopic imaging technique, Neutron Stimulated Emission Computed Tomography (NSECT), to non-invasively measure and image elemental concentrations within the body. The region of interest is interrogated via a beam of high-energy neutrons that excite elemental nuclei through inelastic scatter. These excited nuclei then relax by emitting characteristic gamma radiation. Acquiring the gamma energy spectrum in a tomographic geometry allows reconstruction of elemental concentration images. Our previous studies have demonstrated NSECT's ability to obtain spectra and images of known elements and phantoms, as well as, initial interrogations of biological tissue. Here, we describe the results obtained from NSECT interrogation of a fixed mouse specimen. The specimen was interrogated via a 5MeV neutron beam for 9.3 hours in order to ensure reasonable counting statistics. The gamma energy spectrum was obtained using two High-Purity Germanium (HPGe) clover detectors. A background spectrum was obtained by interrogating a specimen container containing 50mL of 0.9% NaCl solution. Several elements of biological interest including 12C, 40Ca, 31P, and 39K were identified with greater then 90% confidence. This interrogation demonstrates the feasibility of NSECT interrogation of small animals. Interrogation with a commercial neutron source that provides higher neutron flux and lower energy (~2.5MeV) neutrons would reduce scanning time and eliminate background from certain elements.
Initial experimentation with in-line holography x-ray phase-contrast imaging with an ultrafast laser-based x-ray source
Show abstract
We have investigated experimentally and theoretically the imaging performance of our newly constructed in-line
holography x-ray phase-contrast imaging system with an ultrafast laser-based x-ray source. Projection images of nylon
fibers with diameters in the 10-330 μm range were obtained using an ultrafast (100 Hz, 28 fs, 40 mJ) laser-based x-ray
source with Mo and Ta targets and Be filter, and Gaussian spatial-intensity distribution (FWHMS = 5 μm). A cooled
CCD camera (24 μm pitch) with a Gd2OS2 screen coupled via 1:1 optical taper was used (FWHMD = 50 μm). We have investigated nylon-fiber image quality vs. imaging setup geometry and x-ray spectra. The following parameters were
evaluated: contrast, signal-to-noise ratio (SNR), resolution, and sampling. In addition, we performed theoretical
simulation of image formation for the same objects but within a wide range of geometrical parameters. The rigorous
wave-optical formalism was used for modeling of the free-space propagation of x-rays from the object plane to the
detector, and the "projection approximation" was used. We found reasonable agreement between predictions of our
analytical model and the experiments. We conclude that: a) Optimum magnification maximizing contrast and SNR is
almost independent of the source-to-detector (R) distance and depends strongly on the diameter of the fiber. b) The
corresponding maximum values of the contrast and SNR are almost linear with respect to R; the optimum magnification
decreases with fiber diameter. c) The minimum diameter of fiber defines the minimum source-to-object distance R1 if R is fixed and the object is moved.
Detector Technology
A scanning system for intelligent imaging: I-ImaS
Show abstract
I-ImaS (Intelligent Imaging Sensors) is a European project aiming to produce adaptive x-ray imaging systems using Monolithic Active Pixel Sensors (MAPS) to create optimal diagnostic images. Initial systems concentrate on mammography and cephalography.
The on-chip intelligence available to MAPS technology will allow real-time analysis of data during image acquisition, giving the capability to build a truly adaptive imaging system with the potential to create images with maximum diagnostic information within given dose constraints.
In our system, the exposure in each image region is optimized and the beam intensity is a function not only of tissue thickness and attenuation, but also of local physical and statistical parameters found in the image itself. Using a linear array of detectors with on-chip intelligence, the system will perform an on-line analysis of the image during the scan and then will optimize the X-ray intensity in order to obtain the maximum diagnostic information from the region of interest while minimizing exposure of less important, or simply less dense, regions.
This paper summarizes the testing of the sensors and their electronics carried out using synchrotron radiation, x-ray sources and optical measurements.
The sensors are tiled to form a 1.5D linear array. These have been characterised and appropriate correction techniques formulated to take into account misalignments between individual sensors.
Full testing of the mammography and cephalography I-ImaS prototypes is now underway and the system intelligence is constantly being upgraded through iterative testing in order to obtain the optimal algorithms and settings.
Photon-counting gamma camera based on columnar CsI(Tl) optically coupled to a back-illuminated CCD
Show abstract
Recent advances have been made in a new class of CCD-based, single-photon-counting gamma-ray detectors
which offer sub-100 μm intrinsic resolutions.1-7 These detectors show great promise in small-animal SPECT
and molecular imaging and exist in a variety of cofigurations. Typically, a columnar CsI(Tl) scintillator or a
radiography screen (Gd2O2S:Tb) is imaged onto the CCD. Gamma-ray interactions are seen as clusters of signal
spread over multiple pixels. When the detector is operated in a charge-integration mode, signal spread across
pixels results in spatial-resolution degradation. However, if the detector is operated in photon-counting mode, the
gamma-ray interaction position can be estimated using either Anger (centroid) estimation or maximum-likelihood
position estimation resulting in a substantial improvement in spatial resolution.2 Due to the low-light-level nature
of the scintillation process, CCD-based gamma cameras implement an amplfication stage in the CCD via electron
multiplying (EMCCDs)8-10 or via an image intensfier prior to the optical path.1
We have applied ideas and techniques from previous systems to our high-resolution LumiSPECT detector.11, 12
LumiSPECT is a dual-modality optical/SPECT small-animal imaging system which was originally designed to
operate in charge-integration mode. It employs a cryogenically cooled, high-quantum-efficiency, back-illuminated
large-format CCD and operates in single-photon-counting mode without any intermediate amplfication process.
Operating in photon-counting mode, the detector has an intrinsic spatial resolution of 64 μm compared to 134
μm in integrating mode.
A wafer scale active pixel CMOS image sensor for generic x-ray radiology
Show abstract
This paper describes a CMOS Active Pixel Image Sensor developed for generic X-ray imaging systems using standard CMOS technology and an active pixel architecture featuring low noise and a high sensitivity. The image sensor has been manufactured in a standard 0.35 μm technology using 8" wafers. The resolution of the sensor is 3360x3348 pixels of 40x40 μm2 each. The diagonal of the sensor measures little over 190 mm. The paper discusses the floor planning, stitching diagram, and the electro-optical performance of the sensor that has been developed.
Quantitative exploration of performance enhancements offered by active matrix x-ray imagers fabricated on plastic substrates
Show abstract
Development of active matrix imagers fabricated on plastic substrates has become increasingly possible due to widespread efforts to develop the means to create inexpensive, very large area, flexible displays. In addition to benefits associated with cost, robustness, and weight, such novel x-ray imaging devices could provide significant performance improvements by virtue of substrates offering a thinner profile, lower density, and lower atomic number composition, as well as the ability to mechanically conform to non-planar geometries. One potential candidate for advantageous utilization of plastic substrates is that of a high resolution, indirect detection active matrix imager operated at mammographic x-ray energies. Such an imager, configured in an arc and operated in a back illumination geometry, could offer enhanced modulation transfer function and detective quantum efficiency by virtue of reduced optical spread in the scintillator relative to the optical sensor as well as reduction (or elimination) of oblique angles of incidence of primary x-rays. Motivated by such prospects, it is of interest to quantify the degree of performance improvement to be expected from such novel devices. In this paper, we describe a methodology under development by our collaboration to quantitatively examine the performance of indirect detection active matrix imagers incorporating plastic substrates. This methodology is based on the cascaded systems formalism with input provided by, among other things, Monte Carlo simulation of radiation and optical transport in the detector. Finally, illustrative examples of the use of this methodology are presented - although key input parameters are, as yet, insufficiently precise to allow the simulations to accurately depict the complete physical situation.
Dark current and DQE improvements of mercuric iodide medical imagers
Show abstract
A new TFT array has been developed specifically for mercuric iodide (HgI2) deposition. This new TFT array
combined with a modified HgI2 Physical Vapor Deposition (PVD) process provides less than 10 pA/mm2 dark current at
room temperature (22 °C) measured at 1 V/&mgr;m electrical field. This photoconductor (direct) imager was run at 10 fr/s
framerate and gave a measured sensitivity of 19 μC/(R*cm2) using a RQA5 radiation quality x-ray beam (70kVp x-ray
with 21 mm Al filtering). This sensitivity value is higher than the sensitivity reported by our group for any previous
HgI2 imagers.
MTF, NPS and DQE values were also evaluated on this 13 cm x 13 cm size imager with 127 μm pixel pitch.
The MTF value is higher than 40% at the Nyquist frequency (3.9 lp/mm). This is much better than the MTF of a 600
μm CsI scintillator/photodiode (indirect) imager, which is only 16% (Varian internal data) and it is similar to the MTF
value of the a-Se (another photoconductor) imagers. The first frame image lag is less than 8% when the imager was run
at a 10 fr/s framerate.
The low dark current and some noise reduction in the detector electronics, made it possible for the DQE to be
measured down to low fluoroscopic dose levels (< 4 μR/fr). The DQE(0) value is over 50% at a dose of 35 μR/fr and
still about 40% at 3.76 μR/fr.
The 270 μm thick PVD HgI2 layer only absorbs less than 75% of the ~51 keV mean energy X-ray photons (70
kVp RQA5 filtered beam). This means that if the thickness of the HgI2 layer is increased to 500 μm (increasing the
absorption up to over 90%) the DQE(0) should then increase to about 60- 65% (assuming everything else remains
unchanged). This value is close to the 65 - 70 % DQE(0), measured for the indirect (CsI) imagers at higher doses. Such
a high DQE value makes this material competitive both for fluoroscopic and for radiographic applications.
Innovative Imaging II
A sub-matrix method for extracting x-ray coherent scattering form factors from image plate data
Show abstract
The coherent scattering form factor defines a material's small angle x-ray scattering properties. These scattering properties can provide useful medical diagnostic information if properly utilized. Measurement of the coherent scattering form factor is difficult, requiring expensive equipment and long measurement times. We show that it is possible to measure the coherent scattering form factor using standard clinical equipment through a matrix equation. The matrix elements are constructed from knowledge of the input x-ray spectrum. In the ideal case, the form factor can be extracted from this equation by inverting the matrix. For typical x-ray spectra and form factors, however, the matrix tends to be poorly conditioned, leading to large errors upon inversion. We have developed a sub-matrix method that constructs a series of smaller, well-conditioned matrices that can be accurately inverted to give the required form factor. We show through numerical simulations that the sub-matrix method can accurately measure the form factor of common tissue materials. Root mean square deviations of 0.0502 and 0.0804 were calculated for the form factors of water and fat with 90 kV spectra and 0.2 mm of tungsten filtration. Over the measurement range, the form factors vary between approximately 0.5 and 2.5. We show that the optimal spectral shape when using the sub-matrix method is one that is strongly peaked at high energies and that using an improperly chosen spectrum can result in a significant loss of accuracy. We also demonstrate that the sub-matrix method is not readily applicable for measurements of strongly ordered materials.
Prism-array lenses for energy filtering in medical x-ray imaging
Show abstract
Conventional energy filters for x-ray imaging are based on absorbing
materials which attenuate low energy photons, sometimes combined
with an absorption edge, thus also discriminating towards photons of
higher energies. These filters are fairly inefficient, in particular
for photons of higher energies, and other methods for achieving a
narrower bandwidth have been proposed. Such methods include various
types of monochromators, based on for instance mosaic crystals or
refractive multi-prism x-ray lenses (MPL's). Prism-array lenses
(PAL's) are similar to MPL's, but are shorter, have larger
apertures, and higher transmission. A PAL consists of a number of
small prisms arranged in columns perpendicular to the optical axis.
The column height decreases along the optical axis so that the
projection of lens material is approximately linear with a Fresnel
phase-plate pattern superimposed on it. The focusing effect is one
dimensional, and the lens is chromatic. Hence, unwanted energies can
be blocked by placing a slit in the image plane of a desired energy.
We present the first experimental and theoretical results on an
energy filter based on a silicon PAL. The study includes an
evaluation of the spectral shaping properties of the filter as well
as a quantification of the achievable increase in dose efficiency
compared to standard methods. Previously, PAL's have been
investigated with synchrotron radiation, but in this study a medical
imaging setup, based on a regular x-ray tube, is considered.
Phase contrast mammography with synchrotron radiation: physical aspects of the clinical trial
Show abstract
Purpose: The first clinical facility for synchrotron radiation (SR) mammography is now operative at the SYRMEP beamline of ELETTRA, the SR facility in Trieste, Italy. The mammographic facility and the preliminary results of the clinical trial are presented in this contribution.
Method and Materials: The distance between the SR source and the patient is about 30 m; the main features of the X-ray beam are: monochromaticity at ~0.2% bandwith in the energy range 8-35 keV, photon flux of about 108 ph/(mm2 s) and dimensions of 21 cm x 3.5 mm at the compressed breast. An innovative dosimetric system allows the on-line dose control during the examination. The images are acquired by scanning the patient, in prone position, in front of the stationary laminar beam; the average scanning time is about 10 s. The detector is a screen film system; it is at ~2 m from the breast in order to fulfil the so-called Phase Contrast (PhC) requirements. The breast thickness and glandularity defines the optimal beam energy for each examination. The patients are enrolled by radiologists, after routine examinations, on the basis of BI-RADS classification, according the research program approved by the local Ethical Committee.
Results: This communication concerns the first 9 patients underwent the SR PhC mammography; the images match the quality obtained in previous in vitro studies. With reference to conventional mammography the diagnostic quality of the radiological images is better, without increasing the delivered dose to the patient.
Dual-energy contrast enhanced digital breast tomosynthesis: concept, method, and evaluation on phantoms
Show abstract
In this paper, we present the development of dual-energy Contrast-Enhanced Digital Breast Tomosynthesis
(CEDBT). A method to produce background clutter-free slices from a set of low and high-energy projections
is introduced, along with a scheme for the determination of the optimal low and high-energy techniques. Our
approach consists of a dual-energy recombination of the projections, with an algorithm that has proven its performance
in Contrast-Enhanced Digital Mammography1 (CEDM), followed by an iterative volume reconstruction.
The aim is to eliminate the anatomical background clutter and to reconstruct slices where the gray level is
proportional to the local iodine volumetric concentration. Optimization of the low and high-energy techniques
is performed by minimizing the total glandular dose to reach a target iodine Signal Difference to Noise Ratio
(SDNR) in the slices. In this study, we proved that this optimization could be done on the projections, by
consideration of the SDNR in the projections instead of the SDNR in the slices, and verified this with phantom
measurements. We also discuss some limitations of dual-energy CEDBT, due to the restricted angular range
for the projection views, and to the presence of scattered radiation. Experiments on textured phantoms with
iodine inserts were conducted to assess the performance of dual-energy CEDBT. Texture contrast was nearly
completely removed and the iodine signal was enhanced in the slices.
A prototype instrument for adaptive SPECT imaging
Show abstract
We have designed and constructed a small-animal adaptive SPECT imaging system as a prototype for quantifying
the potential benefit of adaptive SPECT imaging over the traditional fixed geometry approach. The optical
design of the system is based on filling the detector with the object for each viewing angle, maximizing the
sensitivity, and optimizing the resolution in the projection images. Additional feedback rules for determining the
optimal geometry of the system can be easily added to the existing control software. Preliminary data have been
taken of a phantom with a small, hot, offset lesion in a flat background in both adaptive and fixed geometry
modes. Comparison of the predicted system behavior with the actual system behavior is presented along with recommendations for system improvements.
Multiplexing radiography based on carbon nanotube field emission x-ray technology
Show abstract
State-of-the-art tomographic imaging technique is based upon of simple serial imaging scheme. The tomographic scanners collect the projection images sequentially in the time domain, by a step-and-shoot process using a single-pixel x-ray source. The inefficient serial data collection scheme severely limits the data collection speed, which is critical for imaging of objects in rapid motion such as for diagnosis of cardiovascular diseases, CT fluoroscopy, and airport luggage inspection. Further improvement of the speed demands an increasingly high x-ray peak workload and gantry rotation speed, both of which have approached the engineering limits. Multiplexing technique, which has been widely adopted in communication devices and in certain analytical instruments, holds the promise to significantly increase the data throughput. It however, has not been applied to x-ray radiography, mainly due to limitations of the current x-ray source technology. Here we report a method for frequency multiplexing radiography (FMR) based on the frequency multiplexing principle and the carbon nanotube field emission x-ray technology. We show the feasibility of multiplexing radiography that enables simultaneous collection of multiple projection images. It has the potential to significantly increase the imaging speed for tomographic imaging without compromising the imaging quality.
System Modeling
Scattered radiation in flat-detector based cone-beam CT: propagation of signal, contrast, and noise into reconstructed volumes
Show abstract
This paper presents a novel framework for the systematic assessment of the impact of scattered radiation in
.at-detector based cone-beam CT. While it is well known that scattered radiation causes three di.erent types of
artifacts in reconstructed images (inhomogeneity artifacts such as cupping and streaks, degradation of contrast,
and enhancement of noise), investigations in the literature quantify the impact of scatter mostly only in terms
of inhomogeneity artifacts, giving little insight, e.g., into the visibility of low contrast lesions. Therefore, for
this study a novel framework has been developed that in addition to normal reconstruction of the CT (HU)
number allows for reconstruction of voxelized expectation values of three additional important characteristics
of image quality: signal degradation, contrast reduction, and noise variances. The new framework has been
applied to projection data obtained with voxelized Monte-Carlo simulations of clinical CT data sets of high
spatial resolution. Using these data, the impact of scattered radiation was thoroughly studied for realistic and
clinically relevant patient geometries of the head, thorax, and pelvis region. By means of spatially resolved
reconstructions of contrast and noise propagation, the image quality of a scenario with using standard antiscatter
grids could be evaluated with great detail. Results show the spatially resolved contrast degradation and
the spatially resolved expected standard deviation of the noise at any position in the reconstructed object. The
new framework represents a general tool for analyzing image quality in reconstructed images.
Validation of simulated point response of columnar phosphor screens
Show abstract
Typical methods to measure the resolution properties of x-ray detectors use slit or edge devices. However,
complete models of imaging systems for system optimization require knowledge of the point-response function
of the detector. In this paper, we report on the experimental methods developed for the validation of the
point-response function of an indirect columnar CsI:Tl detector predicted by Monte Carlo using mantis. We
describe simulation results that replicate experimental resolution measurements using edge and pinhole devices.
The experimental setup consists of a high-resolution CCD camera with a 1-to-1fiber optic faceplate that allows
measurements for different scintillation screens. The results of these experiments and simulations constitute
a resource for the development and validation of the columnar models of phosphor screens proposed as part
of previous work with mantis. We compare experimental high-resolution pinhole responses of two different
CsI(Tl) screens to predictions from mantis. The simulated response matches reasonably well the measurements
at normal and off-normal x-ray incidence angle when a realistic pinhole is used in the simulation geometry. Our
results will be combined with results on Swank factors determined from Monte Carlo pulse-height spectra to
provide a comprehensive validation of the phosphor models, therefore allowing their use for in silico system
optimization.
Monte Carlo package for simulating radiographic images of realistic anthropomorphic phantoms described by triangle meshes
Show abstract
X-ray imaging system optimization increases the benefit-to-cost ratio by reducing the radiation dose to the patient while maximizing image quality. We present a new simulation tool for the generation of realistic medical x-ray images for assessment and optimization of complete imaging systems.
The Monte Carlo code simulates radiation transport physics using the subroutine package PENELOPE, which accurately simulates the transport of electrons and photons within the typical medical imaging energy range. The new code implements a novel object-oriented geometry package that allows simulations with homogeneous objects of arbitrary shapes described by triangle meshes. The flexibility of this code, which uses the industry standard PLY input-file format, allows the use of detailed anatomical models developed using computer-aided design tools applied to segmented CT and MRI data. The use of triangle meshes highly simplifies the ray-tracing algorithm without reducing the generality of the code, since most surface models can be tessellated into triangles while retaining their geometric details. Our algorithm incorporates an octree spatial data structure to sort the triangles and accelerate the simulation, reaching execution speeds comparable to the original quadric geometry model of PENELOPE. Coronary angiograms were simulated using a tessellated version of the NURBS-based Cardiac-Torso (NCAT) phantom. The phantom models 330 objects, comprised in total of 5 million triangles. The dose received by each organ and the contribution of the different scattering processes to the final image were studied in detail.
Noise transfer analysis of base material decomposition methods
Show abstract
A generalized method to evaluate the noise transfer properties of the base material decomposition has been developed. We apply the method to a typical dual-energy CT scan with energy weightings and doses of a 80kV / 140kV scan. For sets {P1, P2} of dual-energy projections with Pi = 10-4.5 ... 1, both the water and bone decomposition and the Compton and Photo Effect decomposition are analyzed. As a figure of merit we determine the noise amplification factors A1, A2. They are given by the ratio of the relative noise of the dual-energy projections B1, B2 to the relative noise of the combined projection data P. The B1, B2 and their variance are simulated by numerical inversion and integration. For the water and bone decomposition an average noise amplification of 3 to 5 is shown. For small contributions of one base material, the noise amplification becomes critically large. In this case the water and bone base material decomposition seems not to be usable for quantitative CT. The Compton and Photo effect decomposition are shown to be more robust in this respect. Physically, both coefficients can only reach zero simultaneously. The Compton coefficient has significantly better noise characteristics than the Photo Effect coefficient. For a partial region of the P1, P2 plane it shows better noise performance than the combined raw data P.
Cardiac Imaging
Motion compensated reconstructions of calcified coronary plaques in cardiac CT
Show abstract
In order to obtain motion-compensated reconstructions of calcified coronary plaques in cardiac CT, the dynamic
trajectory of the plaque must be known rather accurately. The purpose of this study is to evaluate whether
the dynamic trajectories of a plaque extracted from reconstructions provided by a previously developed tracking
algorithm can be used for obtaining motion-compensated reconstructions of this plaque. A single projection
dataset of the modified FORBILD phantom containing a calcified plaque undergoing continuous periodic motion
was acquired with a gantry rotation time of 0.4 s and a heart rate of 90 bpm. Three sets of phase-correlated 4D
ROI images centered on the calcified plaque (labeled G1, G2, and G3) were obtained from this dataset by varying
the numbers of data segments used for cardiac gating (N = 1, 2, 3) during the reconstruction steps of the tracking
algorithm. Dynamic trajectories from each of these datasets were calculated from edge-based segmentations of
these datasets. When compared to the true trajectory (labeled T), root-mean-square (RMS) values of position
for trajectories G1, G2, and G3 were 1.473 mm, 1.166 mm, and 0.736 mm, respectively. Trajectories G1, G2, G3,
and T then were used to obtain motion-compensated reconstructions MC1, MC2, MC3, and MCT, respectively,
at 6.25 ms time intervals over 2 cardiac cycles. The areas (number of pixels) of the plaque then were measured
at all time intervals for each set of reconstructions. When compared against areas obtained for MCT, RMS
values of areas for reconstructions MC1, MC2, and MC3 were 26.888, 12.384, and 4.837, respectively. On visual
inspection, MC3 also exhibited the least motion artifacts at most time intervals.
Proposed diagnostic reference levels for 3 common cardiac interventional procedures in Ireland
Show abstract
Radiation doses for 3 common types of cardiac radiological examinations where
investigated: coronary angiography (CA), percutaneous coronary intervention (PCI) and
pacemaker insertions (PPI). 22 cardiac imaging suites participated in the study. Radiation
dose was monitored for 1804 adult patients using dose area product (DAP) meters.
Operational and examination details such as cardiologist grade, patient details and
examination complexity were recorded for each examination. Both intra and inter-hospital variations where demonstrated by the results. Individual patient DAP values ranged from 136-23,101cGycm2, 475-41,038cGycm2 and 45- 17,192cGycm2 for CA, PCI and PPI respectively, with third quartile values of 4,173cGycm2, 8,836cGycm2 and 2,051cGycm2. Screening times varied from 0.22-27.6mins, 1.8-98mins and 0.33-54.5mins for CA, PCI and PPI respectively.
ECG-gated HYPR reconstruction for undersampled CT myocardial perfusion imaging
Show abstract
In this study we develop a novel ECG-gated method of HYPR (HighlY constrained backPRojection) CT reconstruction for low-dose myocardial perfusion imaging and present its first application in a porcine model. HYPR is a method of reconstructing time-resolved images from view-undersampled projection data. Scanning and reconstruction techniques were explored using x-ray projections from a 50 sec contrast-enhanced axial scan of a 47 kg swine on a 64-slice MDCT system. Scans were generated with view undersampling factors from 2 to 10. A HYPR reconstruction algorithm was developed in which a fully-sampled composite image is generated from views collected from multiple cardiac cycles within a diastolic window. A time frame image for a heartbeat was produced by modifying the composite with projections from the cycle of interest. Heart rate variations were handled by automatically selecting cardiac window size and number of cycles per composite within defined limits. Cardiac window size averaged 35% of the R-R interval for 2x undersampling and increased to 64% R-R using 10x undersampling. The selected window size and cycles per composite was sensitive to synchrony between heart rate, gantry rate, and the view undersampling pattern. Temporal dynamics and perfusion metrics measured in conventional short-scan (FBP) images were well-reproduced in the undersampled HYPR time series. Mean transit times determined from HYPR myocardial time-density curves agreed to within 8% with the FBP results. The results indicate potential for an order of magnitude reduction in dose requirement per image in cardiac perfusion CT via undersampled scanning and ECG-gated HYPR reconstruction.
Cardiac C-arm CT: 4D non-model based heart motion estimation and its application
Show abstract
The combination of real-time fluoroscopy and 3D cardiac imaging on the same C-arm system is a promising technique
that might improve therapy planning, guiding, and monitoring in the interventional suite. In principal, to reconstruct a 3D
image of the beating heart at a particular cardiac phase, a complete set of X-ray projection data representing that phase is
required. One approximate approach is the retrospectively ECG-gated FDK reconstruction (RG-FDK). From the acquired
data set of Ns multiple C-arm sweeps, those projection images which are acquired closest in time to the desired cardiac
phase are retrospectively selected. However, this approach uses only 1/
Ns
of the obtained data. Our goal is to utilize data from
other cardiac phases as well. In order to minimize blurring and motion artifacts, cardiac motion has to be compensated for,
which can be achieved using a temporally dependent spatial 3D warping of the filtered-backprojections. In this work we
investigate the computation of the 4D heart motion based on prior reconstructions of several cardiac phases using RG-FDK.
A 4D motion estimation framework is presented using standard fast non-rigid registration. A smooth 4D motion vector
field (MVF) represents the relative deformation compared to a reference cardiac phase. A 4D deformation regridding by
adaptive supersampling allows selecting any reference phase independently of the set of phases used in the RG-FDK for
a motion corrected reconstruction. Initial promising results from in vivo experiments are shown. The subjects individual
4D cardiac MVF could be computed from only three RG-FDK image volumes. In addition, all acquired projection data
were motion corrected and subsequently used for image reconstruction to improve the signal-to-noise ratio compared to
RG-FDK.
Image-domain motion compensated time resolved 4D cardiac CT
Show abstract
Two major problems with the current electrocardiogram-gated cardiac computed tomography (CT) imaging technique
are a large patient radiation dose (10-15 mSv) and insufficient temporal resolution (83-165 ms). Our long-term goal is to
develop new time resolved and low dose cardiac CT imaging techniques that consist of image reconstruction algorithms
and estimation methods of the time-dependent motion vector field (MVF) of the heart from the acquired CT data.
Toward this goal, we developed a method that estimates the 2D components of the MVF from a sequence of cardiac CT
images and used it to "reconstruct" cardiac images at rapidly moving phases. First, two sharp image frames per heart
beat (cycle) obtained at slow motion phases (i.e., mid-diastole and end-systole) were chosen. Nodes were coarsely
placed among images; and the temporal motion of each node was modeled by B-splines. Our cost function consisted of
3 terms: mean-squared-error with the block-matching, and smoothness constraints in space and time. The time-dependent
MVF was estimated by minimizing the cost function. We then warped images at slow motion phases using
the estimated vector fields to "reconstruct" images at rapidly moving phase. The warping algorithm was evaluated using
true time-dependent motion vector fields and images both provided by the NCAT phantom program. Preliminary results
from ongoing quantitative and qualitative evaluation using the 4D NCAT phantom and patient data are encouraging.
Major motion artifact is much reduced. We conclude the new image-based motion estimation technique is an important
step toward the development of the new cardiac CT imaging techniques.
X-ray Imaging
Configuration of AEC kVp dependence for digital radiography systems
Show abstract
As digital radiography (DR) systems are being increasingly adopted for clinical applications, automatic exposure control (AEC) has remained a critically important component. Unlike traditional screen/film systems, however, there does not currently exist a widely accepted AEC calibration criterion for DR systems. This is due mainly to the signal response characteristics and wide dynamic range of a DR detector, which are inherently different from those of a screen/film system. Consequently, the AEC cutoff dose and its dependence on the kVp selection (i.e., kVp compensation) should be calibrated differently for DR systems. In this paper, we have investigated three possible schemes to set up the AEC compensation based on a constant response of the detector in terms of the signal, receptor dose, or signal-to-noise-ratio (SNR) respectively. The results for each of the setup schemes were evaluated on four different DR detectors (Gd2O2S, CsI(Tl), a-Se, or BaFBrI as x-ray absorption material) based on the measured signal and noise response of the detector under the ISO beam conditions (ISO 6236-1). The results showed that all three setup schemes produced similar results for clinical beams above 70 kVp. Significant differences were observed only at lower kVp (≤60) beams. In addition, schemes of constant signal and constant SNR produced similar results with the only exception for the a-Se detector at low kVp (≤60) beam. These results indicate that the choice of the kVp schemes would be important only for low kVp exams.
The x-ray light valve: a low-cost, digital radiographic imaging system-spatial resolution
Show abstract
In recent years, new x-ray radiographic systems based on large area flat panel technology have revolutionized our
capability to produce digital x-ray radiographic images. However, these active matrix flat panel imagers (AMFPIs) are
extraordinarily expensive compared to the systems they are replacing. Thus there is a need for a low cost digital imaging
system for general applications in radiology. Different approaches have been considered to make lower cost, integrated
x-ray imaging devices for digital radiography, including: scanned projection x-ray, an integrated approach based on
computed radiography technology and optically demagnified x-ray screen/CCD systems. These approaches suffer from
either high cost or high mechanical complexity and do not have the image quality of AMFPIs. We have identified a new
approach - the X-ray Light Valve (XLV). The XLV has the potential to achieve the immediate readout in an integrated
system with image quality comparable to AMFPIs. The XLV concept combines three well-established and hence lowcost
technologies: an amorphous selenium (a-Se) layer to convert x-rays to image charge, a liquid crystal (LC) cell as an
analog display, and an optical scanner for image digitization. Here we investigate the spatial resolution possible with
XLV systems. Both a-Se and LC cells have both been shown separately to have inherently very high spatial resolution.
Due to the close electrostatic coupling in the XLV, it can be expected that the spatial resolution of this system will also
be very high. A prototype XLV was made and a typical office scanner was used for image digitization. The Modulation
Transfer Function was measured and the limiting factor was seen to be the optical scanner. However, even with this
limitation the XLV system is able to meet or exceed the resolution requirements for chest radiography.
Development of a portable instrument for automated measurements of the detective quantum efficiency of x-ray detectors
Show abstract
The scientific community has generally adopted use of the modulation transfer function (MTF) and detective
quantum efficiency (DQE) as primary measures of performance of radiographic detectors. However, measurement
of these parameters is generally restricted to experts in laboratory environments due to the required x-ray
physics knowledge, specialized instrumentation and computational analyses. We have developed a prototype
instrument that automates both the physical measurement and subsequent image analysis to determine the
MTF, noise power spectrum (NPS) and DQE of radiographic and mammographic systems. The instrument
is placed in the x-ray path directly in front of the detector. A series of images are acquired, saved in "raw"
DICOM format and then used to determine the MTF (using the slanted-edge method) and NPS. The number of
incident quanta is calculated from measurements of the incident exposure including corrections for air temperature
and pressure and ionization chamber spectral response. The primary sources of error are backscatter from
the detector and scatter generated within the instrument. These have been minimized to achieve an incident
exposure measurement within 2% of a calibrated electrometer and chamber in free space. The MTF and DQE
of a commercial CsI-based flat-panel detector were measured over a range of incident exposures from 20 uR
to 20 mR per image. Results agreed with both our own laboratory measurements and previously published
measurements performed elsewhere with a similar detector within 2% for the MTF and 5% for the DQE. A
complete DQE analysis of a clinical digital flat-panel detector is completed in 30 minutes and requires no system
modifications.
Multimode C-arm fluoroscopy, tomosynthesis, and cone-beam CT for image-guided interventions: from proof of principle to patient protocols
Show abstract
High-performance intraoperative imaging is essential to an ever-expanding scope of therapeutic procedures ranging from
tumor surgery to interventional radiology. The need for precise visualization of bony and soft-tissue structures with
minimal obstruction to the therapy setup presents challenges and opportunities in the development of novel imaging
technologies specifically for image-guided procedures. Over the past ~5 years, a mobile C-arm has been modified in
collaboration with Siemens Medical Solutions for 3D imaging. Based upon a Siemens PowerMobil, the device includes:
a flat-panel detector (Varian PaxScan 4030CB); a motorized orbit; a system for geometric calibration; integration with
real-time tracking and navigation (NDI Polaris); and a computer control system for multi-mode fluoroscopy,
tomosynthesis, and cone-beam CT. Investigation of 3D imaging performance (noise-equivalent quanta), image quality
(human observer studies), and image artifacts (scatter, truncation, and cone-beam artifacts) has driven the development
of imaging techniques appropriate to a host of image-guided interventions. Multi-mode functionality presents a valuable
spectrum of acquisition techniques: i.) fluoroscopy for real-time 2D guidance; ii.) limited-angle tomosynthesis for fast
3D imaging (e.g., ~10 sec acquisition of coronal slices containing the surgical target); and iii.) fully 3D cone-beam CT
(e.g., ~30-60 sec acquisition providing bony and soft-tissue visualization across the field of view). Phantom and cadaver
studies clearly indicate the potential for improved surgical performance - up to a factor of 2 increase in challenging
surgical target excisions. The C-arm system is currently being deployed in patient protocols ranging from brachytherapy
to chest, breast, spine, and head and neck surgery.
Optimization of a CR system comprising line-scanning and needle image plate technology with respect to examinations of extremities
Show abstract
A CR system based on line-scanning and needle image plate technologies (CsBr), was modified by means of an improved line light source for stimulation and a high aperture line optics for light collection. For comparison, two types of needle image plates were used, i.e. standard and high resolution plates. The image quality performance was investigated in terms of modulation transfer function, detective quantum efficiency and visual impression of both anthropomorphic and technical phantom images, with respect to application for X-ray examinations of extremities.
MTF and DQE at beam quality RQA 3 were clearly improved by the modified scanning components. Further improvement especially with respect to MTF at spatial frequencies beyond 2 LP/mm was obtained via the high resolution image plate, but at reduced low frequency DQE values. The visual inspection of images of a high contrast resolution pattern and an anthropomorphic hand phantom revealed a clear preference for the modified system. The use of the high resolution image plate alone resulted in images with higher sharpness but increased noise. The best performance - at constant X-ray exposure dose - was delivered by the combination of improved optics components with the adapted needle image plate.
Progress in electron-multiplying CCD (EMCCD) based high-resolution high-sensitivity x-ray detector for fluoroscopy and radiography
Show abstract
A new high-resolution, high-sensitivity, low-noise x-ray detector based on EMCCDs has been developed. The EMCCD detector module consists of a 1kx1k, 8μm pixel EMCCD camera coupled to a CsI(Tl) scintillating phosphor via a fiber optic taper (FOT). Multiple modules can be used to provide the desired field-of-view (FOV). The detector is capable of acquisitions over 30fps. The EMCCD's variable gain of up to 2000x for the pixel signal enables high sensitivity for fluoroscopic applications. With a 3:1 FOT, the detector can operate with a 144μm effective pixel size, comparable to current flat-panel detectors. Higher resolutions of 96 and 48μm pixel size can also be achieved with various binning modes. The detector MTFs and DQEs were calculated using a linear-systems analysis. The zero frequency DQE was calculated to be 59% at 74 kVp. The DQE for the 144μm pixel size was shown to exhibit quantum-noise limited behavior down to ~0.1μR using a conservative 30x gain. At this low exposure, gains above 30x showed limited improvements in DQE suggesting such increased gains may not be necessary. For operation down to 48µm pixel sizes, the detector instrumentation noise equivalent exposure (INEE), defined as the exposure where the instrumentation noise equals the quantum-noise, was <0.1μR for a 20x gain. This new technology may provide improvements over current flat-panel detectors for applications such as fluoroscopy and angiography requiring high frame rates, resolution, dynamic range and sensitivity while maintaining essentially no lag and very low INEE. Initial images from a prototype detector are also presented.
Breast Imaging
A contrast-detail comparison of computed mammotomography and digital mammography
Show abstract
We use a contrast-detail observer study to compare performance of a novel 3D computed mammotomography (CmT)
system with a commercially developed full-field digital mammography (FFDM) system. A contrast-detail phantom
comprised of uniform acrylic spheres of various diameters was developed and placed in a variety of mediums
including uniform water (simulating low contrast lesions within a uniform background), water and acrylic yarn
(simulating low contrast lesions with over/under-lying structure), oil only (simulating higher contrast lesions in a
uniform background), and oil and acrylic yarn (simulating higher contrast lesions with over/under-lying structure).
For CmT, the phantom was placed in a 14.6 cm diameter uncompressed breast phantom and projections acquired using
a simple circular orbit, W-target tube, 60 kVp tube potential, 0.05 cm Ce filtration, 4 mAs per projection, and a CsI(Tl)
digital x-ray detector. Reconstructions used an iterative OSTR algorithm. For FFDM, the phantom was placed in a
5.3-cm-thick compressed breast phantom. Single CC-view mammograms were acquired using a clinical W-target tube
with 50 um Rh filtration, 28 kVp, photo-timed mAs per our clinical mammography operation, and a Selenium-based
flat-panel detector (Mammomat Novation, Siemens). Six observers evaluated the images in terms of the number of
detectable spheres. FFDM performed significantly better for the low contrast lesions in uniform water background
(p<0.05). However, CmT performed significantly better for all other cases (p<0.05). Results indicate that CmT shows
significant advantage in soft tissue detection over FFDM in otherwise low contrast dense breasts.
Evaluation of lesion distortion at various CT system tilts in the development of a hybrid system for dedicated mammotomography
Show abstract
A hybrid SPECT-CT system for dedicated 3D breast imaging (mammotomography) is currently under development. Each imaging system will be placed on top of a single rotation stage and moved in unison azimuthally, with the SPECT system additionally capable of polar and radial motions. In this initial prototype, the CT system will initially be positioned at a fixed polar tilt. Using a phantom with three tungsten wires, the MTF of the CT system was measured in 3D for different CT system tilts. A phantom with uniformly arranged 0.5cm diameter acrylic spheres was suspended in air in the CT field of view, and also placed at multiple locations and orientations inside an oil-filled breast phantom to evaluate the effect of CT system tilt on lesion visibility and distortion. Projection images were collected using various simple circular orbits with fixed polar tilts ranging between ±15°, and complex 3D saddle trajectories including combined polar and azimuthal motions at maximum polar tilt angles. Reconstructions were performed using an iterative reconstruction algorithm on 4x4 binned projection images with 0.508mm3 voxels. There was minor variation in the MTF in the imaged volume for the CT system at all trajectories, potentially due to the use of an iterative reconstruction algorithm. Results from the spherical cross phantoms indicated that there was more reconstruction inaccuracy and geometric distortion in the reconstructed slices with simple circular orbits with fixed tilt in contrast to complex 3D trajectories. Line profiles further showed a cupping artifact in planes farther away from the flat plane of the x-ray cone beam placed at different tilts. However, this cupping artifact was not seen for images acquired with complex 3D trajectories. This indicated that cupping artifacts can also be caused by undersampled cone beam data. These findings generally indicate that despite insufficient sampling with the cone beam imaging geometry, it is possible to place the CT system at a stationary polar tilt with the CT tube positioned upward such that a patient can be comfortably placed above the system and allow complete sampling near the top of the pendant, uncompressed breast and chest wall. However, a complex 3D trajectory allows for more complete sampling of the entire image volume.
Performance characterization of a volumetric breast ultrasound scanner
Show abstract
The purpose of this project was to design, build, and characterize the performance of a volume breast ultrasound (VBUS) scanner that images the pendant breast. VBUS scanner design includes a: 1) clinical ultrasound scanner and transducer; 2) scanning table with a hole for the pendant breast; 3) rotational gantry; 4) probe mounting assembly; 5) compressionless breast stabilization device; 6) acquisition, control, reconstruction, and display software. Performance assessment characterized a variety of parameters including: spatial resolution, uniformity, and distortion using high and low contrast test objects in both horizontal and vertical scanning modes. The VBUS scanner modules have been constructed and initial performance evaluated. Approximately 300 scans are acquired over 360 degrees in 18 seconds. Reconstruction requires 25 ms per slice. Test object images depicted hyper- and hypo-echoic masses and demonstrated good resolution, soft tissue contrast and reduced speckle compared to conventional US images. Although scanner performance is satisfactory, additional developments including reduced spacing transducer - scanned object spacing and corrections for sound velocity and aberrations will improve operation. Future work will continue system optimization.
Investigating novel patient bed designs for use in a hybrid dual-modality dedicated 3D breast imaging system
Show abstract
A hybrid SPECT-CT system for dedicated 3D breast cancer imaging (mammotomography) is in development. Using
complex 3D imaging acquisition trajectories, the versatile integrated system will be capable of contouring and imaging
an uncompressed breast suspended in a 3D volume located below a radio-opaque patient bed, providing co-registered
volumetric anatomical and functional information. This study examines tradeoffs involved in the design of the patient
bed to satisfy concomitant and competing technical and ergonomic requirements specific to this imaging paradigm. The
complementary source-detector arrangement of the CT system is geometrically more restrictive than that of the single
detector SPECT system. Additionally, the compact dimensions and size of the CT system components (primarily the x-ray
tube) are key constraints on the bed design and so the focus is concentrated there. Using computer-aided design
software, several design geometry options are examined to simultaneously consider and optimize the following
parameters: image magnification, imaged breast volume, azimuthal imaging span, and patient comfort. Several CT
system source to image distances are examined (55-80cm), as well as axial patient tilt up to 35°. An optimal patient bed
design for a completely under-bed hybrid imaging system was determined. A 60cm SID, magnification factor of ~1.5,
and patient bed angled at ~15° provided the optimal dimensions. Additional bed dimensions allow the CT projection
beam to nearly entirely image the chest wall, however at the cost of reduced angular sampling for CT. Acquired x-ray
mammotomographic image data is used to assess the feasibility of this reduced angle acquisition approach.
Tomosynethsis
Methodology of NEQ (f) analysis for optimization and comparison of digital breast tomosynthesis acquisition techniques and reconstruction algorithms
Show abstract
As a new three-dimensional imaging technique, digital breast tomosynthesis allows the reconstruction of an arbitrary
set of planes in the breast from a limited-angle series of projection images. Though several tomosynthesis algorithms
have been proposed, no complete optimization and comparison of different tomosynthesis acquisition techniques for
available methods has been conducted as of yet. This paper represents a methodology of noise-equivalent quanta
NEQ (f) analysis to optimize and compare the efficacy of tomosynthesis algorithms and imaging acquisition
techniques for digital breast tomosynthesis. It combines the modulation transfer function (MTF) of system signal
performance and the noise power spectrum (NPS) of noise characteristics. It enables one to evaluate the
performance of different acquisition parameters and algorithms for comparison and optimization purposes. An
example of this methodology was evaluated on a selenium-based direct-conversion flat-panel Siemens Mammomat
Novation prototype system. An edge method was used to measure the presampled MTF of the detector. The MTF
associated with the reconstruction algorithm and specific acquisition technique was investigated by calculating the
Fourier Transform of simulated impulse responses. Flat field tomosynthesis projection sequences were acquired and
then reconstructed. A mean-subtracted NPS on the reconstructed plane was studied to remove fixed pattern noise.
An example of the application of this methodology was illustrated in this paper using a point-by-point Back
Projection correction (BP) reconstruction algorithm and an acquisition technique of 25 projections with 25 degrees
total angular tube movement.
Improved in-plane visibility of tumors using breast tomosynthesis
Show abstract
The purpose of this work was to evaluate and compare the visibility of tumors in digital mammography (DM) and breast tomosynthesis (BT) images. Images of the same women were acquired on both a DM system (Mammomat Novation, Siemens) and a BT prototype system adapted from the same type of DM system. Simulated 3D tumors (average dimension: 8.4 mm x 6.6 mm x 5 mm) were projected and added to each DM image as well as each BT projection image prior to 3D reconstruction. The same beam quality and approximately the same total absorbed dose were used for each breast image acquisition on both systems. Two simulated tumors were added to each of thirty breast scans, yielding sixty cases. A series of 4-alternative forced choice (4-AFC) human observer performance experiments were conducted in order to determine what projected tumor signal intensity in the DM images would be needed to achieve the same detectability as in the reconstructed BT images. Nine observers participated. For the BT experiment, when the tumor signal intensity on the central projection was 0.010 the mean percent of correct responses (PC) was measured to be 81.5%, which converted to a detectability index value (d') of 1.96. For the DM experiments, the same detectability was achieved at a signal intensity determined to be 0.038. Equivalent tumor detection in BT images were thus achieved at around four times less projected signal intensity than in DM images, indicating that the use of BT may lead to earlier detection of breast cancer.
A mathematical model approach toward combining information from multiple image projections of the same patient
Show abstract
In this study, we used a mathematical observer model to combine information obtained from multiple angular projections of the same breast to determine the overall detection performance of a multi-projection breast imaging system in detectability of a simulated mass. 82 subjects participated in the study and 25 angular projections of each breast were acquired. Projections from a simulated 3 mm 3-D lesion were added to the projection images. The lesion was assumed to be embedded in the compressed breast at a distance of 3 cm from the detector. Hotelling observer with Laguerre-Gauss channels (LG CHO) was applied to each image. Detectability was analyzed in terms of ROC curves and the area under ROC curves (AUC). The critical question studied is how to best integrate the individual decision variables across multiple (correlated) views. Towards that end, three different methods were investigated. Specifically, 1) ROCs from different projections were simply averaged; 2) the test statistics from different projections were averaged; and 3) a Bayesian decision fusion rule was used. Finally, AUC of the combined ROC was used as a parameter to optimize the acquisition parameters to maximize the performance of the system. It was found that the Bayesian decision fusion technique performs better than the other two techniques and likely offers the best approximation of the diagnostic process. Furthermore, if the total dose level is held constant at 1/25th of dual-view mammographic screening dose, the highest detectability performance is observed when considering only two projections spread along an angular span of 11.4°.
Effect of acquisition parameters on image quality in digital tomosynthesis
Show abstract
Digital tomosynthesis (DTS) is emerging as an advanced imaging technique that enables volumetric slice imaging with a detector typically used for projection radiography. An understanding of the interactions between DTS acquisition parameters and characteristics of the reconstructed slice images is required for optimizing the acquisition protocols of various clinical applications. This paper presents our investigation of the effects and interactions of acquisition parameters, including sweep angle, number of projections, and dose, on clinically relevant image-quality metrics. Metrics included the image characteristics of in-slice resolution, depth resolution, image noise level, and presence of ripple.
Phantom experiments were performed to characterize the relationship between the acquisition parameters and image quality. Results showed that the depth resolution was mainly dependent on sweep angle. Visibility of ripple was determined by the projection density (number of projections divided by sweep angle), as well as properties of the imaged object. Image noise was primarily dependent on total dose and not significantly affected by the number of projections. These experimental and theoretical results were confirmed using anthropomorphic phantoms and also used to develop clinical acquisition protocols. Assessment of phantom and clinical images obtained with these protocols revealed that the use of acquisition protocols optimized for a given clinical exam enables rapid, low-dose, high quality DTS imaging for diverse clinical applications including abdomen, hand, shoulder, spine, and chest.
We conclude that DTS acquisition parameters have a significant effect on image quality and should be tailored for the imaged anatomy and desired clinical application. Relationships developed in this work will guide the selection of acquisition protocols to improve image quality and clinical utility of DTS for a wide variety of clinical exams.
Optimization of detector operation and imaging geometry for breast tomosynthesis
Show abstract
In breast tomosynthesis there are tradeoffs between resolution, noise and acquisition speed for a given glandular dose. The purpose of the present work is to investigate the dependence of tomosynthesis imaging performance on system configuration, which includes detector operational modes and image acquisition geometry. A prototype Siemens breast tomosynthesis system with maximum angular range of +/- 25 degrees was used in our investigation. The system was equipped with an amorphous selenium (a-Se) full field digital mammography detector with pixel size of 85µm. The detector can be read out with full resolution or 2x1 binning (binning in the tube travel direction), which increases the image readout rate and decreases the degradation effect of electronic noise. The total number of views can be varied from 11 to 49, and filtered back projection (FBP) method was used to reconstruct the tomosynthesis images. We investigated the effects of detector operational modes (binning) and imaging geometry (view angle and number) on temporal performance and spatial resolution of the projection images. The focal spot blur due to continuous tube travel was measured for different acquisition geometry, and its effect on in-plane presampling modulation transfer function (MTF) was compared to that due to pixel binning. A three-dimensional cascaded linear system model was developed for tomosynthesis to predict the 3D MTF, NPS and DQE. The results were compared with experimental measurements, and reasonable agreement was achieved. The understanding of the relationship between the 3D and projection image quality will lead to optimization of the x-ray spectrum, imaging geometry and reconstruction filters for digital breast tomosynthesis.
Circular tomosynthesis implemented with a clinical interventional flat-panel based C-arm: initial performance study
Show abstract
There exists a strong desire for a platform in which researchers may investigate planar tomosynthesis (i.e. all
source positions reside in a single plane that is parallel to the reconstructed image planes) trajectories directly
on an interventional C-arm system. In this work we describe an experimental system designed to accomplish this
aim, as well as the potential of this system for testing multiple aspects of the tomosynthetic image acquisition
process. The system enables one to evaluate the effect of the physical imaging parameters on the image quality,
as well as the effect of the reconstruction algorithm utilized. The experimental data collection for this work is
from the Innova 4100 (Flat-panel based interventional C-arm system manufactured by GE Healthcare). The
system is calibrated using a phantom with known geometrical placement of multiple small metallic spheres.
Initial performance was assessed with three physical phantoms and performance was assessed by varying: the
reconstruction algorithm (backprojection, filtered backprojection), the half tomographic angle (15°, 25°, 35°),
and the angular sampling (20,40,80 views / acquisition). Initial results demonstrate the ability to well differentiate
simulated vessels separated by 1 cm, even with the modest half tomographic angle of 15° and modest sampling
of 20 views/acquisition.
CT Systems
Perfusion analysis using a wide coverage flat-panel volume CT: feasibility study
M. Grasruck,
R. Gupta M.D.,
B. Reichardt,
et al.
Show abstract
We developed a Flat-panel detector based Volume CT (VCT) prototype scanner with large z-coverage. In that prototype scanner a Varian 4030CB a-Si flat-panel detector was mounted in a multi slice CT-gantry (Siemens Medical Solutions) which provides a 25 cm field of view with 18 cm z-coverage at isocenter. The large volume covered in one rotation can be used for visualization of complete organs of small animals, e.g. rabbits. By implementing a mode with continuous scanning, we are able to reconstruct the complete volume at any point in time during the propagation of a contrast bolus. Multiple volumetric reconstructions over time elucidate the first pass dynamics of a bolus of contrast resulting in 4-D angiography and potentially allowing whole organ perfusion analysis.
We studied to which extent pixel based permeability and blood volume calculation with a modified Patlak approach was possible. Experimental validation was performed by imaging evolution of contrast bolus in New Zealand rabbits. Despite the short circulation time of a rabbit, the temporal resolution was sufficient to visually resolve various phases of the first pass of the contrast bolus. Perfusion imaging required substantial spatial smoothing but allowed a qualitative discrimination of different types of parenchyma in brain and liver. If a true quantitative analysis is possible, requires further studies.
Evaluation of the spatial resolution of a dedicated breast CT system using computer simulation
Show abstract
Spatial resolution is one of the most crucial parameters for an imaging system. The modulation transfer function (MTF) was physically measured using wire images on a prototype breast CT scanner previously1. In this study, the effects on MTF from different components in the imaging chain, including the x-ray focal spot distribution, detector lag, and x-ray detector MTF were physically measured. The contributions of these three factors and gantry motion affecting CT system resolution were studied using computer simulation. The CT system MTF was computed and the role of each factor was studied independently. The simulated MTF results demonstrated that the x-ray focal spot size and detector MTF have an effect on the system resolution, while the scanner motion degrades only the azimuthal MTF, with greater degradation occurring further from isocenter where greater rotational velocities occur. The azimuthal MTF of this system has a cutoff frequency of 2.0 cycles/mm at the isocenter but degrades to 1.0 cycles/mm at the periphery. The radial MTF has a cutoff frequency of 2.0 cycles/mm, at both the isocenter and periphery. The comparison between the computer simulated and physically measured MTF values demonstrates reasonable accuracy in the simulation process. The results from computer simulation also suggest ways in which the spatial resolution can be improved by system modification.
Novel C-arm based cone-beam CT using a source trajectory of two concentric arcs
Show abstract
The first results from an interventional C-arm based computed tomography system where a complete source trajectory was used are presented. A scan with two arcs which are joined approximately at the center of their paths (CC trajectory) is utilized here. This trajectory satisfies Tuy's sufficiency condition for a large volume, but is not well
populated with PI-lines. Therefore, a non-PI-line based reconstruction
method is required. The desire for high dose efficiency led to the selection of an equal weighting based method. An FBP type reconstruction algorithm which was derived for two orthogonal concentric circles was utilized for reconstruction.
The concept of a virtual image object was used to relate the projections from the two acquired non-orthogonal arcs to projections of a virtual object from two orthogonal arcs. Geometrical calibration is vital when performing tomography from an interventional system, and was incorporated here with the use of a homogeneous virtual projection matrix. The results demonstrate a significant reduction in cone-beam artifacts when the complete source trajectory is utilized.
Microcomputed tomography with a photon-counting x-ray detector
Show abstract
In this work we used a novel CdTe photon counting x-ray detector capable of very high count rates to perform x-ray micro-computed tomography (microCT). The detector had 2 rows of 384 square pixels each 1 mm in size. Charge signals from individual photons were integrated with a shaping time of ~60 ns and processed by an ASIC located in close proximity to the pixels. The ASIC had 5 energy thresholds with associated independent counters for each pixel. Due to the thresholding, it is possible to eliminate dark-current contributions to image noise. By subtracting counter outputs from adjacent thresholds, it is possible to obtain the number of x-ray photon counts in 5 adjacent energy windows. The detector is capable of readout times faster than 5 ms. A prototype bench-top specimen μCT scanner was assembled having distances from the tube to the object and detector of 11 cm and 82 cm, respectively. We used a conventional x-ray source to produce 80 kVp x-ray beams with tube currents up to 400 μA resulting in count rates on the order of 600 kcps per pixel at the detector. Both phantoms and a dead mouse were imaged using acquisition times of 1.8 s per view at 1° steps around the object. The count rate loss (CRL) characteristics of the detector were measured by varying the tube current and corrected for using a paralyzable model. Images were reconstructed using analytical fan-beam reconstruction. The reconstructed images showed good contrast and noise characteristics and those obtained from different energy windows demonstrated energy-dependent contrast, thus potentially allowing for material decomposition.
Evaluation of noise power spectra of CT images
Show abstract
Using a commercial clinical CT scanner (GE lightspeed), nine CT scans were performed on a 20 cm diameter plastic
pipe filled with water. The mAs was varied from 10 to 400mAs and the beam energy was varied from 80 to140kVp. For
each scan three volume datasets were reconstructed using different filters. Noise power spectrum (NPS) curves were
measured to examine the effect of varying kVp, mAs and reconstruction filter on the noise content. Sixteen slices from
each of the reconstructed volumes were used to compute the NPS; the central 192x192 pixels of each slice were split into
four overlapping regions of interest (ROI) of 128x128 pixels. A total of 64 ROI were used per scan. The magnitude
squared of the 2D Fourier transform of each ROI was computed. The mean of the 64 2D results was averaged over radial
frequency, yielding a 1D NPS. The overall shape of the NPS was dependent on the reconstruction filter used. The
magnitude of the curves decreased with the increase of mAs or kVp. kVp, mAs, and the reconstruction filter can be
adjusted to modulate the amount of noise present in resulting CT volumes, but the effect these values have on the patient
must be considered. The relationship between NPS and the Noise Equivalent Quanta (NEQ) makes trends in NPS
important and is the motivation for this evaluation and future research.
Signal Corrections
Evaluation of scatter effects on image quality for breast tomosynthesis
Show abstract
Digital breast tomosynthesis uses a limited number of low-dose x-ray projections to produce a three-dimensional (3D) tomographic reconstruction of the breast. The purpose of this investigation was to characterize and evaluate the effect of scatter radiation on image quality for breast tomosynthesis.
Generated by a Monte Carlo simulation method, scatter point spread functions (PSF) were convolved over the field of view (FOV) to estimate the distribution of scatter for each angle of tomosynthesis projection. The results demonstrated that in the absence of scatter reduction techniques, the scatter-to-primary ratio (SPR) levels for the average breast are quite high (~0.4 at the centre of mass), and increased with increased breast thickness and with larger FOV.
Associated with such levels of x-ray scatter are cupping artifacts, as well as reduced accuracy in reconstruction values. The effect of x-ray scatter on the contrast, noise, and signal-difference-to-noise ratio (SDNR) in tomosynthesis reconstruction was measured as a function of tumour size. For example, the contrast in the reconstructed central slice of a tumour-like mass (14 mm in diameter) was degraded by 30% while the inaccuracy of the voxel value was 28%, and the reduction of SDNR was 60%. We have quantified the degree to which scatter degrades the image quality over a wide range of parameters, including x-ray beam energy, breast thickness, breast diameter, and breast composition. However, even without a scatter rejection device, the contrast and SDNR in the reconstructed tomosynthesis slice is higher than that of conventional mammographic projection images acquired with a grid at an equivalent total exposure.
Improved scatter correction for x-ray conebeam CT using primary modulation
Show abstract
Recently, we proposed a scatter correction method for x-ray imaging using primary modulation. A primary
modulator with spatially variant attenuating materials is inserted between the x-ray source and the object to
make the scatter and part of the primary distributions strongly separate in the Fourier domain. Linear filtering
and demodulation techniques suffice to extract and correct the scatter for this modified system. The method has
been verified by computer simulations and preliminary experimental results on a simple object. In this work, we
improve performance by using a new primary modulator with a higher modulation frequency and by refining the
algorithm. The improved method is evaluated experimentally using a pelvis phantom. The imaging parameters
are chosen to match the Varian Acuity CT simulator, where scatter correction has been shown to be challenging
due to complicated artifact patterns. The results using our approach are compared with those without scatter
correction, and with scatter estimated and corrected using a slit measurement as a pre-scan. The comparison
shows that the primary modulation method greatly reduces the scatter artifacts and improves image contrast.
Using only one single scan, this method achieves CT HU accuracy comparable to that obtained using a slit measurement as a pre-scan.
Evaluation of scatter mitigation strategies for x-ray cone-beam CT: impact of scatter subtraction and anti-scatter grids on contrast-to-noise ratio
Show abstract
The large contribution of scatter to cone-beam computed tomography (CBCT) x-ray projections significantly degrades
image quality, both through streaking and cupping artifacts and by loss of low contrast boundary detectability. The goal
of this investigation is to compare the efficacy of three widely used scatter mitigation methods: subtractive scatter
correction (SSC); anti-scatter grids (ASG); and beam modulating with bowtie filters; for improving signal-to-noise ratio
(SNR), contrast, contrast-to-noise ratio (CNR) and cupping artifacts. A simple analytic model was developed to predict
scatter-to-primary ratio (SPR) and CNR as a function of cylindrical phantom thickness. In addition, CBCT x-ray
projections of a CatPhan QA phantom were measured, using a Varian CBCT imaging system, and computed, using an
inhouse Monte Carlo photon-transport code to more realistically evaluate the impact of scatter mitigation techniques.
Images formed with uncorrected sinograms acquired without ASGs and bow-tie filter show pronounced cupping
artifacts and loss of contrast. Subtraction of measured scatter profiles restores image uniformity and CT number
accuracy, but does not improve CNR, since the improvement in contrast almost exactly offset by the increase in relative
x-ray noise. ASGs were found to modestly improve CNR (up to 20%, depending ASG primary transmission and
selectivity) only in body scans, while they can reduce CNR for head phantoms where SPR is low.
Motion artifact reduction in fan-beam and cone-beam computed tomography via the fan-beam data consistency condition (FDCC)
Show abstract
Motion contamination in computed tomography projection data causes significant artifacts in the reconstructed images. If during the tomographic acquisition the object is relatively stationary during a portion of the acquisition, and then
moves significantly, the projection data after the motion will be inconsistent with the projection data during the period of relative stationarity. The fan-beam data consistency condition (FDCC) provides a means to directly estimate motion contaminated projection data based on all of the projection data acquired. Thus, the FDCC may be used to
combat many types of motion contamination in computed tomography. This approach to motion artifact correction is novel as none of the previous methods for artifact correction utilized a direct estimation of motion contaminated data.
Additionally, this methodology depends upon only a small amount of a priori information and is not based on a motion model. Another distinguishing feature of this method is that it operates directly in the projection space, and is completely independent of
the reconstruction algorithm utilized.
An example of clinical relevance of coronary motion artifact reduction is presented using both simulated projection data as well as projection data acquired with a porcine
model using a state-of-the-art 64 row volumetric CT scanner. Significant reduction in motion related artifacts
is achieved in both the simulation case and the porcine model.
Resolution and noise trade-off analysis for volumetric CT
Show abstract
In this study, we investigate the relationship between quantum noise and spatial resolution for volumetric CT. Both theoretical analysis and experiments were performed to investigate their relationship. In theory, quantum noise can be derived from its relationship to dose, in-plane spatial resolution, recon kernel, and signal-to-noise ratio (SNR). In the experiments, by scanning a Teflon sphere phantom, the 3-D MTF was measured from the edge profile along the spherical surface. Cases of different resolutions (and noise levels) were generated by adjusting recon kernel. To reduce bias, the total photon fluxes were matched: 120kVp, 260mA, and 1sec per gantry rotation. In the end, all data sets were reconstructed using modified FDK algorithm under the same condition: FOV=10cm and slice thickness=0.625mm. Finally, we investigated the efficiency of an image-space adaptive smoothing filter as a noise reduction tool and its impact on spatial resolution. The theoretical analysis indicated that the variance of noise is proportional to at least 4th power of the spatial resolution. Our experimental results supported this conclusion by showing the relationship is 4.6th (helical) or 5th (axial) power. Results also showed that, with properly designed image-space smoothing filters, it is feasible to reduce quantum noise (and the power relationship to a lower order) with smaller loss of spatial resolution.
Sinogram restoration of dual focal spot CT data
Show abstract
The raw data generated by a computed tomography (CT) machine are not readily usable for reconstruction. This is the result of various system non-idealities, and although the deterministic nature of corruption effects like crosstalk and afterglow permits removal through deconvolution, there is the drawback that deconvolution increases noise.
Methods that perform raw data correction combined with noise suppression are commonly termed sinogram restoration methods. The need for sinogram restoration arises, for example, when photon counts are low and non-statistical reconstruction algorithms like filtered backprojection are used.
Many modern CT machines offer a so-called dual focal spot (DFS) mode, which serves the goal of increased radial sampling by switching the focal spot between two positions on the anode plate during the scan. Although the focal spot mode does not play a role with respect to how the data are affected by the above mentioned corruption effects, it needs to be taken into account, if regularized sinogram restoration is to be applied to the data.
This work points out the subtle difference in processing that sinogram restoration for DFS requires, how it is correctly employed within the penalized maximum likelihood sinogram restoration algorithm, and what impact that has on image quality.
Cone Beam Reconstruction
A Poisson likelihood iterative reconstruction algorithm for material decomposition in CT
Show abstract
Emerging photon-counting detectors with energy discrimination ability for x-ray CT perform binning according
to the energy of the incoming photons. Multiple output channels with different energy thresholds can be obtained
in one irradiation. The energy dependency of attenuation coefficients can be described by a linear combination
of basis functions, e.g., Compton scatter and photo-electric effect; their individual contributions can be differentiated
by using the multiple energy channels hence material characterization is made possible. Conventional
analytic approach is a two-step process. First decompose in the projection domain to obtain the sinograms
corresponding to the coefficients of the basis functions, then apply FBP to obtain the individual material components.
This two-step process may have poor quality and quantitative accuracy due to the lower counts in the
separate energy channels and approximation errors propagated to the image domain from projection domain
decomposition. In this work we modeled the energy dependency of linear attenuation coefficients in our problem
formulation and applied the optimality transfer principle to derive a Poisson-likelihood based algorithm for
material decomposition from multiple energy channels. Our algorithm reconstructs the coefficients of the basis
functions directly therefore the separate non-linear estimation step in the projection domain as in conventional
approaches is avoided. We performed simulations to study the accuracy and noise properties of our method.
We synthesized the linear attenuation coefficients at a reference energy and compared with standard attenuation
values provided by NIST. We also synthesized the attenuation maps at different effective energy bin centers
corresponding to the different energy channels and compared the synthesized images with reconstructions from
standard fan-beam FBP methods. Preliminary simulations showed that our reconstructed images have much better noise properties.
Limited view angle tomographic image reconstruction via total variation minimization
Show abstract
In tomosynthesis, cone-beam projection data are acquired from a
few of view angles, which are not sufficient for an exact
reconstruction of an image object using state-of-the-art image
reconstruction algorithms. In the case of parallel-beam
projections, the well-known projection-slice theorem may be
utilized to transform the parallel-beam projections into the
Fourier space of an image object. Due to the limited range of view
angles, the available projection data can only populate a portion
of Fourier space. Moreover, the angular sampling rate of the
populated portion of the Fourier space may not satisfy the Nyquist
criterion. Thus, reconstructed images using direct Fourier
inversion contain severe streaking and distortion artifacts. In
this paper, we present a novel image reconstruction method via
minimizing the total variation (TV) of the reconstructed image for
limited view angle X-ray computed tomography. Specifically, the
missing data points in Fourier space, due to either the limited
range or undersampling of view angles, are iteratively filled
using the following two constraint conditions: (1) the total
variation of the reconstructed image is minimized and (2)
reconstructed image maintains fidelity to the sampled data in the
Fourier space. Using analytical phantoms, numerical simulations
were conducted to validate the new image reconstruction method.
Images are compared with two other image reconstruction methods in
terms of image artifact level and noise properties. Numerical
results demonstrated that the new image reconstruction algorithm
is superior to direct Fourier inversion reconstruction algorithm
and the projection onto convex sets (POCS) image reconstruction
algorithm.
Analytical cone-beam reconstruction using a multi-source inverse geometry CT system
Show abstract
In a 3rd generation CT system, a single source projects the entire field of view (FOV) onto a large detector opposite the source. In multi-source CT imaging, a multitude of sources sequentially project a part of the FOV on a much smaller detector. These sources may be distributed in both the trans-axial and axial directions in order to jointly cover the entire FOV. Scan data from multiple sources in the axial direction provide complementary information, which is not available in a conventional single-source CT system. In this work, an analytical 3D cone-beam reconstruction algorithm for multi-source CT is proposed. This approach has three distinctive features. First, multi-source data are re-binned transaxially to multiple offset third-generation datasets. Second, data points in sinograms from multiple source sets are either accepted or rejected for contribution to the backprojection of a given voxel. Third, instead of using a ramp filter, a Hilbert transform is combined with a parallel derivative to form the filtering mechanism. Phantom simulations are performed using a multi-source CT geometry and compared to conventional 3rd generation CT geometry. We show that multi-source CT can extend the axial scan coverage to 120mm without cone-beam artifacts, while a third-generation geometry results in compromised image quality at 60mm of axial coverage. Moreover, given that the cone-angle in the proposed geometry is limited to 7 degrees, there are no degrading effects such as the Heel effect and scattered radiation, unlike in a third-generation geometry with comparable coverage. An additional benefit is the uniform flux profile resulting in uniform image noise throughout the FOV and a uniform dose absorption profile.
Gated cone-beam CT imaging of the thorax: a reconstruction study
Show abstract
In radiotherapy, the respiratory motion of the patient in treatment position is determined from gated cone-beam CT images. This method consists in selecting projections according to a respiratory signal for the reconstruction of a CT image of each respiratory state. This selection can be based on the amplitude or the phase of the signal. The number of selected projections also depends on the width of the gating window. The present study compares different reconstructions using a dynamic digital phantom of the thorax and a known respiratory signal. We applied both regular and irregular motions to this phantom and reconstructed the exhale state using different widths. We evaluated these reconstructions with the signal-to-noise ratio, the contrast-to-noise ratio and a blur criterion. In the case of a regular motion, there was no difference between the amplitude and the phase. The signal quality was high, even for the smallest width, and the blur increased with width. In the case of an irregular motion, the difference was noticeable. Amplitude-based reconstructions suffered from severe artifacts with the smallest width because there were respiratory cycles for which no projection was selected. This drawback is overcome by increasing the width of the gating window. Phase-based reconstructions also allowed to avoid artifacts, whatever the width. But the blur was higher, even for the smallest width applied. These results suggest that the gating process must be adjusted in order to select at least one projection per respiratory cycle. Phase gating is a robust way to achieve this goal when respiration is irregular. Amplitude gating may be more effective in terms of blur, but the width must be carefully chosen to avoid severe artifacts. Finally, we observed the potential of dynamic reconstruction by using a motion model to deform different gated CT images toward a common reference and compute the weighted mean. The resulting CT image suffered less from artifacts than each gated CT image separately even if artifacts were still visible.
A practical reconstruction algorithm for CT noise variance maps using FBP reconstruction
Show abstract
Accurate prediction of reconstructed noise in computed tomography (CT) images is important for purposes of
system design, optimization and evaluation. A large body of work describes noise prediction methods for CT,
the vast majority of which assume stationarity of both noise and signal processes. Consequently, these methods
are usually applied to and evaluated using simple phantoms, and only a portion of the image is scrutinized.
In this work, we derive a practical method for reconstructing CT noise variance maps for arbitrary objects.
Photon Poisson noise and system electronic noise are considered. The final formula has the same structure as
that of the filtered backprojection (FBP) formula, but with different weighting factors and convolution kernels.
The algorithm is verified using computer simulations of the Shepp-Logan phantom, and a good match is found
between the predicted noise map from one single noisy scan and the measured noise using 128 noisy scans.
As compared to other proposed noise models, our complementary work provides a method of noise prediction
by simple adaptation of FBP reconstruction algorithms. The result is a tool that can be useful for system
optimization and evaluation tasks as well as the design of reconstruction filters.
Advanced Reconstruction
A method for atlas-based volumetric registration with surface constraints for optical bioluminescence tomography in small animal imaging
Show abstract
Atlases are normalized representations of anatomy that can provide a standard coordinate system for in vivo
imaging studies. For Optical Bioluminescence Tomography (OBT) in small animals, the animal's surface topography
can be reconstructed from structured light measurements, but internal anatomy is unavailable unless
additional CT or MR images are acquired. We present a novel method for estimating the internal organ structure
of a mouse by warping a labeled 3D volumetric mouse atlas with the constraint that the surfaces of the two
should match. Surface-constrained harmonic maps used for this bijective warping are computed by minimizing
the covariant harmonic energy. We demonstrate the application of this warping scheme in OBT, where scattering
and absorption coefficients of tissue are functions of the internal anatomy and hence, better estimates
of the organ structures can lead to a more accurate forward model resulting in improved source localization.
We first estimated the subject's internal geometry using the atlas-based warping scheme. Then the mouse was
tessellated and optical properties were assigned based on the estimated organ structure. Bioluminescent sources
were simulated, an optical forward model was computed using a finite-element solver, and multispectral data
were simulated. We evaluate the accuracy of the forward model computed using the warped atlas against that
assuming a homogeneous mouse model. This is done by comparing each model against a 'true' optical forward
model where the anatomy of the mouse is assumed known. We also evaluate the impact of anatomical alignment
on bioluminescence source localization.
3D bioluminescent source localization of different depths with spectrum information and adaptive finite element analysis
Show abstract
In this contribution, a novel adaptive finite element based source
localization method is proposed and developed, which adopts the
spectrum characteristic of bioluminescent source. In this method,
multiple sets of surface measured data corresponding to the
discretized spectral bands reduce the ill-posedness of BLT. Then,
the adaptive mesh refinement using a posteriori error
estimation helps not only to avoid the dimension disaster arisen
from the multi-spectral measured data, but also to improve the BLT
reconstruction. In addition, a posteriori permissible source
region selection further reduces the ill-posedness of BLT and
enhances the reconstruction stability and speed. Finally, the
simulation verification of the proposed method is performed with the
Monte Carlo (MC) based synthetic data. The accurate localization of
one and two bioluminescent sources in different depths shows the
effectiveness and potential of this algorithm for BLT.
An iterative method for the reconstruction of the coronary arteries from rotational x-ray angiography
Show abstract
Three-dimensional (3D) reconstruction of the coronary arteries offers great advantages in the diagnosis and treatment of cardiovascular diseases, compared to two-dimensional X-ray angiograms. Besides improved roadmapping, quantitative analysis of vessel lesions is possible. To perform 3D reconstruction, rotational projection data of the selectively contrast agent enhanced coronary arteries are acquired with simultaneous ECG recording. For the reconstruction of one cardiac phase, the corresponding projections are selected from the rotational sequence by nearest-neighbor ECG gating. This typically provides only 5-10 projections per cardiac phase. The severe angular undersampling leads to an ill-posed reconstruction problem.
In this contribution, an iterative reconstruction method is presented which employs regularizations especially suited for the given reconstruction problem. The coronary arteries cover only a small fraction of the reconstruction volume. Therefore, we formulate the reconstruction problem as a minimization of the L1-norm of the reconstructed image, which results in a spatially sparse object. Two additional regularization terms are introduced: a 3D vesselness prior, which is reconstructed from vesselness-filtered projection data, and a Gibbs smoothing prior. The regularizations favor the reconstruction of the desired object, while taking care not to over-constrain the reconstruction by too detailed a-priori assumptions. Simulated projection data of a coronary artery software phantom are used to evaluate the performance of the method. Human data of clinical cases are presented to show the method's potential for clinical application.
Image reconstruction in digital breast tomosynthesis by total variation minimization
Show abstract
The total variation (TV) minimization algorithm for image reconstruction in few-view computed
tomography is applied to image reconstruction in digital breast tomosynthesis. In the TV minimization
algorithm, there is a parameter that regulates how close the estimated data should be to
the actual projection data. The effect of this parameter on the reconstructed images is investigated.
In addition, realistic noise is added to the simulated projection data.
Planar tomosynthesis reconstruction in a parallel-beam framework via virtual object reconstruction
Show abstract
A framework for
image reconstruction from planar tomosynthesis trajectories
(i.e. all source positions reside in a single plane)
is presented.
The parallel beam geometry is a convenient starting point in deriving
reconstruction algorithms, both analytic and iterative, as the
relation between frequency space and image space is well known.
We present a method for utilizing
parallel beam reconstruction algorithms in an internally consistent
manner. The concept of a virtual image object is utilized. This virtual image
object has the property that cone-beam projections through the real object
are directly related to parallel-beam projections of the virtual object. The
virtual object may then be reconstructed using any algorithm derived for
parallel beam projections. Finally, an affine transform may be applied to the
virtual image object in order to generate the reconstruction
result. In the implementation described here the
backprojection operation is performed such that the real image object is reconstructed
without introducing a rebinning in image space. Image reconstruction
comparisons are given for a standard filtered backprojection (FBP) type algorithm where parallel
projections were assumed in the algorithm derivation. Numerical simulations
were performed for a C-arm type geometry and parallel beam FBP reconstructions
using the virtual object are compared with the standard backprojection algorithm.
Finally, a comparison was made between the new parallel beam reconstruction and
the standard approximation where the cone-beams are assumed to be approximately
parallel beams and a cone-beam backprojection is employed.
A reduction in streaking artifacts was observed using the new algorithm compared with the standard
approximation.
Sub-pixel compounding from elasticity imaging data
Show abstract
Sub-pixel compounding is a technique that synthesizes the information of an image sequence to form a betterresolved
and speckle reduced image. To avoid extra data acquisition time and patient exposure, reuse of the existing data
is highly desired. In elasticity imaging, a set of images with slight changes due to deformation is produced, which
provides an ideal input for the sub-pixel compounding process. In this paper, a brief review of the resolution
enhancement techniques in ultrasound imaging will be provided, and then, a diffusion-regularized, least square approach
is presented for sub-pixel compounding image reconstruction. Based on the results, we suggest that (1) B-mode images
from elastic imaging are suitable data for sub-pixel compounding and a speckle noise reduced higher-resolution image is
a co-product of elasticity imaging; (2) for breast diagnosis, resolution improvement is of strong interest since better
depiction of the interior and exterior structures of a tumor provides important detection and diagnostic information; (3) a
similar approach could be extended to elasticity imaging with other modalities.
Poster Session: Radiography/Computed Tomography
Experimental benchmarking of a Monte Carlo dose simulation code for pediatric CT
Show abstract
In recent years, there has been a desire to reduce CT radiation dose to children because of their susceptibility and
prolonged risk for cancer induction. Concerns arise, however, as to the impact of dose reduction on image quality and
thus potentially on diagnostic accuracy. To study the dose and image quality relationship, we are developing a
simulation code to calculate organ dose in pediatric CT patients. To benchmark this code, a cylindrical phantom was
built to represent a pediatric torso, which allows measurements of dose distributions from its center to its periphery.
Dose distributions for axial CT scans were measured on a 64-slice multidetector CT (MDCT) scanner (GE Healthcare,
Chalfont St. Giles, UK). The same measurements were simulated using a Monte Carlo code (PENELOPE, Universitat de
Barcelona) with the applicable CT geometry including bowtie filter. The deviations between simulated and measured
dose values were generally within 5%. To our knowledge, this work is one of the first attempts to compare measured
radial dose distributions on a cylindrical phantom with Monte Carlo simulated results. It provides a simple and effective
method for benchmarking organ dose simulation codes and demonstrates the potential of Monte Carlo simulation for
investigating the relationship between dose and image quality for pediatric CT patients.
Low dose applications of lightspeed VCT in cardiac imaging
Show abstract
We introduced and evaluated the techniques LightSpeed VCT uses to reduce X-ray dose and image noise in cardiac helical CT applications. These techniques include the use of much improved VCT data acquisition system (VDAS) with reduced electronic noise; cardiac bowtie that redistributes X-rays to have more signals for heart and much less flux to the peripheries; adaptive post-processing filters to reduce cardiac image noise; and ECG modulated tube currents to concentrate X-ray dose for desired cardiac phases. Phantom and patient scans were used to evaluate the dose saving and noise reduction potentials of these techniques. The results demonstrated that the improved VDAS reduced image noise 15-20% for cardiac imaging. With same scan technique, the use of cardiac bowtie reduced about 10% dose in terms of CTDIw measurement and clinical evaluation demonstrated additional 7% image noise reduction and equivalent image quality with cardiac bowtie vs. regular body bowtie. The adaptive filter generated 15-20% noise reduction while maintaining image resolution and artery sharpness. Finally, the use of ECG modulated mA method provided up to 50% dose reduction based on CTDIw measurements, but the saving potentials depended on the heart rate and cardiac phase selection. For heart rate of 60bpm and ±15% cardiac phase margin, the average dose reduction could be 30%. Since these dose and noise reduction methods are inclusive and can be combined to produce even greater dose/noise reduction. It is reasonable to believe that with VCT we maybe able to acquire cardiac helical CT images with only 30-40% of the dose of older generation 16-slice CT scanners with similar noise level and same slice thickness.
Radiation dose from MDCT using Monte Carlo simulations: estimating fetal dose due to pulmonary embolism scans accounting for overscan
E. Angel,
C. Wellnitz,
M. Goodsitt,
et al.
Show abstract
Pregnant women with shortness of breath are increasingly referred for CT Angiography to rule out Pulmonary Embolism (PE). While this exam is typically focused on the lungs, extending scan boundaries and overscan can add to the irradiated volume and have implications on fetal dose. The purpose of this work was to estimate radiation dose to the fetus when various levels of overscan were encountered.
Two voxelized models of pregnant patients derived from actual patient anatomy were created based on image data. The models represent an early (< 7 weeks) and late term pregnancy (36 weeks). A previously validated Monte Carlo model of an MDCT scanner was used that takes into account physical details of the scanner. Simulated helical scans used 120 kVp, 4x5 mm beam collimation, pitch 1, and varying beam-off locations (edge of the irradiated volume) were used to represent different protocols plus overscan. Normalized dose (mGy/100mAs) was calculated for each fetus.
For the early term and the late term pregnancy models, fetal dose estimates for a standard thoracic PE exam were estimated to be 0.05 and 0.3 mGy/100mAs, respectively, increasing to 9 mGy/100mAs when the beam-off location was extended to encompass the fetus.
When performing PE exams to rule out PE in pregnant patients, the beam-off location may have a large effect on fetal dose, especially for late term pregnancies. Careful consideration of ending location of the x-ray beam - and not the end of image data - could result in significant reduction in radiation dose to the fetus.
Methodology for determining dose reduction for chest tomosynthesis
Show abstract
Digital tomosynthesis is an imaging technique that reconstructs tomographic planes in an object from a set of projection
images taken over a fixed angle1. Preliminary results show that this technique increases the detectability of lung
nodules2. Current settings acquire images with approximately the same exposure as a screen-film lateral. However, due
to the increased detectability of lung nodules from the removal of overlying structures, patient dose may be reduced
while still maintaining increased sensitivity and specificity over conventional chest radiographs. This study describes a
simulation method that provides realistic reduced dose images by adding noise to digital chest tomosynthesis images in
order to simulate lower exposure settings for the purpose of dose optimization. Tomosynthesis projections of human
subjects were taken at dose levels which were specified based on either patient thickness or a photo-timed digital chest
radiograph acquired prior to tomosynthesis acquisition. For the purposes of this study, subtle nodules of varying size
were simulated in the image for demonstration purposes before the noise simulation in order to have a known truth for
nodule location and to evaluate the effect of additive noise on tumor detection. Noise was subsequently added in order to
simulate 3/4, 1/2, and 1/4 of the original exposure in each projection. The projections were then processed with the MITS
algorithm to produce slice images. This method will be applied to a study of dose reduction in the future using human
subject cases.
Coupling the use of anti-scatter grid with analytical scatter estimation in cone beam CT
Show abstract
Cone-Beam Computed Tomography (CBCT) enables three-dimensional imaging with isotropic resolution. X-ray scatter estimation is a big challenge for quantitative CBCT imaging: even in the presence of anti-scatter grid, the scatter level is significantly higher on cone beam systems compared to collimated fan beam systems. The effects of this scattered radiation include cupping artifacts, streaks, and quantification inaccuracies. In this paper, a scatter management process for tomographic projections, without supplementary on-line acquisition, is presented. The scattered radiation is corrected using a method based on scatter calibration through off-line acquisitions. This is combined with on-line analytical transformation based on physical equations, to perform an estimation adapted to the object observed. This approach has been previously applied to a system without anti-scatter grid. The focus of this paper is to show how to combine this approach with an anti-scatter grid. First, the interest of the grid is evaluated in terms of noise to signal ratio and scatter rejection. Then, the method of scatter correction is evaluated by testing it on an anthropomorphic phantom of thorax. The reconstructed volume of the phantom is compared to that obtained with a strongly collimated conventional multi-slice CT scanner. The new method provides results that closely agree with the conventional CT scanner, eliminating cupping artifacts and significantly improving quantification.
Initial application of digital tomosynthesis to improve brachytherapy treatment planning
Show abstract
We present preliminary investigations that examine the feasibility of incorporating volumetric images generated using
digital tomosynthesis into brachytherapy treatment planning. The Integrated Brachytherapy Unit (IBU) at our facility
consists of an L-arm, C-arm isocentric motion system with an x-ray tube and fluoroscopic imager attached. Clinically,
this unit is used to generate oblique, anterior-posterior, and lateral images for simple treatment planning and dose
prescriptions. Oncologists would strongly prefer to have volumetric data to better determine three dimensional dose
distributions (dose-volume histograms) to the target area and organs at risk. Moving the patient back and forth to CT
causes undo stress on the patient, allows extensive motion of organs and treatment applicators, and adds additional time
to patient treatment. We propose to use the IBU imaging system with digital tomosynthesis to generate volumetric
patient data, which can be used for improving treatment planning and overall reducing treatment time. Initial image data
sets will be acquired over a limited arc of a human-like phantom composed of real bones and tissue equivalent material.
A brachytherapy applicator will be incorporated into one of the phantoms for visualization purposes. Digital
tomosynthesis will be used to generate a volumetric image of this phantom setup. This volumetric image set will be
visually inspected to determine the feasibility of future incorporation of these types of images into brachytherapy
treatment planning. We conclude that initial images using the tomosynthesis reconstruction technique show much
promise and bode well for future work.
CatSim: a new computer assisted tomography simulation environment
Show abstract
We present a new simulation environment for X-ray computed tomography, called CatSim. CatSim provides a research platform for GE researchers and collaborators to explore new reconstruction algorithms, CT architectures, and X-ray source or detector technologies. The main requirements for this simulator are accurate physics modeling, low computation times, and geometrical flexibility. CatSim allows simulating complex analytic phantoms, such as the FORBILD phantoms, including boxes, ellipsoids, elliptical cylinders, cones, and cut planes. CatSim incorporates polychromaticity, realistic quantum and electronic noise models, finite focal spot size and shape, finite detector cell size, detector cross-talk, detector lag or afterglow, bowtie filtration, finite detector efficiency, non-linear partial volume, scatter (variance-reduced Monte Carlo), and absorbed dose. We present an overview of CatSim along with a number of validation experiments.
Dual-energy contrast enhanced digital mammography using a new approach for breast tissue canceling
Show abstract
In this study, we propose a novel approach to dual-energy
contrast-enhanced digital mammography, with the development of a
dual-energy recombination algorithm based on an image chain model
and the determination of the associated optimal low and high-energy
techniques. Our method produces clutter-free iodine-equivalent
images and includes thickness correction near the breast border.
After the algorithm description, the optimal low and high-energy
acquisition techniques are determined to obtain a compromise between
image quality and glandular dose. The low and high-energy techniques
were chosen to minimize the glandular dose for a target Signal
Difference to Noise Ratio (SDNR) in the dual-energy recombined
image. The theoretical derivation of the iodine SDNR in the
recombined image allowed the prediction of the optimal low and
high-energy techniques. Depending on the breast thickness and
glandular percentage, the optimal low-energy kVp and mAs ranged from
24kVp (Mo/Mo or Mo/Rh) to 35kVp (Rh/Rh), and from 60 to 90mAs
respectively, and the high-energy kVp and mAs ranged from 40kVp to
47kVp (Mo/Cu), and from 80mAs to 290mAs. We proved the better
performance of our algorithm compared to the classic weighted
logarithmic subtraction method in terms of patient dose and also in terms of texture
cancelation, through the use of artificial textured images. Values of iodine contrast measured on phantom were
close to the expected iodine thickness. Good correlation was found
between the measured and theoretical iodine SDNR in the dual-energy
images, which validates our theoretical optimization of the
acquisition techniques.
Dedicated dental volumetric and total body multislice computed tomography: a comparison of image quality and radiation dose
Show abstract
Aim of this work is to compare the performances of a Xoran Technologies i-CAT Cone Beam CT for dental applications with those of a standard total body multislice CT (Toshiba Aquilion 64 multislice) used for dental examinations. Image quality and doses to patients have been compared for the three main i-CAT protocols, the Toshiba standard protocol and a Toshiba modified protocol. Images of two phantoms have been acquired: a standard CT quality control phantom and an Alderson Rando® anthropomorphic phantom. Image noise, Signal to Noise Ratio (SNR), Contrast to Noise Ratio (CNR) and geometric accuracy have been considered. Clinical image quality was assessed. Effective dose and doses to main head and neck organs were evaluated by means of thermo-luminescent dosimeters (TLD-100) placed in the anthropomorphic phantom. A Quality Index (QI), defined as the ratio of squared CNR to effective dose, has been evaluated. The evaluated effective doses range from 0.06 mSv (i-CAT 10 s protocol) to 2.37 mSv (Toshiba standard protocol). The Toshiba modified protocol (halved tube current, higher pitch value) imparts lower effective dose (0.99 mSv). The conventional CT device provides lower image noise and better SNR, but clinical effectiveness similar to that of dedicated dental CT (comparable CNR and clinical judgment). Consequently, QI values are much higher for this second CT scanner. No geometric distortion has been observed with both devices. As a conclusion, dental volumetric CT supplies adequate image quality to clinical purposes, at doses that are really lower than those imparted by a conventional CT device.
Semi-empirical scattering correction model for MSCT
Show abstract
Scattering in CT is a major process that may result in severe image artifacts.
In order to suppress this scattering, most CT scanners are equipped with post collimation anti-scatter grid along the fan beam direction. In the longitudinal direction (z-direction) no hardware solution has been implemented since the scattering is negligible in narrow coverage CT scanners. As the coverage becomes wider in recent MSCT scanners the scattering level in the z-direction increases significantly. This scattering increase in z-direction, results in image artifacts appearing as dark shadows along highly attenuating directions.
In the present work we measure the scattering level in the z-direction for a wide coverage scanner, using various phantoms.
Based on the results, a semi-empirical model for scattering correction in MSCT is presented and validated.
The proposed model is based on a subtraction of a low frequency offset. This offset is proportional to the scattering, corresponding to the detector that has the lowest signal at each rotation angle.
To validate the model, we first calculate this low frequency offset directly from the scatter measurements and apply it to the data acquired with the wide coverage. We then use a semi-empirical function to estimate the scattering fraction from the raw data, and use it to replace the directly measured scatter values in the correction scheme.
Applying the proposed semi-empirical scatter correction model to the data acquired with the wide coverage, the scattering signal is significantly decreased. The images reconstructed from the corrected data exhibit a clear reduction of the artifact level.
Scattering phenomena in MSCT: measurements and analysis
Show abstract
In multi-slice CT (MSCT), as the coverage becomes wider, the scattering contribution along the longitudinal direction (z) to the detectors' signal increases. The scattering results in image artifacts, appearing as dark shadows between highly attenuating objects.
In this work we measure the scattering level systematically, using phantoms of various sizes, shapes, and materials. We study the dependencies and their effect on the scattering amount.
We derive an empirical function for the scattering fraction, based on the maximal attenuation at each rotation angle. The function contains a single constant (SC). The variation of SC as a function of the different phantoms is analyzed, showing a clear dependence on the minimal water equivalent axis of each phantom.
The strong dependence of the scattering fraction on the maximal attenuation along each view is shown. This phenomenon can be correlated to a single scatter process along the z axis in the presence of an anti-scatter grid along the direction of the detectors.
The dependence of SC on the minimal axis indicates an additional significant scatter process.
The results validate that the scattering level estimation can be achieved using the derived function, with a minimal variation in the solution parameters. Hence, enabling the introduction of this scatter estimation into an MSCT scattering correction scheme.
Optimized anti-scatter grids for flat panel detectors
Show abstract
Anti-scatter grids are well established in the field of X-ray projection imaging. In general these grids consist of a large
number of parallel lead lamellae separated by X-ray-transparent material. This regular structure defines the characteristic
grid frequency. Modern X-ray imaging systems apply digital receptors, i.e. image intensifiers coupled to a CCD camera
or solid state flat-panel detector. Combining a digital detector and an anti-scatter grid may lead to Moire artifacts. This
results from sampling an analog X-ray image with signal components higher than half the sampling frequency. Especially
in high dose DSA images (Digital Subtraction Angiography) these irritating artifacts may be visible to the user.
In this paper we present a concept for minimizing these grid artifacts: Signal propagation in the detector is modeled by
three steps, scintillator MTF, aperture MTF, and sampling. Since the scintillator MTF is irrelevant for the grid optimization
process, we focus on aperture MTF and sampling. From the given geometry of the detector elements the corresponding
2D Fourier transform is calculated. An evaluation for typical grid frequencies, i.e. arcs around the origin of the 2D
Fourier transform, results in profiles exhibiting pronounced minima. From the respective angle values for these minima,
grid orientation can be optimized for minimum Moire disturbances.
Simulation results for typical detector pixel geometries and for grid frequencies used in practice are validated by measurement
for two different anti-scatter grids on a Siemens angiographic system with a digital flat-panel detector.
On the development of a Gaussian noise model for scatter compensation
Show abstract
The underlying mechanism in projection radiography as well as in computed tomography (CT) is
the accumulative attenuation of a pencil x-ray beam along a straight line. However, when a
portion of photons is deviated from their original path by scattering, it is not valid to assume that
these photons are the survival photons along the lines connecting the x-ray source and the
individual locations where they are detected. Since these photons do not carry the correct spatial
information, the final image is contaminated. Researchers are seeking techniques to reduce
scattering, and hence, improve image quality, by scatter compensation. Previously, we presented
a post-acquisition scatter compensation technique based on an underlying statistical model. We
used the Poisson noise model, which assumed that the signals in the detector individually
followed the Poisson process. Since most x-ray detectors are energy integrating rather than
photon counting, the Poisson noise model can be improved by taking this property into account.
In this study, we developed a Gaussian noise model by the matching-of-the-first-two-moments
method. The Maximum Likelihood Estimator of the scatter-free image was derived via the
expectation maximization (EM) technique. The maximum a posteriori estimate was also
calculated. The Gaussian noise model was preliminarily evaluated on a full-field digital
mammography system.
Clinical usefulness of automatic phase selection in coronary CT angiography (CTA)
Show abstract
With high-speed multislice helical CT, the time needed to select the optimal cardiac phase accounts for a large percentage of the coronary CT angiography examination time because the scan time is short. To reduce the phase selection time, we have developed an automatic cardiac phase selection algorithm and implemented it in the Aquilion 64 scanner. This algorithm calculates the absolute sum of the differences between two raw data sets for subsequent cardiac phases (e.g., 4% and 0%) and generates a velocity curve representing the magnitude of cardiac motion velocity for the entire heart volume. Normally, the velocity curve has two local minimum slow-motion phases corresponding to end-systole and mid-diastole. By applying these local minimum phases in reconstruction, stationary cardiac images can be reconstructed automatically. In this report, the algorithm for generating the velocity curve and the processing time for selecting the optimal cardiac phase are discussed. The accuracy of this method is compared with that of the conventional manual method. In the manual method, a sample plane containing all four cardiac chambers was selected, reconstruction was performed for all phases at 2% intervals, and images were visually evaluated. Optimal phase selection required about 5 min/exam. With automatic phase selection, optimal phase selection required only about 1 min/exam, and the cardiac phases were close to those selected using the manual method. Automatic phase selection substantially reduces the time needed to select the optimal phase and increases patient throughput. Moreover, the influence of operator skill in selecting the optimal phase is minimized.
Enhancement of edge response in same matrix size of x-ray CT image without special image processing
Show abstract
Multi-detector row computed tomography (MDCT) allows high-edge-response imaging in z-direction by thinner
slice compared with the conventional single-detector row CT (SDCT). The edge response in x- and y-direction is,
however, basically depending on the reconstruction function or the image processing. In this work, a method of
enhancing the edge response in x- and y-direction, without special image processing, is proposed. In this method,
projection data of a patient is acquired by high-spatial-frequency sampling, and an edge-enhanced image is
reconstructed with same matrix size as a conventional image by averaging the projection data. Edge-enhancement
effect in this method employs nonlinearlity of the projection data, and special image processing is not required.
In order to verify this proposed method, a large water phantom that consists of five resinous rods and a small
one that the similitude rate is 1/2, and is topologically the same as the former large phantom, were scanned, and
high-spatial-frequency sampling was simulated. After that, reconstructed images were obtained by averaging
the high-spatial-frequency sampled data and edge gradients for some rods were obtained and estimated. As a
result, although image noise increased slightly, edge gradients are improved 25 to 97 % without special image
processing.
Application of time sampling in brain CT perfusion imaging for dose reduction
Show abstract
The purpose of this study is to determine a stable sampling rate not to be affected by sampling shift for reducing radiation exposure with time sampling and interpolation in cerebral perfusion CT examination. Original images were obtained every 1 second for 40 time series from 3 patients, respectively. Time sampling was performed with sampling intervals (SI) from 2 to 10 seconds. Sampling shift was applied from +1 to SI-1 for each sampling rate. For each patient, 30 tissue concentration time-course data were collected, and arterial input curves were fitted by gamma-variate function. The sinc function was introduced for interpolation. Deconvolution analysis based on SVD was performed for quantifying perfusion parameters. The perfusion values through time-varying sampling and interpolation were statistically compared with the original perfusion values. The mean CBF values with increase of sampling interval and shift magnitude from the collected data had a wider fluctuation pattern centering around the original mean CBF. The mean CBV values had a similar tendency to the mean CBF values, but a relatively narrower deviation. The mean MTT values were fluctuated reversely to the trend of the mean CBF values. The stable sampling interval for quantifying perfusion parameters with lower radiation exposure was statistically acceptable up to 4 seconds. These results indicate that sampling shift limits sampling rate for acquiring acceptable perfusion values. This study will help in selecting more reasonable sampling rate for low-radiation-dose CT examination.
Robust temporal resolution of MSCT cardiac scan by rotation-time update scheme based on analysis of patient ECG database
Show abstract
We have analyzed 144 ECG wave-forms that were taken during cardiac CT exams to determine in retrospect the
optimized timing for updating the gantry rotation-time. A score was defined, according to the number of heart beats
during X-ray on, which fulfill the temporal resolution (tR)condition, tR<100mSec. The temporal resolution calculation
was based on dual-cycle π/2 sector segmentation, where the data required for any image is collected during two heart
cycle. The results yield a significant improvement of the tR score with the rotation-time update method relative to using
a fixed minimal rotation-time of the gantry. The analysis suggest that full heart scan with better than 100mSec temporal
resolution per slice can routinely be achieved in 128 slices MSCT scanner by performing gantry rotation-time -update
after patient starts its breath hold. At these conditions the required breath-hold time is expected to be less than 15
seconds.
Development of the translating and rotating volume computed tomography (TRVCT)
Show abstract
We describe a novel system named TRVCT (Translating and Rotating Volume Computed Tomography), developed for computed tomography image from large object with simple method and low price. Tomogram images can be acquired when the object is translating and rotating simultaneously with vertical linear array detector. This method is different from the normal X-ray CT completely. We used fan-beam X-ray, and the direction of the detector and rotating axis are in parallel. Because a hundred or thousand tomograms with Z-axis from just one scanning, it has excellent Z-axis resolution and has an advantage that can improve the resolution in X-Y plane with changing translating speed and frequency of data acquisition. There is no ring artifact that is generated frequently in the third generation CT scanner. So, we can have high resolution tomograms from this TRVCT system. The TRVCT can be used to acquire images for large object like tire, engine, or whole car, and it can remove the scattering from X-ray for high resolution images.
Dose reduction of up to 89% while maintaining image quality in cardiovascular CT achieved with prospective ECG gating
Show abstract
We present the results of dose and image quality performance evaluation of a novel, prospective ECG-gated Coronary
CT Angiography acquisition mode (SnapShot Pulse, LightSpeed VCT-XT scanner, GE Healthcare, Waukesha, WI), and
compare it to conventional retrospective ECG gated helical acquisition in clinical and phantom studies. Image quality
phantoms were used to measure noise, slice sensitivity profile, in-plane resolution, low contrast detectability and dose,
using the two acquisition modes. Clinical image quality and diagnostic confidence were evaluated in a study of 31
patients scanned with the two acquisition modes. Radiation dose reduction in clinical practice was evaluated by tracking
120 consecutive patients scanned with the prospectively gated scan mode. In the phantom measurements, the
prospectively gated mode resulted in equivalent or better image quality measures at dose reductions of up to 89%
compared to non-ECG modulated conventional helical scans. In the clinical study, image quality was rated excellent by
expert radiologist reviewing the cases, with pathology being identical using the two acquisition modes. The average
dose to patients in the clinical practice study was 5.6 mSv, representing 50% reduction compared to a similar patient
population scanned with the conventional helical mode.
Poster Session: Breast Imaging
Characterization of a prototype tabletop x-ray CT breast imaging system
Show abstract
Planar X-ray mammography is the standard medical imaging modality for the early detection of breast cancer. Based on
advancements in digital flat-panel detector technology, dedicated x-ray computed tomography (CT) mammography is a
modality under investigation that offers the potential for improved breast tumor imaging. We have implemented a
prototype half cone-beam CT breast imaging system that utilizes an indirect flat-panel detector. This prototype can be
used to explore and evaluate the effect of varying acquisition and reconstruction parameters on image quality. This
report describes our system and characterizes the performance of the system through the analysis of Modulation Transfer
Function (MTF) and Noise Power Spectrum (NPS). All CT reconstructions were made using Feldkamp's filtered
backprojection algorithm. The 3D MTF was determined by the analysis of the plane spread function (PlSF) derived
from the surface spread function (SSF) of reconstructed 6.3mm spheres. 3D NPS characterization was performed
through the analysis of a 3D volume extracted from zero-mean CT noise of air reconstructions. The effect of varying
locations on MTF and the effect of different Butterworth filter cutoff frequencies on NPS are reported. Finally, we
present CT images of mastectomy excised breast tissue. Breast specimen images were acquired on our CTMS using an
x-ray technique similar to the one used during performance characterization. Specimen images demonstrate the inherent
CT capability to reduce the masking effect of anatomical noise. Both the quantitative system characterization and the
breast specimen images continue to reinforce the hope that dedicated flat-panel detector, x-ray cone-beam CT will
eventually provide enhanced breast cancer detection capability.
Iodine contrast cone-beam CT imaging of breast cancer
Show abstract
An iodine contrast agent, in conjunction with an X-ray cone beam CT imaging system, was used to clearly image three,
biopsy verified, cancer lesions in two patients. The lesions were approximately in the 10 mm to 6 mm diameter range.
Additional regions were also enhanced with approximate dimensions down to 1 mm or less in diameter. A flat panel
detector, with 194 μm pixels in 2 x 2 binning mode, was used to obtain 500 projection images at 30 fps with an 80 kVp
X-ray system operating at 112 mAs, for an 8-9 mGy dose - equivalent to two view mammography for these women.
The patients were positioned prone, while the gantry rotated in the horizontal plane around the uncompressed, pendant
breasts. This gantry rotated 360 degrees during the patient's 16.6 sec breath hold. A volume of 100 cc of 320 mg/ml
iodine-contrast was power injected at 4 cc/sec, via catheter into the arm vein of the patient. The resulting 512 x 512 x
300 cone beam CT data set of Feldkamp reconstructed ~(0.3 mm)3 voxels were analyzed. An interval of voxel contrast
values, characteristic of the regions with iodine contrast enhancement, were used with surface rendering to clearly
identify up to a total of 13 highlighted volumes. This included the three largest lesions, that were previously biopsied and
confirmed to be malignant. The other ten highlighted regions, of smaller diameters, are likely areas of increased contrast
trapping unrelated to cancer angiogenesis. However the technique itself is capable of resolving lesions that small.
A computer simulation for evaluating dual-energy contrast-enhanced breast tomosynthesis
Show abstract
Previous studies have suggested that iodine contrast-enhanced breast tomosynthesis can be helpful in the
characterization of suspicious abnormalities. Dual-energy, contrast-enhanced breast tomosynthesis involves acquiring
low- and high-energy acquisitions after the administration of the contrast agent, and therefore can simplify the
procedure and reduce the effect of patient motion. In this study, a computer simulation is developed for use in
investigating optimal parameters for dual-energy, contrast-enhanced breast tomosynthesis. The simulation allows for the
selection of various polyenergetic x-ray spectra and x-ray filters, and models x-ray transport through a voxelized breast
phantom, as well as signal and noise propagation through an indirect CsI based imager. A compressed breast phantom
that models the non-uniform parenchymal structure of the breast is used. Irregular lesions were simulated by using a
stochastic growth algorithm. Simulations of dual-energy subtraction, contrast-enhanced breast tomosynthesis show that
using x-ray filters to form quasi-monochromatic high- and low-energy spectra above and below the iodine K-edge can
substantially reduce background structure, and increase lesion conspicuity as compared to single shot contrast-enhanced
BT.
Quantitative flow phantom for contrast-enhanced breast tomosynthesis
Show abstract
The use of contrast agents can help to overcome a lack of intrinsic radiographic contrast between malignant and benign breast tissue by taking advantage of the properties of tumour angiogenesis. Studies of contrast-enhanced mammography have demonstrated increased lesion conspicuity and have shown that this technique provides information on contrast uptake kinetics. It has been suggested that malignant and benign lesions can be differentiated in part by their uptake kinetics, so this additional data may lead to more accurate diagnoses. Tomosynthesis is a 3D x-ray imaging technique that permits lesion depth localization and increased conspicuity in comparison with 2D x-ray projection techniques. This modality, used in combination with contrast agents, promises to be a sensitive method of breast cancer detection. To develop the technique of contrast-enhanced breast tomosynthesis, a dynamic flow phantom has been constructed to provide the same types of imaging challenges anticipated in the clinical setting. These challenges include a low-contrast tumour space, relevant temporal contrast agent uptake and washout profiles, and a need for quantitative analysis of enhancement levels. The design of a flow phantom will be presented that includes a dynamic tumour space, a background that masks the tumour space in images without contrast enhancement, and flow characteristics that simulate tumour contrast agent uptake and washout kinetics. The system is calibrated to relate signal to concentration of the contrast agent using a well plate filled with iodinated water. Iodine detectability in the flow phantom is evaluated in terms of the signal-difference-to-noise ratio for various tomosynthesis image acquisition parameters including number of acquired angular views, angular extent, and reconstruction voxel size.
Dual energy contrast enhanced breast imaging optimization using contrast to noise ratio
Show abstract
The properties of dual energy contrast enhanced breast imaging have been analyzed by imaging a 4 cm breast equivalent phantom consisting of adipose and glandular equivalent plastics. This phantom had superimposed another thin plastic which incorporated a 2 mm deep cylinder filled with iodinated contrast media. The iodine projected thicknesses used for this study was 3 mg/cm2. Low and high energy spectra that straddle the iodine K-edge were used. Critical parameters such as the energy spectra and exposure are discussed, along with post processing by means of nonlinear energy dependent function. The dual energy image was evaluated using the relative contrast to noise ratio of a 2.5 mm x 2.5 mm region of the image at the different iodine concentrations incorporating different breast composition with respect to the noniodinated areas. Optimum results were achieved when the low and high-energy images were used in such a way that relative contrast to noise ratio of the iodine with respect to the background tissue was maximum. A figure of merit suggests that higher noise levels can be tolerated at the benefit of lower exposure. Contrast media kinetics of a phantom incorporating a water flow of 20.4 ml/min through the plastic cylinder suggests that time domain imaging could be performed with this approach. The results suggest that optimization of dual energy contrast enhanced mammography has the potential to lead to the development of perfusion digital mammography.
Breast density mapping based upon system calibration, x-ray techniques, and FFDM images
Show abstract
Clinical studies have correlated a high breast density to a women's risk of breast cancer. A breast density measurement
that can quantitatively depict the volume distribution and percentage of dense tissues in breasts would be very useful for
risk factor assessment of breast cancer, and might be more predictive of risks than the common but subjective and
coarse 4-point BIRADS scale. This paper proposes to use a neural-network mapping to compute the breast density
information based upon system calibration data, x-ray techniques, and Full Field Digital Mammography (FFDM)
images. The mapping consists of four modules, namely, system calibration, generator of beam quality, generator of
normalized absorption, and a multi-layer feed-forward neural network. As the core of breast density mapping, the
network accepts x-ray target/filter combination, normalized x-ray absorption, pixel-wise breast thickness map, and x-ray
beam quality during image acquisition as input elements, and exports a pixel-wise breast density distribution and a
single breast density percentage for the imaged breast. Training and testing data sets for the design and verification of
the network were formulated from calibrated x-ray beam quality, imaging data with a step wedge phantom under a
variety x-ray imaging techniques, and nominal breast densities of tissue equivalent materials. The network was trained
using a Levenberg-Marquardt algorithm based back-propagation learning method. Various thickness and glandular
density phantom studies were performed with clinical x-ray techniques. Preliminary results showed that the neural
network mapping is promising in accurately computing glandular density distribution and breast density percentage.
A novel cone beam breast CT scanner: system evaluation
Show abstract
The purpose of the study is to characterize the imaging performance of the recently built novel cone beam breast CT (CBBCT) scanner. This CBBCT scanner system has one x-ray source and one flat panel detector (Varian's PaxScan 4030CB) mounted on a rotating assembly. A patient table is mounted above the rotating tube/detector assembly. The table has a hole through it that allows a woman's breast to hang pendant in the imaging volume at the rotation axis. The tube/detector assembly rotates around the rotation axis and acquires multiple 2D projection images of the uncompressed breast located at the rotation axis in 10 seconds. Slip ring technology allows continuous rotation of the x-ray tube/detector assembly concentric to the opening in the table to achieve multiple circle scans. Also, it has a controlled vertical motion during the rotation to perform a spiral scan over 20 cm of travel. The continuous 360° rotation is designed to have speeds up to 1 rev/sec. This system was validated through a series of breast-imaging phantom studies and and patient studies. The results show that the image quality of the CBBCT scanner is excellent and all phantom masses (tissue-equivalent carcinomas) and calcifications as well as human subjects' masses, calcifications and abnormalities can be detected faithfully using the CBBCT technique with a glandular dose level less than or equal to that of a single two-view mammography exam. The results indicate that the CBBCT imaging system has much better detectability of small breast tumors compared to the conventional mammography system.
Optimization of image quality in breast tomosynthesis using lumpectomy and mastectomy specimens
Show abstract
The purpose of this study was to determine how image quality in breast tomosynthesis (BT) is affected when acquisition
modes are varied, using human breast specimens containing malignant tumors and/or microcalcifications. Images of
thirty-one breast lumpectomy and mastectomy specimens were acquired on a BT prototype based on a Mammomat
Novation (Siemens) full-field digital mammography system. BT image acquisitions of the same specimens were
performed varying the number of projections, angular range, and detector signal collection mode (binned and nonbinned
in the scan direction). An enhanced filtered back projection reconstruction method was applied with constant
settings of spectral and slice thickness filters. The quality of these images was evaluated via relative visual grading
analysis (VGA) human observer performance experiments using image quality criteria. Results from the relative VGA
study indicate that image quality increases with number of projections and angular range. A binned detector collecting
mode results in less noise, but reduced resolution of structures. Human breast specimens seem to be suitable for
comparing image sets in BT with image quality criteria.
Segmentation-free estimation of volume changes in 3D ultrasound of breast lesion phantoms
Show abstract
Estimation of volume change of structures in response to treatment or growth during breast screening exams is a challenge primarily because of ill-defined boundary. Some treatment procedures alter the lesion completely out of its original shape. In this paper, we present an overview of our recent work on identifying a technique based on Image Volume Based Registration (IVBaR) for estimation of volume. We propose that as long as a region of interest around the lesion can be identified, the exact boundary information would not be necessary. Here, we assume that the surrounding tissue remains nearly unaffected by the treatment procedure, an assumption that is valid in many cases. It is the motion of this tissue in response to changes in the central tumor that would be tracked and used to estimate the change in tumor volume.
Investigation of the use of iodinated contrast agent in a proposed flat-panel CT mammography system
Show abstract
In considering a breast CT system, it is important to note that the spectral attenuation profile of a tumor is very similar to that of fibro-glandular tissue. Preliminary evidence based on imaging breast specimens suggest that the CT number of a malignant breast tumor is very similar to that of surrounding fibro-glandular tissue. Therefore, it is expected that radiologists will probably rely more on tumor morphology to distinguish a malignant tumor from fibro-glandular tissue than an increase in contrast per se. Previous studies have shown that iodinated contrast agents can increase the effective attenuation coefficient yielded by a breast tumor thereby providing increased CT tumor contrast. In order to characterize how the intravenous administration of an iodinated contrast agent can affect the performance of CT breast imaging, a computer simulation of such a system was conducted. The two primary goals of this investigation were first to determine how mean glandular dose, choice of x-ray energy spectrum, and iodine contrast agent density affect tumor detection, and second to determine what effect Compton and Rayleigh scattering have on the variability of the attenuation coefficient yielded by CT mammography. The first goal was achieved by making use of a modified version of the Bakic (Med. Phys. 2003) digital breast phantom to model the uncompressed breast, and a 0.5 cm sphere representing a breast tumor was digitally inserted into the ductal region of this phantom. Several projection sets were generated with the tumor containing various densities of iodine contrast agent, different x-ray energy spectra, and different mean glandular dosage (MGD) levels . Slices through the tumor were extracted from the reconstructions of these projections and were used in human observer studies to determine tumor detectability. The second goal was achieved by using the GATE (Geant 4 Application for Tomographic Emission) Monte-Carlo software package to compute the scattering incident on the flat panel detector for an x-ray projection, then using the aforementioned Bakic phantom, a 0.5 cm sphere representing a breast tumor attenuation and a 3.0 mg/ml of Iodinated contrast agent were inserted at various locations with varying attenuation for 100 projection sets with scatter, and 100 projections without scatter. Histograms of the resulting effective attenuation coefficients yielded by Feldkamp filtered backprojection were plotted and compared.
Novel single x-ray absorptiometry method to solve for volumetric breast density in mammograms with paddle tilt
Serghei Malkov,
Jeff Wang,
John Shepherd
Show abstract
We report on the design and validation of a breast tissue equivalent phantom for automated measurement of
breast composition in film/screen and digital mammography systems. This phantom is a multi-step phantom made of a
single material and containing nine lead positioning markers around its periphery. The markers allow for the phantom
position to be solved relative to the x-ray gantry. The phantom was adhered to the top of the mammographic
compression paddle such that it projected an attenuation image onto the unused corner of the image without overlapping
with CC- nor MLO-view breast edges. The markers and their centroids were identified using automatic morphological
image processing operations. The phantom, and thus the paddle, orientation is then obtained by minimizing a simple
least-square error function of the difference between a pseudo projection image of the phantom markers at known
coordinates and the actual marker image. Fibroglandular-equivalent breast attenuation values were found directly from
step phantom projections. Fat attenuation values were derived from the attenuation coefficient ratios of fat to
fibroglandular tissue. Finally, breast density was calculated by comparing image pixel values to the fat/fibroglandular
references at the same thickness. Multiple scans of a test object (a density step phantom with 7 densities) at nine
different compression thicknesses and six paddle-tilt angles were acquired. We found the precision for determining the
breast thickness to be 0.015 cm (standard deviation) and for determining individual paddle angles to be 0.05 degrees.
Multiple clinical studies using the technique on film/screen and digital mammography machines are also currently under
way.
Breast positioning system for full field digital mammography and digital breast tomosynthesis system
Show abstract
This paper will present a new breast positioning system for amorphous selenium (a-Se) based full field digital mammography (FFDM) system, which is also a platform of tomosynthesis prototype. Clinical images demonstrate that this method is capable extending the breast away from the chest wall, and maximizing the breast volume.
Breast positioning system consists of two transparent moving sheets that apply traction of the breast controlled by motor. Sheets are under and above the compressed breast. Breast positioning sheets pull the breast into the imaging area during the compression. Digital mammography system is based on amorphous selenium flat panel detector (FPD) technology where the overall thickness of the selenium structure is 200 μm, and the pixel size on this detector is 85 μm.
Preliminary results will be presented. Clinical study showed increment of the breast volume imaged, and it brought up to 1.0 cm - 2.0 cm more breast tissue. New breast position system also holds a promise of slight decrement of compression force used in the examination.
Maximizing the exposured breast tissue is complicated, but important aspect in the breast cancer detection and diagnosis. Increasing the field of view with an additional volume of breast tissue imaged is a key point in digital mammography and
digital breast tomosynthesis (DBT).
Analysis of patient bed positioning in SPECT-CT imaging for dedicated mammotomography
Show abstract
Patient positioning on a bed is an integral part of accurate imaging for dedicated 3D breast imaging. For both dedicated
breast SPECT (single photon emission computed tomography) and breast CT (computed tomography or computed
mammotomography, CmT) which are under development in our lab, maximum access to the breast in the imaging
system's field of view is required to obtain the largest imaged breast volume. Accurate bed positioning will be necessary
as it may be integrated with a guided biopsy apparatus. Thus, a patient bed with flexible 3D positioning capability is
being integrated into the various independent and hybrid 3D imaging systems. The customized bed has both manual and
computer controlled positioning capability, and the accuracy and reproducibility of the system are being characterized.
Computer controlled positioning and feedback provide seemingly reproducible results. However, gross movements may
vary in their accuracy to the given input position. While linear with slopes near 1.0 and intercepts near 0.0cm, lateral
(Y) movement translates less than the input amount, while axial (X) movement translates farther than the input amount.
Vertical (Z) directional movement follows a quadratic shift with a small dc component with or without added weight on
the table. A variety of patient imaging conditions along with x-ray image data are evaluated to demonstrate the
reproducibility of positioning accuracy. Individual directional repositioning accuracy is found to be better than multiple,
combined directional repositioning accuracy. Imaging results indicate a reproducibility (error) of less than 1mm, which
may be suitable for SPECT imaging but perhaps not for higher resolution dedicated breast CT. However, for the
independent SPECT system, bed motion is not necessary because the detector's line of sight can already acquire data at
the chest wall.
Glandular segmentation of cone beam breast CT volume images
Show abstract
Cone beam breast CT (CBBCT) has potential as an alternative to mammography for screening breast cancer while limiting the radiation dose to that of a two-view mammogram. A clinical trial of CBBCT has been underway and volumetric breast images have been obtained. Although these images clearly show the 3D structure of the breast, they are limited by quantum noise due to dose limitations. Noise from these images adds to the challenges of glandular/adipose tissue segmentation. In response to this, an automated method for reducing noise and segmenting glandular tissue in CBBCT images was developed.
A histogram based 2-means clustering algorithm was used in conjunction with a seven-point 3D median filter to reduce quantum noise. Following this, a 2D parabolic correction was applied to flatten the adipose tissue in each slice to reduce system inhomogeneities. Finally, a median smoothing algorithm was applied to further reduce noise for optimal segmentation.
The algorithm was tested on actual breast scan volume data sets for subjective analysis and on a 3D mathematical phantom to test the algorithm.
Subjective comparison of the actual breast scans with the denoised and segmented volumes showed good segmentation with little to no noticeable degradation. The mathematical phantom, after denoising and segmentation, was found to accurately measure the percent glandularity within 0.03% of the actual value for the phantom containing larger spherical shapes, but was only able to preserve small micro-calcification sized spheres of 0.8 and 1.0 mm, and small fibers with diameters of 1.2 and 1.4 mm.
Evaluation of a new breast-shaped compensation filter for a newly built breast imaging system
Show abstract
A new breast-shaped compensation filter has been designed and fabricated for breast imaging using our newly built
breast imaging (CBCTBI) system, which is able to scan an uncompressed breast with pendant geometry. The shape of
this compensation filter is designed based on an average-sized breast phantom. Unlike conventional bow-tie
compensation filters, its cross-sectional profile varies along the chest wall-to-nipple direction for better compensation for
the shape of a breast. Breast phantoms of three different sizes are used to evaluate the performance of this compensation
filter. The reconstruction image quality was studied and compared to that obtained without the compensation filter in
place. The uniformity of linear attenuation coefficient and the uniformity of noise distribution are significantly improved,
and the contrast-to-noise ratios (CNR) of small lesions near the chest wall are increased as well. Multi-normal image
method is used in the reconstruction process to correct compensation flood field and to reduce ring artifacts.
Evaluation of physical image characteristics of phase contrast mammography
Show abstract
Recently, the system called PCM (phase contrast mammography) came to be applied to clinical examinations. The PCM images are acquired as 1.75x-magnified images using small focal spot, then reduced to real object's size by image processing. The PCM images had the feature that edges of objects were enhanced. It was reported that the edges were enhanced because of refraction of X rays. We measured physical image characteristics of the PCM, to compare image characteristics of the PCM with those of a conventional mammography. Specifically, response characteristics to objects and noise characteristics in the spatial frequency domain were measured. As the results, compared with the conventional mammography, response characteristics of the PCM were high. If the placement conditions of objects were changed, response characteristics were also changed. And if the shape of objects was changed, response characteristics also were changed. Noise characteristics of the PCM were better than those of the conventional mammography. Furthermore, in order to investigate why edges of objects in the PCM images were enhanced, we simulated image profiles which would be obtained if X rays were refracted. And, we found the possibility that edge enhancements of the PCM images were based on refraction of X rays.
Digital breast tomosynthesis geometry calibration
Show abstract
Digital Breast Tomosynthesis (DBT) is a 3D x-ray technique for imaging the breast. The x-ray tube, mounted on a gantry, moves in an arc over a limited angular range around the breast while 7-15 images are acquired over a period of a few seconds. A reconstruction algorithm is used to create a 3D volume dataset from the projection images. This procedure reduces the effects of tissue superposition, often responsible for degrading the quality of projection mammograms. This may help improve sensitivity of cancer detection, while reducing the number of false positive results.
For DBT, images are acquired at a set of gantry rotation angles. The image reconstruction process requires several geometrical factors associated with image acquisition to be known accurately, however, vibration, encoder inaccuracy, the effects of gravity on the gantry arm and manufacturing tolerances can produce deviations from the desired acquisition geometry. Unlike cone-beam CT, in which a complete dataset is acquired (500+ projections over 180°), tomosynthesis reconstruction is challenging in that the angular range is narrow (typically from 20°-45°) and there are fewer projection images (≈7-15). With such a limited dataset, reconstruction is very sensitive to geometric alignment. Uncertainties in factors such as detector tilt, gantry angle, focal spot location, source-detector distance and source-pivot distance can produce several artifacts in the reconstructed volume. To accurately and efficiently calculate the location and angles of orientation of critical components of the system in DBT geometry, a suitable phantom is required.
We have designed a calibration phantom for tomosynthesis and developed software for accurate measurement of the geometric parameters of a DBT system. These have been tested both by simulation and experiment. We will present estimates of the precision available with this technique for a prototype DBT system.
A new approach to digital breast tomosynthesis for breast cancer screening
Show abstract
Digital breast tomosynthesis (DBT) is being proposed as a replacement for conventional mammography for breast cancer screening. However, there are limitations to DBT that reduce its effectiveness for screening, principally, difficulty in imaging microcalcifications and increased reading times by radiologists. We propose a method to overcome these limitations.
Our proposed method is to divide the total dose given to the patient unequally such that one projection uses at least half of the dose and the remaining dose is divided over the remaining projections. We assume that in screening with DBT, only a single view is obtained using twice the dose of a conventional mammogram. All the projection images are used in the reconstruction. The 2D projection image that received the highest dose is analyzed by a computer-aided detection (CADe) scheme for microcalcifications. The radiologist views the 3D image set, with mass CADe, principally to search for masses and the 2D image to search for clustered microcalcifications with CADe. Since the 3D image set is for mass detection, the image can be reconstructed using larger sized pixels. This will reduce computation time and image noise. In principle, radiologists can review the tomosynthesis slices faster since they do not have to search for microcalcifications.
We believe that by producing both a high resolution, "standard" dose 2D image and a lower resolution 3D image set, both calcifications and masses can be optimally imaged and detected in a time efficient manner.
Development of a model for breast tomosynthesis image acquisition
Show abstract
Tomosynthesis is emerging as a promising modality for breast imaging. Several manufacturers have developed prototype
units and have acquired clinical and phantom data. Scanning configurations of these prototypes vary. So far, studies
relating scanning configuration to image quality have been limited to those geometries that could be implemented on a
particular prototype. To overcome this limitation, we are developing a model of breast tomosynthesis image acquisition
system, which models the formation of the x-ray image and x-ray detector.
The x-ray image of an object is computed analytically for a polychromatic x-ray beam. Objects consist of volumetric
regions that are bounded by either a planar, ellipsoidal, cylindrical or conical surface, allowing for a variety of objects. xray
scatter is computed by convolving the image with a scatter point-spread function. Poisson noise according to the
entrance exposure is added to the image.
The x-ray detector in this model is composed of a phosphor screen followed by a detector array. X-ray interactions in the
screen are modeled as depth-dependent. The optical output of the screen is converted into digital units using a gain factor
which was assumed to be Gaussian distributed.
To validate this data model, we acquired images of a contrast-detail phantom on a stereotactic biopsy unit. The x-ray
source is mounted on an arm that pivots in a plane about the detector center. The x-ray detector consists of a Min-R type
screen fiber-optically coupled to a CCD camera.
To compare actual and simulated data, we compared line profiles as well as several automatically extracted image
features such as contrast-to-noise ratio, contrast, area and radial gradient index. Good agreement was found between
simulation and physical data, indicating that we can now use this model to explore image quality for various
tomosynthesis scanning configurations.
Poster Session: Innovative Imaging
Super-resolution ultrasound tomography: a preliminary study with a ring array
Show abstract
Ultrasound tomography attempts to retrieve the structure of an object by exploiting the interaction of acoustic
waves with the object. A fundamental limit of ultrasound tomography is that features cannot be resolved if they
are spaced less than λ/2 apart, where λ is wavelength of the probing wave, regardless of the degree of accuracy of
the measurements. Therefore, since the attenuation of the probing wave with propagation distance increases as
λ decreases, resolution has to be traded against imaging depth. Recently, it has been shown that the λ/2 limit is
a consequence of the Born approximation (implicit in the imaging algorithms currently employed) which neglects
the distortion of the probing wavefield as it travels through the medium to be imaged. On the other hand, such a
distortion, which is due to the multiple scattering phenomenon, can encode unlimited resolution in the radiating
component of the scattered field. Previously, a resolution better than λ/3 has been reported in these proceedings
[F. Simonetti, pp. 126 (2006)] in the case of elastic wave probing. In this paper, we demonstrate experimentally
a resolution better than λ/4 for objects immersed in a water bath probed by means of a ring array which excites
and detects pressure waves in a full view configuration.
Development of qualitative near infrared vascular imaging system with tuned aperture computed tomography
Show abstract
We developed a novel system for imaging and qualitative analysis of surface vessels using near infrared (NIR) radiation with tuned aperture computed tomography (TACT®), even if the NIR cannot transmit through thick regions. NIR-sensitive CCD camera was surrounded by sixty light emitting diodes (alternating wavelengths of 700 nm and 810 nm), and could only detect the NIR from the subcutaneous tissue. We obtained multiple near infrared projections of surface vessels at each wavelength in accordance with the optical aperture theory within one second. Then, we created tomograms using the TACT program, and determined the venous oxygenation index (VOI), which reflected the oxygen saturation level, calculated from the image signals at each wavelength. This system produced thinner NIR tomograms under 0.5 mm. The change in VOI after load test calculated from NIR tomograms was more sensitive than that from NIR images without tomography. Our novel system makes it possible to non-invasively obtain NIR tomograms and accurately analyze changes in oxygen saturation.
Adaptive MOEMS mirrors for medical imaging
Show abstract
This paper presents micro-electro-mechanical-systems (MEMS) optical elements with high angular deflection arranged in arrays to perform dynamic laser beam focusing and scanning. Each element selectively addresses a portion of the laser beam. These devices are useful in medical and research applications including laser-scanning microscopy, confocal microscopes, and laser capture micro-dissection. Such laser-based imaging and diagnostic instruments involve complex laser beam manipulations. These often require compound lenses and mirrors that introduce misalignment, attenuation, distortion and light scatter. Instead of using expensive spherical and aspherical lenses and/or mirrors for sophisticated laser beam manipulations, we propose scalable adaptive micro-opto-electro-mechanical-systems (MOEMS) arrays to recapture optical performance and compensate for aberrations, distortions and imperfections introduced by inexpensive optics. A high-density array of small, individually addressable, MOEMS elements is similar to a Fresnel mirror. A scalable 2D array of micro-mirrors approximates spherical or arbitrary surface mirrors of different apertures. A proof of concept prototype was built using PolyMUMPTM due to its reliability, low cost and limited post processing requirements. Low-density arrays (2x2 arrays of square elements, 250x250μm each) were designed, fabricated, and tested. Electrostatic comb fingers actuate the edges of the square mirrors with a low actuation voltage of 20 V - 50 V. CoventorWareTM was used for the design, 3D modeling and motion simulations. Initial results are encouraging. The array is adaptive, configurable and scalable with low actuation voltage and a large tuning range. Individual element addressability would allow versatile uses. Future research will increase deflection angles and maximize reflective area.
Energy and dose considerations for diffraction enhanced CT in small animal studies
Show abstract
Diffraction enhanced imaging (DEI) uses monochromatic x-rays coupled to an analyzer crystal to extract information
about the refraction of x-rays within the object. Studies of excised biological tissues show that DEI has significant
contrast-to-noise ratio (CNR) advantages for soft tissue when compared to standard radiography. DEI differs from
conventional CT in that its refraction contrast depends on x-ray energy as 1/E, thus the energy and dose considerations
for conventional CT will be inappropriate. The goal of this study was to assess the optimal energy for in vivo CT
imaging of a mouse head to obtain the largest soft tissue refraction CNR. Through a theoretical model, optimum
refraction CNR for mouse brain imaging was found to be about 20 keV. The findings were tested experimentally using
the DEI system at the X15A beamline of the National Synchrotron Light Source. Using the parameters for optimized
refraction CNR (20 keV, silicon [333] reflection), large image artifacts were caused by DEI's scatter-rejection
properties. By increasing the x-ray energy and using a lower order diffraction, silicon [111], soft tissue features within
the brain, including the hippocampus, could be resolved.
Characterization of a novel microCT detector for small animal computed tomography (CT)
Show abstract
The development of new small animal imaging techniques such as high-speed microCT and low-dose microCT often
requires investigating optimal detector parameters and imaging techniques. This paper presents an approach to develop
a low-dose microCT detector based on a novel, back-thinned, back-illuminated electron multiplying CCD (EMCCD)
coupled to a high performance microcolumnar CsI scintillator via a fiberoptic taper. Our goal is to achieve high
DQE(0) for X-ray energies typically used in small animal imaging (40 to 80 kVp), providing high quality imaging at
substantially reduced dose.
Towards achieving this goal we have developed a novel EMCCD camera fitted with a fiberoptic window. To
enhance the imaging area we fabricated additional fiberoptic tapers measuring 3:1 and 6:1 in demagnification ratio,
mechanically coupled to the EMCCD.
The high sensitivity and internal gain of the EMCCD is further exploited in our system design by the use of a
thick CsI screen. These screens not only provide higher absorption for 40 to 80 kVp X-rays, but even at ~200 &mgr;m
thickness maintain a high resolution of up to 11 lp/mm.
This paper outlines the quantitative performance of each detector component and the detector as a whole.
While the detector demonstrated the potential for achieving the targeted DQE performance, it also showed that
mechanical coupling of the tapers to the CCD results in unacceptable light loss, and that direct CCD-to-taper bonding
and using new versions of large-area EMCCD chips would be better options.
Dual-energy cone-beam micro-CT for animal imaging: preliminary study
Show abstract
We have recently developed a prototype cone-beam micro-CT system for studying new image reconstruction
algorithms experimentally as well as for small animal imaging. Dual energy methods have widely been investigated
and applied for eliminating beam hardening artifacts in the reconstructed images and for performing quantitative CT.
We have proposed earlier a pre-reconstruction-type dual energy method which has the advantage that the cone-beam
CT data from two measurements with different x-ray energy spectra are first processed to obtain quantities that are
approximately consistent with the x-ray transform, and then image reconstruction is performed. In this work, the
proposed method is applied to the cone-beam micro-CT data acquired by our prototype system. Only the thickness
maps of the object based on projection images are shown in this work as a preliminary.
A system model for pinhole SPECT simulating edge penetration, detector, and pinhole response and non-uniform attenuation
Show abstract
Small animal SPECT using low energy photons of I-125 and approaching resolutions of microscopic
levels, imaging parameters such as pinhole edge penetration, detector blur, geometric response, detector
and pinhole misalignment, and gamma photon attenuation and scatter can have increasingly noticeable
and/or adverse effects on reconstructed image quality. Iterative reconstruction algorithms, the widelyaccepted
standard for emission tomography, allow modeling of such parameters through a system
matrix. For this Monte Carlo simulation study, non-uniform attenuation correction was added to the
existing system model. The model was constructed using ray-tracing and further included corrections
for edge penetration, detector blur, and geometric aperture response. For each ray passing through
different aperture locations, this method attenuates a voxel's contribution to a detector element along
the photon path, which is then weighted according to a pinhole penetration model. To lower the
computational and memory expenses, symmetry along the detector axes and an incremental storage
scheme for the system model were used. For evaluating the nonuniform attenuation correction method,
3 phantoms were designed of which projection images were simulated using Monte Carlo methods. The
first phantom was used to examined skin artifacts, the second to simulate attenuation by bone, and the
third to generate artifacts of an air-filled space surrounded by soft tissue. In reconstructions without
attenuation correction, artifacts were observed with up to a 40% difference in activity. These could
be corrected using the implemented method, although in one case overcorrection occurred. Overall,
attenuation correction improved reconstruction accuracy of the radioisotope distribution in the presence
of structural differences.
Analytical deconvolution for improvement in spatial resolution of the In-111 coincidence camera
Show abstract
The In-111 coincidence camera we previously proposed can significantly increase detection efficiency because collimators are no longer needed. However, the initial simulations indicated that spatial resolution was too poor for medical imaging. To improve the resolution, we derived an analytical deconvolution algorithm in this study. In the derivation, 1-D Fourier transform for the shift-invariant point spread function (PSF) with respect to the detector bin location t was carried out analytically. The Fourier transform is approximately a linear function of the source-to-detector distance s when s is greater than 5 cm and its variation over s is much slower than that of any extensive source. The Fourier transform of the PSF can thus be taken out of the integration over s with reasonable accuracy and its inversion is the deconvolution kernel. A low-pass filter was applied to the deconvolved Fourier transform to suppress high-frequency oscillation. Applying the derived deconvolution algorithm to computer simulated phantoms, we achieved a resolution of 2 cm for s = 10 cm. Compared to the pre-deconvolution resolution of 19 cm, this is a huge improvement but is still poor. The errors caused by the approximations made in the derivation can be further reduced and also the high-frequency behavior of the deconvolved Fourier transform can be improved using better deconvolution techniques. Monte Carlo simulations for more realistic sources with image noise should be performed for further evaluation.
Mean absorbed dose to mouse in micro-CT imaging with an ultrafast laser-based x-ray source
Show abstract
We have investigated theoretically the mean absorbed dose to the mouse in our newly constructed, in-line
holography, x-ray phase-contrast, in-vivo, micro-CT system with an ultrafast laser-based x-ray (ULX) source. We
assumed that the effective mouse diameter was 30 mm and the x-ray detector required minimum 30 μGy per frame
to produce high quality images. The following laser target-filter combinations were considered: Ag-Ag, Mo-Mo, Sn-
Sn. In addition, we considered narrow-pass multilayer x-ray mirrors. The corresponding ULX spectra were obtained
using a CZT solid-state spectrometer. The approach used for dose computation was similar to human dose
estimation. The mouse was modeled as a tissue-equivalent cylinder located at the isocenter with diameter 30 mm
and density 1g/cm3. A layer of dermis (skin and fur) with 1 mm thickness was also modeled. Imparted energy per
volume was estimated for 1 keV wide x-ray energy intervals in the 6-100 keV range. Monte Carlo simulations were
performed using the SIERRA code previously validated using 30 mm diameter PMMA phantom. The results
obtained indicate that: a) the mean absorbed dose for ULX is less than or equal to that from a W-anode micro-CT
tube operating at 30-40 kVp with 0.5 or 1.0 mm Al; b) for filter thickness above 100 μm, Sn-Sn results in the
highest dose, followed by Ag-Ag and Mo-Mo; c) the multilayer x-ray mirror with FWHM ≤ 10 keV produces
significantly lower dose than metallic foil filters. We conclude that ULX can provide better dose utilization than a
microfocal x-ray tube for in vivo microtomography applications.
Evaluation of frequency multiplexing radiography based on multi-pixel x-ray technology
Show abstract
We have recently demonstrated the feasibility of frequency multiplexing radiography (FMR) technique based on the frequency division multiplexing (FDM) principle and the carbon nanotube field emission x-ray technology. The key component of the FMR technique is a multi-pixel carbon nanotube field emission x-ray source. The prototype multi-pixel x-ray source has a linear array of nine field emission x-ray pixels. By programming the control electronics, the multi-pixel x-ray source can generate spatially and temporally modulated x-ray radiation. During the multiplexing imaging process, all the x-ray pixels were turned on simultaneously with each beam modulated at different frequency. The superimposed x-ray signals generated by the multi-pixel x-ray source were captured using a high speed flat panel x-ray detector over a certain period of time. The collected composite images were then demultiplexed using a Fourier transform based algorithm to recover the original nine projection images from different view angles. The FMR technique can in principle increase the imaging speed and reduce the x-ray peak workload for applications such as computed tomography (CT). In this paper we evaluated the performance of this new radiographic imaging technique based on our simulation and experiment results. Imaging artifacts caused by the cross-talk among different frequency subchannels have been studied and the importance of orthogonal frequency division multiplexing (OFDM) has been demonstrated.
Modeling and testing of a non-standard scanning device with dose reduction potential
Show abstract
A non-standard scanning device with dose-reduction potential was proposed at the SPIE Medical Imaging conference
2006. The new device obtains the Radon data after the X-ray beam is collimated through a special
mask. This mask is combined with a new geometry that permits an efficient data collection, thus the device
has the potential of reducing the dose by a factor of two. In this work, we report a prototype of the new device
and experimental data acquisition using only the mask of the new scanning geometry. In order to obtain the
optimal parameters for the scanning device, several factors have been considered, including detector elements
and shielding shape, fan beam angle, speed of the source rotation and materials employed. The calibration of the
detector elements needs especial attention, due to the dependence of the detector response on the energy of the
X-rays. A simplfied version of the device was designed and mounted. Phantom data were acquired using this
prototype and were used to test the performance of the new design. The results obtained are highly promising,
even though the prototype developed does not make use yet of all the potential features proposed in the theory.
Imaging with Iridium photons: an application in brachytherapy
Show abstract
In external-beam radiotherapy efforts are currently devoted to research on image-guided verification techniques. In brachytherapy the situation is far less advanced; usually, there is no treatment verification imaging. We are studying the possibility to use the photons emitted from a conventional 192Ir brachytherapy source for High Dose Rate (HDR) treatments, when inserted in a patient. We investigated whether the images can be used for dose delivery verification and to interrupt faulty dose deliveries. Phantoms were built to accommodate a remote controlled HDR 192Ir source. Images were collected with an x-ray intensifier, and predicted from calculations based on ray-tracing. For a bone/tissue/air/lung phantom with the source on top of the phantom measured contrasts were 8% (bone/tissue), 19% (tissue/lung) and 26% (lung/bone). When a thick Lucite slab was added on top of the contrast phantom, the contrasts decreased to 3, 7 and 10%, respectively, indicating that phantom scatter is an important issue. Differences between measured and simulated images and the influence of scatter were quantified. From this feasibility study it is concluded that imaging with 192Ir photons is possible but that work on scatter rejection through simulation and anti-scatter grids is needed.
Poster Session: Detectors
Photon counting pixel architecture for x-ray and gamma-ray imaging applications
Show abstract
Photon counting is emerging as an alternative detection technique to conventional photon integration. In photon counting systems, the value of each image pixel is equal to the number of photons that are absorbed by the radiation detector. The proposed pixel architecture provides a method for energy windowing and serial readout for low-dose gamma-ray imaging. Each pixel is comprised of a radiation detector and integrated analog and digital circuitry. A prototype was developed on a printed circuit board (PCB) using discrete electronic components. In this research, we present the experimental results for the operation of the photon counting pixel with energy windowing and investigate the compromise between pixel noise level and photon count rate.
Amplified pixel sensor architectures for low dose computed tomography using silicon thin film technology
Show abstract
Cone beam computed tomography (CBCT) has been recently reported using flat panel imagers (FPI). Here, detector
technology capable of high speed imaging, high spatial resolution, large volume coverage, better contrast resolution and,
in particular, lowered patient dose is required. Employing active matrix flat panel imagers (AMFPIs) as cone beam CT
detectors has been proposed as a solution for improving volume coverage, contrast and resolution; however, clinical
evaluations have shown that they suffer from low speed read out.
Unlike passive pixel architecture which is currently the state-of-the-art technology for AMFPIs, our preliminary studies
have shown that novel amplified pixel sensor (APS) architectures can overcome the low readout speed, and moreover,
they provide gain which can be traded for higher frame rate and lower X-ray doses. Although APS architectures can
meet the high dynamic range and low noise requirements of CT imaging, linearity and variations between pixel
characteristics are major issues. In this study we will investigate novel APS architectures to address these concerns.
Multidetector-row CT with a 64–row amorphous silicon flat panel detector
Show abstract
A unique 64-row flat panel (FP) detector has been developed for sub-second multidetector-row CT (MDCT). The intent
was to explore the image quality achievable with relatively inexpensive amorphous silicon (a-Si) compared to existing
diagnostic scanners with discrete crystalline diode detectors. The FP MDCT system is a bench-top design that consists
of three FP modules. Each module uses a 30 cm x 3.3 cm a-Si array with 576 x 64 photodiodes. The photodiodes are
0.52 mm x 0.52 mm, which allows for about twice the spatial resolution of most commercial MDCT scanners. The
modules are arranged in an overlapping geometry, which is sufficient to provide a full-fan 48 cm diameter scan. Scans
were obtained with various detachable scintillators, e.g. ceramic Gd2O2S, particle-in-binder Gd2O2S:Tb and columnar
CsI:Tl. Scan quality was evaluated with a Catphan-500 performance phantom and anthropomorphic phantoms. The FP
MDCT scans demonstrate nearly equivalent performance scans to a commercial 16-slice MDCT scanner at comparable
10 - 20 mGy/100mAs doses. Thus far, a high contrast resolution of 15 lp/cm and a low contrast resolution of 5 mm @
0.3 % have been achieved on 1 second scans. Sub-second scans have been achieved with partial rotations. Since the
future direction of MDCT appears to be in acquiring single organ coverage per scan, future efforts are planned for
increasing the number of detector rows beyond the current 64- rows.
Comparison of multi-arm VRX CT scanners through computer models
Show abstract
Variable Resolution X-ray (VRX) CT scanners allow imaging of different sized anatomy at the same level of
detail using the same device. This is achieved by tilting the x-ray detectors so that the projected size of the
detecting elements is varied producing reconstructions of smaller fields of view with higher spatial resolution.1
The detector can be divided in two or more separate segments, called arms, which can be placed at different
angles, allowing some flexibility for the scanner design. In particular, several arms can be set at different angles
creating a target region of considerably higher resolution that can be used to track the evolution of a previously
diagnosed condition, while keeping the patient completely inside the field of view (FOV).2 This work presents
newly-developed computer models of single-slice VRX scanners that allow us to study and compare different
configurations (that is, various types of detectors arranged in any number of arms arranged in different geometries)
in terms of spatial and contrast resolution. In particular, we are interested in comparing the performance
of various geometric configurations that would otherwise be considered equivalent (using the same equipment,
imaging FOVs of the same sizes, and having a similar overall scanner size). For this, a VRX simulator was
developed, along with mathematical phantoms for spatial resolution and contrast analysis. These tools were
used to compare scanner configurations that can be reproduced with materials presently available in our lab.
Effect of multiple dopants on the quantum efficiency of LiF thermoluminescent dosimeters (TLD) and BaFX (X = Br,Cl,I) storage phosphors
Show abstract
The radiant quantum efficiency (RQE) of x-ray phosphors is defined as the ratio of the emitted luminescent power and
the power absorbed by the material. For a given x-ray imaging technique, a high RQE phosphor means a lower x-ray
exposure to patients and a better image quality. To improve RQE, phosphors such as LiF and BaFX:Eu ( X = Br, Cl, I)
host lattices, which are commonly used in medical dosimetry and imaging system, are always doped with Cu, Ti, or Tb.
Experimental observations showed that these dopants can increase phosphor RQE significantly. In this study, we
theoretically investigated the effect of additional dopants on the RQE of LiF:Mg and BaFX:Eu host lattices using
Density Functional Theory (DFT) in the Local Density Approximation (LDA). Self-consistent charge density
calculations were performed. The energy loss function L(w) was obtained and used to calculate the RQE for different
phosphors. The results showed that additional dopants produced changes in the optical properties of the phosphors,
particularly the energy loss function L(w). Doping with more substitutional impurities increased the RQE of all host
lattices except the BaFI lattice where the RQE decreased.
Evaluation of Moire artifacts with stationary anti-scatter grids in amorphous selenium-based flat panel x-ray detector system
Show abstract
In digital imaging systems, using anti-scatter grids may lead to arise moire artifacts. Recently, amorphous seleniumbased
(direct-conversion) flat panel X-ray detector systems were developed. An important advantage of a-Se is its high
spatial resolution. However, the high resolution of a-Se potentially introduces more moire artifacts.1 The aim of present
study was to choice optimal anti-scatter grids in amorphous selenium-based flat panel X-ray system, and to demonstrate
how to arise moire artifacts. We simulated the sampling process in the spatial domain equivalent to the sampling aperture
function in the spatial frequency domain. Moire patterns appeared with the different period and contrast. The period and
the contrast of moire artifacts were varied with combinations of the sampling conditions and the strip density of antiscatter
grids.
A simple all-digital PET system
Show abstract
Positron emission tomography~(PET) systems employ mixed-signal front-end to carry out relatively simple, and ad hoc, processing of the charge pulses generated upon event detection. To obtain, and maintain over time, proper calibrations of the mixed-signal circuitry for generating accurate event information is a challenging task due to the simplicity of the event processing, and the huge number of channels and multiplexing of the input signals found in modern PET systems. It is also difficult to modify or extend the event-processing technologies when needs arise because it would involve making changes to the circuitry. These limitations can be circumvented by applying digital signal-processing technologies for analyzing event pulses generated in PET. With digital technologies, optimized event-processing algorithms can be implemented and they can be modified or extended with ease when needed. The resulting PET data-acquisition (DAQ) system is easier to calibrate and maintain, can generate more accurate event information, and has better extendibility. In this paper, we present our work toward developing a scalable all-digital DAQ system for PET, built upon a personal-computer platform for reducing cost. We will present the overall architecture of this digital DAQ system, and describe our implementations of several components of the system.
Comparison of compound semiconductor radiation films deposited by screen printing method
Show abstract
Flat-panel direct conversion detectors used in compound substance of semiconductor are being studied for digital x-ray imaging. Recently, such detectors are deposited by physical vapor deposition(PVD) generally. But, most of semiconductors (HgI2, PbI2, TlBr, PbO) deposited by PVD method have shown difficult fabrication process and instability for large area x-ray imaging. Consequently, in this paper, we propose applicable potentialities for screen printing method that is coated on a substrate easily. It is compared to electrical properties among semiconductors such as HgI2, PbI2, PbO, CdTe under investigation for direct conversion detectors. Each film detector was coated onto the substrate and Substrates of 2cmx5cm had been used to evaluate performance of semiconductor radiation detectors. Fabricated films consisted of ~25 to 35 μm thick layer of semiconductor. Dark current, sensitivity, linearity, lag, and morphologic property were measured for evaluation of films performance. Dark current of PbO was acquired the lowest, and dark current of HgI2 at the operation voltage of ~1V/μm was observed 8pA/mm2. Sensitivity is observed higher about ten times than the others. And then HgI2 is observed the best SNR in four materials. In four semiconductors, it is shown in good linearity. Such a value is not better than PVD process, but it is easy to be fabricated in high quality for large area X-ray Imaging. Our future efforts will concentrate on optimization of growth of film thickness that is coated onto a-Si TFT array.
Comparison in image quality and noise component of columnar phosphor plate and powder phosphor plate
Show abstract
For Computed Radiography (CR) systems that use a columnar phosphor plate (CPP) and a powder phosphor plate (PPP), we designed the systems to obtain the best image quality. To determine the optimum phosphor layer thickness for each phosphor plate, the relationship between the intensity and spatial spread of photo-stimulated luminescence (PSL), and the phosphor layer thickness of the phosphor plate is quantitatively clarified. Next, to determine the stimulation light intensity, we measured PSL, modulation transfer function (MTF) and detective quantum efficiency (DQE) by varying the stimulation light intensity, using the determined optimum phosphor layer thickness. We also investigated the noise components of each phosphor plate. Results show that, compared to the PPP, the CPP is more favorable in allowing thicker phosphor layer without reduction in MTF. As the result of the relationship between the layer thickness and the PSL, noise analysis, it was confirmed that the CPP could detect PSL in the deep region of the phosphor layer without reducing the intensity of PSL. This suggests that in comparison to the PPP, the CPP can make efficient use of X-ray information, thereby promising to enhance image quality and to reduce exposure dose.
Image quality of the front exposure system and the back exposure system in the indirect (x-ray-to-light conversion) digital radiography system
Show abstract
In X-ray-to-light conversion digital radiography, we compared the image quality of a system in which photodetection is done from the X-ray incident surface (hereafter referred to as a front exposure system) and a system in which photodetection is done from the back side opposite the X-ray incident surface (hereafter referred to as a back exposure system). Modulation transfer function (MTF) and detective quantum efficiency (DQE) measurements were performed using the method IEC prescribes. Both MTF and DQE were higher with the front exposure system than with the back exposure system, with the former delivering better image quality. This difference can be accounted for by differences in the distribution of absorbed X-ray doses in the phosphor layer, the readout efficiency, which varies as a function of depth in the phosphor layer, and depth-dependent blurs of light. Furthermore, we determined changes in image quality incurred by varying the quality of X-rays, the thickness of the phosphor layer and the crystal structure of phosphors. The advantage of the front exposure system becomes more pronounced with decreasing X-ray tube voltage, increasing phosphor layer thickness, and the use of phosphors in powder form.
Fluorozirconate-based glass-ceramic storage phosphors for digital mammography
Show abstract
We developed a class of (semi-) transparent glass-ceramic storage phosphors for digital mammography. The glass ceramics are based on europium-doped fluorozirconate glasses, which were additionally doped with chlorine to initiate the nucleation of barium chloride nanoparticles therein. The glass ceramic is able to convert ionizing radiation into stable electron-hole pairs, which can be read out afterwards with a scanning laser beam in a so-called "photostimulated luminescence" (PSL) process. A number of experiments were done to measure materials and engineering parameters relevant to a point scanning readout system, and to allow projection of the Detective Quantum Efficiency (DQE) for the proposed x-ray storage phosphor system. These included measurement of the required stimulating exposure (laser power density times pixel dwell time), and integrated PSL signal (or "gain", expressed as the number of detected electrons per absorbed x-ray). Measurements of optical light spreading of the stimulating laser light were also done, since this effect determines the MTF of the scanning system. Calculations of x-ray absorption vs. imaging plate composition and thickness, and x-ray beam spectrum, were also completed. Finally, the measured parameters were used to project DQE vs. spatial frequency for the proposed detector, and to compare with commercially available electronic mammography systems.
A new x-ray imaging technique for radiography mode of flat-panel imager
Show abstract
A digital radiography system using a flat-panel imager, which has a novel imaging technique for a radiography mode, has been developed. A radiographic image captured by the new imaging technique has a wide dynamic range in comparison with conventional radiographic images. The purpose of this presentation is to show the basic performance of the image quality acquired by the new imaging technique and compare it with an image taken by a conventional technique.
The flat-panel imager has a gain switching capability, normally used in a dynamic imaging mode for a cone-beam CT study. The gain switching method has two gain settings and switches between them automatically, depending on the incident dose to each pixel of flat-panel imager. As a result of the gain switching method, an image having wide dynamic range is achieved. In this study, we applied the gain switching method to the radiography mode, and achieved a radiographic image with wider dynamic range than a conventional radiograph. Furthermore, we have also developed an algorithm for calibration of the gain switching method in radiography mode.
Poster Session: Performance Assessment
A new paradigm in portal imaging QA: fast measurements of modulation transfer function (MTF) and detective quantum efficiency (DQE) using line-pair bar patterns
Show abstract
In portal imaging, the role of electronic portal imaging devices (EPIDs) to implement complex radiation therapy
protocols is crucial, and regular quality assurance (QA) of EPID image quality is necessary to ensure treatment efficacy.
The modulation transfer function (MTF), noise power spectrum (NPS) and detective quantum efficiency (DQE) are
universal metrics that can completely describe radiographic image quality. These metrics are independent of
measurement geometry and the user, and allow intra and inter detector performance evaluation. Though NPS and DQE
calculations are straightforward, conventional MTF measurements using slit/edge phantoms are difficult and time
consuming. Therefore, such measurements are not feasible within routine clinical QA. Currently, EPID performance is
monitored using image quality indices obtained from commercial QA phantoms. Such methods may be qualitative or
sensitive to imaging conditions, and hence not appropriate for rigorous QA. In this work, a simple and quick method for
EPID QA is presented based on a line-pair bar-pattern for fast MTF measurement, coupled with standard NPS and DQE
measurements, all of which can be carried out within two minutes. The method to determine MTF from line-pair
modulations was developed based on theoretical analyses and Monte Carlo simulations to identify accurate and reliable
measurement conditions. The bar-pattern based QA method was tested with two clinical EPIDs and found to be in
excellent agreement with slit/edge measurements. It was also implemented for weekly QA checks, and compared with
established commercial QA methods. This bar-pattern based QA was more sensitive to potential decrease of EPID image quality.
Characterization of a CMOS detector for limited-view mammography
Show abstract
Sensors based on complementary metal oxide semiconductors (CMOS) technology have recently been
considered for mammography applications. CMOS offers the advantages of lower cost and relative ease
of fabrications. We report on the evaluation of a CMOS imager (C9730DK, Hamamatsu Corporation)
with 14-bit digitization and 50-micron detector element (del) resolution. The imager has an active area of
5 x 5 cm and uses 160-micron layer of needle-crystal CsI (55 mg/cc) to convert x-rays to light. The
detector is suitable for spot and specimen imaging and image-guided biopsy. To evaluate resolution
performance, we measured the modulation transfer function (MTF) using the slanted edge method. We
also measured the normalized noise power spectrum (NNPS) using Fourier analysis of uniform images.
The MTF and NNPS were used to determine the detective quantum efficiency (DQE) of the detector. The
detector was characterized using a molybdenum target/molybdenum filter mammography x-ray source
operated at 28 kVp with 44mm of PMMA added to mimic clinical beam quality (HVL = 0.62 mm Al).
Our analysis showed that the imager had a linear response. The MTF was 28% at 5 lp/mm and 8% at 10
lp/mm. The product of the NNPS and exposure showed that the detector was quantum limited. The DQE
near 0 lp/mm was in the 55-60% range. The DQE and MTF performance of the CMOS detector are
comparable to published values for other digital mammography detectors.
Investigation of the Z-axis resolution of breast tomosynthesis mammography systems
Show abstract
Digital Tomosynthesis Mammography (DTM) is a promising modality that can improve breast
cancer detection. DTM acquires low-dose mammograms at a number of projection angles over a limited
angular range and reconstructs the 3D breast volume. DTM can provide depth information to separate
overlapping breast tissues occurred in conventional mammograms, thereby facilitating detection of subtle
lesions. In this work, we investigated the impact of the imaging parameters and reconstruction methods on
the Z-axis resolution in DTM systems. The Z-axis resolution represents the ability of the DTM system to
distinguish adjacent objects along the depth direction. A DTM system with variable image acquisition
parameters was modeled. In this preliminary study, a computer phantom containing a high-density point
object embedded in an air volume was used. We simulated a range of DTM conditions by generating an
appropriate number of PV images in 3° increments covering a total tomosynthesis angle from ±15° to ±30°.
The Simultaneous Algebraic Reconstruction Technique (SART) was used for reconstruction of the imaged
volume from the noise-free projection data and the results were compared to those of back-projection
method. Vertical line profiles along the Z-axis and through the object center were extracted from the
reconstructed volume and the full-width-at-half-maximum (FWHM) of the normalized intensity profile was
used to evaluate the Z-axis resolution. Preliminary results demonstrated that while the Z-axis resolution
remains almost constant as a function of depth within a 5-cm-thick volume, it is strongly affected by the PV
angular range such that the depth resolution improves with increasing total tomosynthesis angle. The depth
resolution also depends on the reconstruction algorithm employed; the SART method is superior to the
simple back-projection method in terms of depth resolution.
Performance analysis of a CsI-based flat panel detector in a cone beam variable resolution x-ray system
Show abstract
A new Cone-Beam CT (CBCT) system is introduced that uses the concept of Variable Resolution X-ray (VRX)
detection, which has previously been demonstrated to significantly increase spatial resolution for small objects.
An amorphous silicon Flat Panel Detector (FPD) with a CsI scintillator (PaxScan 2020, Varian, Salt Lake City,
UT) is coupled with a micro-focus x-ray tube (35 - 80 kVp, 10 - 250 μA) to form a CBCT. The FPD is installed
on a rotating arm that can be adjusted to any angle θ, called the VRX angle, between 90° and 0° with respect to
the x-ray direction. A VRX angle of 90° for the detector corresponds to a conventional CBCT whereas a VRX
angle of 30° means that the detector is tilted 90° - 30° = 60° from its perpendicular position. Tilting the FPD in
this manner reduces both the line-spread function width and the sampling distance by a factor of sin(&thgr;), thereby
increasing detector spatial resolution proportionately. An in-house phantom is used to measure the MTF of the
reconstructed CT images using different VRX angles. An increase by a factor of 1.67 ± 0.007 is observed in the
MTF cutoff frequency at 30° compared to 90° in images acquired at 75 kVp. Expected theoretical value for this
case is 2.0. The new Cone-Beam Variable Resolution X-ray (CB-VRX) CT system is expected to significantly
improve the images acquired from small objects - such as small animals - while exploiting the opportunities
offered by a conventional CBCT.
The CT image standardization based on the verified PSF
Show abstract
This study discusses a method of CT image quality standardization that uses a point-spread function (PSF) in MDCT.
CT image I(x,y,z) is represented by the following formula: I(x,y,z) = O(x,y,z)***PSF(x,y,z). Standardization was
performed by measuring the three-dimensional (3-D) PSFs of two CT images with different image qualities. The image
conversion method was constructed and tested using the 3-D PSFs and CT images of the CT scanners of three different
manufacturers. The CT scanners used were Lightspeed QX/i, Somatom Volume Zoom, and Brilliance-40. To obtain the
PSF(x,y) of these CT scanners, the line spread functions of the respective reconstruction kernels were measured using a
phantom described by J.M. Boone. The kernels for each scanner were: soft, standard, lung, bone, and bone plus (GE);
B20f, B40f, B41f, B50f, and B60f (Siemens); and B, C, D, E, and L (Philips). Slice sensitivity profile (SSP) were
measured using a micro-disk phantom (50 μm* φ1 mm) with 5 mm slice thickness and beam pitch of 1.5 (GE, Siemens)
and 0.626 (Philips). 3-D PSF was verified using an MDCT QA phantom. Real chest CT images were converted to
images with contrasting standard image quality. Comparison between the converted CT image and the original standard
image showed good agreement. The usefulness of the image conversion method is discussed using clinical CT images
acquired by CT scanners produced by different manufacturers.
Minimum dose calculation for different imaging tasks in digital projection radiography
Show abstract
In diagnostic radiology it is a common interest of patient and medical staff to keep the exposure as low as reasonably
achievable (ALARA). In spite of this task being well known there is no consensus about how low the exposure for a
specific diagnostic examination can be. The methods presented in this paper allow for the mathematical determination of
the lowest exposure necessary to perform basic, well defined imaging tasks in digital radiography. The model quantifies
how different demands on the result influence the necessary exposure. Dependences on parameters describing the beam
quality used and the detector are implemented into the model. A strong rise of the necessary exposure to detect a certain
contrast of lower amplitude or with higher certainty was determined. The effects of a change of the energy of the
irradiation are as a first step investigated via the connected change of the transmission of a main absorber. By
specification of the specimen to be observed the result is connected to beam energies which can be correlated with the
energy dependent response of a realistic detector system. The calculations give basic information about the best exposure
in a simplified view of patient dimensions and diagnostic needs. Especially in pediatric radiology optimized adaptations
to the patient and the clinical question are expected to take great effects due to the great variations of patient sizes.
Validation of software for QC assessment of MTF and NPS
Show abstract
Modulation transfer function (MTF), noise power spectrum (NPS), and detective quantum efficiency (DQE) are widely accepted measures of digital radiographic system performance. However the implementation of these measurement methodologies has been limited to a handful of researchers using an assortment of techniques. A prototype edge tool and easy-to-use software program, which can generate MTF, NPS, and DQE results quickly and easily in the field, have been developed. The edge tool consists of 1mm or 250 μ thick tungsten with two polished edges. Edge and NPS data were obtained and analyzed by 3 investigators using three analysis methods: Method A, the software under development for this report; Method B, code available on the web site of one of the investigators [Saunders and Samei, Med. Phys. 33, 308-319 (2006)]; and Method C, code developed by two other of the investigators [Samei and Flynn, Med Phy. 30, 608-622, (2003)]. In all cases the differences between the results using Method B and Method A were less than 1%. The differences between Method A and Method C were larger, up to 5.26%. NPS were calculated using Method A and B. The results were very close, with average errors less than 2.5% for exposures of 27.3, 9.3, and 2.7 μGy. Analysis of data for a 10 cm misalignment shows no significant error for either the 250 μ or 1mm edge. The method developed gives results that correlate closely with results obtained from established methods. The software is easy-to-use and flexible in its application. The Edge Tool developed has the necessary precision to accurately determine the MTF values of the system. Further validation of NPS and DQE is ongoing.
Software tools dedicated for an automatic analysis of the CT scanner quality control images
Show abstract
This paper deals with the CT scanner images quality control, which is an important part of the quality control process of the CT scanner, which consists of making measurement in images of dedicated phantoms.
Standard methods consist of scan explorations of phantoms that contain different specific patterns1, 2. These methods rely on manual measurements with graphics tools in corresponding images (density, position, length...) or automatic measurements developed in softwares3, 4 that use some masks to determine the region of interest (ROI). The problem of these masks is that they may produce wrong results in case of misalignment of the phantom.
We propose a new method that consists, firstly of developing software tools that are capable of performing an automated analysis of CT images of three standard phantoms, LAP5 , GEMS6 and CATPHAN6007, in terms of slice thickness, spatial resolution, low and high level contrast, noise and uniformity. The method we have developed is completely automatic because it uses some protocols and special treatments in the images to automatically detect the position and the size of the ROI. Secondly, to test the performances of our software tools, we develop two digital phantoms which reproduce the exact geometry and composition of the physical phantoms, i.e. some perfect CT images of the real phantoms, and a complete set of distorted digital phantoms which represent the "perfect" phantom distorted by noise and blur calibrated functions to test the performances of our automated analysis software.
Optimization of image quality and average glandular dose in CR mammography
Show abstract
The concept proposed by EUREF which determines the AGD between upper limit corresponding to the acceptable level
of AGD and lower limit corresponding to lower limit of image quality was applied to CR Mammography, and the
resulting object thickness tracking and tube voltage tracking were determined.
EUREF specifies threshold contrast visibility for a 5 cm of PMMA. In accordance with this definition, the lower limit of
CNR for a 5 cm of PMMA was determined by measuring the CNR at the lower limit of AGD where threshold contrast
visibility was just acceptable. The obtained lower limit of CNR was then multiplied by the object thickness
compensation factor to estimate the lower limits of CNR for all object thicknesses. AGDs are now determined for each
lower limit of CNR to obtain thickness tracking and tube voltage tracking characteristics. Although the limited range of
examined target/filter combinations and tube voltages should be taken into consideration, these tracking characteristics
for constant CNR differ in their profiles from those of a screen-film system with AEC for providing constant optical
density. Among our findings, we found that a lower AGD is achieved while maintaining CNR for a thick object when
using a combination of target/filter and tube voltage that generates higher X-ray energy compared to the combination
given in the EUREF's typical spectra per PMMA thickness (Mo/Rh 32 kV at PMMA 4cm and Rh/Rh 28 kV at PMMA
6 cm, for example).
We also found that the thickness tracking characteristics for constant S value behaves similarly to constant CNR under
certain conditions of the target/filter combination and tube voltage.
Complete MTF evaluation of two cone beam CT systems
Show abstract
The Modulation Transfer Function (MTF) of any system is the frequency response to a delta signal. This response is degraded by several factors such as the inherent veiling glare of the detector and
the focal spot size among others. Consequently, the MTF has been one of the physical characteristics that
are commonly used to quantitatively measure the physical performance of a system. In this article, the
MTF of two flat panel detectors (FPD) and of two Cone Beam CT systems is evaluated. First, the MTF of
PaxScan 2520 and of PaxScan 4030CB is evaluated. One of the standard techniques to evaluate the MTF
of a FPD is by using an edge of a metal with high atomic number. For instance, it has been suggested by
IEC 62220-1 to use an opaque edge to evaluate the MTF of a FPD. Yet, it was found that different metals
yield slightly different MTF. In this study, the effects on the MTF evaluation of different metals and exposure parameters was studied and analyzed. First, the MTF was evaluated using different kVps and exposures levels. Second, the MTF was evaluated using aluminum edges of different thickness. Third, the MTF was evaluated using the following four different metals: Aluminum, Copper, Steel and Lead. Finally, the MTF obtained previously were compared to the MTF obtained by using a pinhole. In the second part of this study, the MTF of two systems using these two FPDs were also evaluated using different wires, filters and acquisition modes. The preliminary results demonstrate that the MTF is independent of kVp and exposure level. Yet, it is dependent on the material used to evaluate it.
Automatic quality control of digitally reconstructed radiograph computation and comparison with standard methods
Show abstract
Conformal radiotherapy helps to deliver an accurate and effective cancer treatment by exactly targeting the
tumor. In this purpose, softwares of the treatment planning system (TPS) compute every geometric parameters
of the treatment. It is essential to control the quality of them because the TPS performances are directly
connected with the precision on the treated region. The standard method to control them is to use physical
test objects (PTOs).1, 2 The use of PTOs introduces uncertainties in the quality assessment because of the CT
scan. Another method to assess the quality of these softwares is to use digital test objects (DTOs).3-5 DTOs
are exactly known in a continuous and a discrete way. Thus the assessment of the TPS quality can be more
accurate and faster. The fact that the DTO characteristics are well known allows to calculate a theoretical result.
The comparison of the TPS and this theoretical results leads to a quantitative assessment of the TPS softwares
quality. This work presents the control of major quality criteria of digitally reconstructed radiograph (DRR)
computation: ray divergence, ray incidence and spatial resolution. Fully automated methods to control these
points have been developed. The same criteria have been tested with PTO and the quality assessments by the
two methods have been compared. The DTO methods appeared to be much more accurate because computable.
Physical and psychophysical characterization of a GE senographe DS clinical system
Show abstract
Indirect-conversion FFDM systems usually present a lower spatial resolution, with respect to the direct-conversion one. This can put serious issues in mammography, since high resolution is required. Digital software has been developed for restoring the losses in spatial resolution caused by blurring in the scintillation phosphor. GE Senographe DS system gives users the possibility of using such restoration. Basically, a filtering can be performed on the acquired images, by activating the FineView software option. In this work we present a complete characterization of a clinical system, in terms of MTF, NPS, DQE, and contrast-detail analysis. Figures of merit have been calculated on images acquired with and without the FineView software. The effects of the restoration software are investigated, both on image quality parameters, and on contrast-detail visibility. The MTF of the FFDM system is improved when FineView is activated. On the other hand, NPS presents noticeably changes, especially at high frequencies. DQE is fairly independent from the exposure, when FineView filter is not activated, whereas it presents a clear spread over the exposures, when FineView is activated. CDMAM analysis does not show significant differences between images with or without the restoration filter. Besides, the Mo/Mo beam seems to provide slightly better results than the Rh/Rh one.
Virtual adaptation of physical phantoms to datasets derived from clinical tomographic examinations
Show abstract
Thoracic radiography was simulated making use of a virtual pediatric model created from tomographic data
(voxelphantom) of a child eight weeks old. The dataset was scaled down to fit the dimensions of a premature newborn.
The simulation allows a quantitative and spatially resolved analysis of the x-ray image generation. The transmission
behavior of different anatomical regions present in the voxelphantom was compared to the output of the simulation of a
step-like phantom made from aluminum and PMMA. The step like structure of the simulated model can be easily built
and statements about its x-ray related behavior can be directly validated by means of experiments. A thin contrast plate
was placed on each step to make the determination of dependences e.g. between the applied radiation energy and the
contrast to noise ratio possible.
Performance evaluation of a direct computed radiography system by means of physical characterization and contrast detail analysis
Show abstract
The aim of this study is to determine the performance of a direct CR reader, named "FCR Velocity U Focused Phosphor (FP)". The system is based on a CsBr columnar structured crystal, and the system's read out is based on the "linescan technology" that employs a wide-view CCD. The system's physical performance was tested by means of a quantitative analysis, with calculation of the modulation transfer function (MTF), noise power spectrum (NPS) and detective quantum efficiency (DQE). Image quality was assessed by performing a contrast-detail analysis. The results are compared with those obtained with the well known CR system Fuji FCR XG5000, and the new one Kodak DirectView CR 975. For all the measurements the standard radiation quality RQA-5 was used. The relationship between signal amplitude and entrance air kerma is logarithmic for all the systems and the response functions were used to linearize the images before the MTF (edge method) and NPS calculations. The contrast detail analysis has been achieved by using the well known CDRAD phantom and a customized software designed for automatic computation of the contrast-detail curves. The three systems present similar MTFs, whereas the Fuji Velocity U FP system, thanks to its greater efficiency, has a better behavior in terms of NNPS, especially at low frequencies. That allows the system based on columnar phosphor to provide a better DQE. CDRAD analysis basically confirms that the structured phosphor used in the Velocity system improves the visibility of some details. This is especially true for medium and large details.
A new method for evaluation of slice sensitivity profiles (SSPz) for spatial variation in 64-channel MSCT
Show abstract
The purpose of this study is to evaluate Slice- sensitive profile on Z-axis (SSPz) at any point in XY plane of multi-slice
helical CT (MSCT) by new method that we are proposing. It is very important to understand MSCT physical property to
provide high-integrity information and get the best possible images. Quality evaluation around the images has been
gaining more importance by needs of proof diversity on evaluation of CT physical property. However, in conventional
method of image quality edge of the images, we had some unignorable data variation in every measurement. Considering
the development, we assumed that in acquiring helical, the relative position of tube trajectory for the measuring object
becomes the problem. Setting the proper scan interval and acquiring the data continuously, we devised the method for
property evaluation with controlling tube trajectory. We obtained SSPZ in multiple positions and measured full width at
half maximum (FWHM) distribution in axial plane. As a result, we found periodic variation in FWHM especially around
the images. The degree of the variation changes depends on helical pitch (HP).If HP is small, the variation is also small,
and if HP become larger, the variation also become larger. We figured out the variation in SSPz and FWHM planes
through the method that we proposed.
How do kV and mAs affect CT lesion detection performance?
Show abstract
The purpose of this study was to investigate how output (mAs) and x-ray tube voltage (kV) affect lesion detection in CT imaging. An adult Rando phantom was scanned on a GE LightSpeed CT scanner at x-ray tube voltages from 80 to 140 kV, and outputs from 90 to 360 mAs. Axial images of the abdomen were reconstructed and viewed on a high quality monitor at a soft tissue display setting. We measured detection of 2.5 to 12.5 mm sized lesions using a 2 Alternate Forced Choice (2-AFC) experimental paradigm that determined lesion contrast (I) corresponding to a 92% accuracy (I92%) of lesion detection. Plots of log(I92%) versus log(lesion size) were all approximately linear. The slope of the contrast detail curve was ~ -1.0 at 90 mAs, close to the value predicted by the Rose model, but monotonically decreased with increasing mAs to a value of ~ -0.7 at 360 mAs. Increasing the x-ray tube output by a factor of four improved lesion detection by a factor of 1.9 for the smallest lesion (2.5 mm), close to the value predicted by the Rose model, but only by a factor of 1.2 for largest lesion (12.5 mm). Increasing the kV monotonically decreased the contrast detail slopes from -1.02 at 80 kV to -0.71 at 140 kV. Increasing the x-ray tube voltage from 80 to 140 kV improved lesion detection by a factor of 2.8 for the smallest lesion (2.5 mm), but only by a factor of 1.7 for largest lesion (12.5 mm). We conclude that: (i) quantum mottle is an important factor for low contrast lesion detection in images of anthropomorphic phantoms; (ii) x-ray tube voltage has a much greater influence on lesion detection performance than x-ray tube output; (iii) the Rose model only predicts CT lesion detection performance at low x-ray tube outputs (90 mAs) and for small lesions (2.5 mm).
Method for the determination of the modulation transfer function (MTF) in 3D x-ray imaging systems with focus on correction for finite extent of test objects
Show abstract
It is well known that rotational C-arm systems are capable of providing 3D tomographic X-ray images with much higher spatial resolution than conventional CT systems. Using flat X-ray detectors, the pixel size of the detector typically is in the range of the size of the test objects. Therefore, the finite extent of the "point" source cannot be neglected for the determination of the MTF. A practical algorithm has been developed that includes bias estimation and subtraction, averaging in the spatial domain, and correction for the frequency content of the imaged bead or wire. Using this algorithm, the wire and the bead method are analyzed for flat detector based 3D X-ray systems with the use of standard CT performance phantoms. Results on both experimental and simulated data are presented. It is found that the approximation of applying the analysis of the wire method to a bead measurement is justified within 3% accuracy up to the first zero of the MTF.
In-plane artifacts in breast tomosynthesis quantified with a novel contrast-detail phantom
Show abstract
The purpose of this work was to develop a contrast-detail phantom that can be used to evaluate image quality in breast
tomosynthesis (BT) and as a first step use it to evaluate in-plane artifacts with respect to object size and contrast. The
phantom was constructed using a Polylite® resin as bulk material, as it has x-ray mass attenuation properties similar to
polymethyl methacrylate (PMMA), a common phantom material in mammography. Six different materials -
polyoxymethylene (POM), bakelite®, nylon, polycarbonate (PC), acrylonitrilebutadienestyrene (ABS) and polyethene
(PE) - were selected to form the phantom details. For each of the six materials, five spherical objects were manufactured
(diameters of 4, 8, 12, 16, and 20 mm) resulting in 30 objects that were embedded with their centres approximately
aligned at the central plane of a 26 mm thick Polylite® block (210 mm x 300 mm). A 20 mm thick PMMA block was
added to yield a phantom with attenuation properties similar to 45 mm PMMA that could simulate a so-called standard
breast (50 mm thick, 50% glandular tissue). Images of the phantom were acquired using a BT prototype system that
employs filtered backprojection for image reconstruction. The magnitude of the in-plane artifacts was evaluated and was
found to increase linearly with increasing contrast (signal) level and size of the embedded objects. The contrast-detail
phantom was found to be a useful tool for evaluating BT in-plane artifacts and might also be used to study out-of-plane
artifacts and the effect of different acquisition and reconstruction parameters on image quality in BT.
Poster Session: Signal Analysis
Sound-speed and attenuation imaging of breast tissue using waveform tomography of transmission ultrasound data
Show abstract
Waveform tomography results are presented from 800 kHz ultrasound transmission scans of a breast phantom, and from an in vivo ultrasound breast scan: significant improvements are demonstrated in resolution over time-of-flight reconstructions. Quantitative reconstructions of both sound-speed and inelastic attenuation are recovered. The data were acquired in the Computed Ultrasound Risk Evaluation (CURE) system, comprising a 20 cm diameter solid-state ultrasound ring array with 256 active, non-beamforming transducers.
Waveform tomography is capable of resolving variations in acoustic properties at sub-wavelength scales. This was verified through comparison of the breast phantom reconstructions with x-ray CT results: the final images resolve variations in sound speed with a spatial resolution close to 2 mm.
Waveform tomography overcomes the resolution limit of time-of-flight methods caused by finite frequency (diffraction) effects. The method is a combination of time-of-flight tomography, and 2-D acoustic waveform inversion of the transmission arrivals in ultrasonic data. For selected frequency components of the waveforms, a finite-difference simulation of the visco-acoustic wave equation is used to compute synthetic data in the current model, and the data residuals are formed by subtraction. The residuals are used in an iterative, gradient-based scheme to update the sound-speed and attenuation model to produce a reduced misfit to the data. Computational efficiency is achieved through the use of time-reversal of the data residuals to construct the model updates. Lower frequencies are used first, to establish the long wavelength components of the image, and higher frequencies are introduced later to provide increased resolution.
Implementation of a fully 3D system model for brain SPECT with fan-beam-collimator OSEM reconstruction with 3D total variation regularization
Show abstract
In order to improve tomographically reconstructed image quality, we have implemented a fully 3D reconstruction, using
an ordered subsets expectation maximization (OSEM) algorithm for fan-beam collimator (FBC) SPECT, along with a
volumetric system model-fan-volume system model (FVSM), a modified attenuation compensation, a 3D depth- and
angle-dependent resolution and sensitivity correction, and a 3D total variation (TV) regularization. SPECT data were
acquired in a 128x64 matrix, in 120 views with a circular orbit. The numerical Zubal brain phantom was used to
simulate a FBC HMPAO Tc-99m brain SPECT scan, and a low noise and scatter-free projection dataset was obtained
using the SimSET Monte Carlo package. A SPECT scan for a mini-Defrise phantom and brain HMPAO SPECT scans
for five patients were acquired with a triple-head gamma camera (Triad 88) equipped with a low-energy high-resolution
(LEHR) FBC. The reconstructed images, obtained using clinical filtered back projection (FBP), OSEM with a line-length
system model (LLSM) and 3D TV regularization, and OSEM with FVSM and 3D TV regularization were
quantitatively studied. Overall improvement in the image quality has been observed, including better axial and transaxial
resolution, better integral uniformity, higher contrast-to-noise ration between the gray matter and the white matter, and
better accuracy and lower bias in OSEM-FVSM, compared with OSEM-LLSM and clinical FBP.
Hybrid geodesic region-based curve evolutions for image segmentation
Show abstract
In this paper we present a gradient descent flow based on a novel energy functional that is capable of producing
robust and accurate segmentations of medical images. This flow is a hybridization of local geodesic active
contours and more global region-based active contours. The combination of these two methods allows curves
deforming under this energy to find only significant local minima and delineate object borders despite noise,
poor edge information, and heterogeneous intensity profiles. To accomplish this, we construct a cost function
that is evaluated along the evolving curve. In this cost, the value at each point on the curve is based on the
analysis of interior and exterior means in a local neighborhood around that point. We also demonstrate a novel
mathematical derivation used to implement this and other similar flows. Results for this algorithm are compared
to standard techniques using medical and synthetic images to demonstrate the proposed method's robustness
and accuracy as compared to both edge-based and region-based alone.
Quantitative analysis of 3D stent reconstruction from a limited number of views in cardiac rotational angiography
Show abstract
Percutaneous coronary angioplasty consists in conducting a guidewire carrying a balloon and a stent through
the lesion and deploying the stent by balloon inflation. A stent is a small 3D complex mesh hardly visible in
X-ray images : the control of stent deployment is difficult although it is important to avoid post intervention
complications. In a previous work, we proposed a method to reconstruct 3D stent images from a set of 2D
cone-beam projections acquired in rotational acquisition mode. The process involves a motion compensation
procedure based on the position of two markers located on the guidewire in the 2D radiographic sequence.
Under the hypothesis that the stent and markers motions are identical, the method was shown to generate a
negligible error. If this hypothesis is not fulfilled, a solution could be to use only the images where motion is
weakest, at the detriment of having a limiter number of views. In this paper, we propose a simulation based
study of the impact of a limited number of views in our context. The chain image involved in the acquisition of
X-ray sequences is first modeled to simulate realistic noisy projections of stent animated by a motion close to
cardiac motion. Then, the 3D stent images are reconstructed using the proposed motion compensation method
from gated projections. Two gating strategies are examined to select projection in the sequences. A quantitative
analysis is carried out to assess reconstruction quality as a function of noise and acquisition strategy.
A wavelet based method for SPECT reconstruction with non-uniform attenuation
Show abstract
SPECT (single photon emission computed tomography) is a tomography technique that can greatly show information about the metabolic activity in body and improve clinical diagnosis. In SPECT, because of photoelectric absorption and Compton scattering, the emitted gamma photons are attenuated inside the body before arriving at the detector. The goal of quantitative SPECT reconstruction is to obtain an accurate reconstructed image of the radioactivity distribution in the interested area of a human body, so the compensation for non-uniform attenuation is necessary in the quantitative SPECT reconstruction. In this paper, based on the explicit inversion formula for the attenuated Radon transform discovered by R. Novikov, we present a wavelet based SPECT reconstruction algorithm with non-uniform attenuation. We know that the wavelet transform has characteristics of multi-resolution analysis and localized analysis, and these characteristics can be applied to de-noising and localized reconstruction. Simulation results show that our wavelet based SPECT reconstruction algorithm is accurate.
3D view weighted cone-beam backprojection reconstruction for digital tomosynthesis
Show abstract
Cone-beam filtered backprojection (CB-FBP) is one of the major reconstruction algorithms for digital tomosynthesis. In conventional FBP, the photon fluxes in projections are evenly distributed along the X-ray beam. Due to the limited view angles and finite detector dimensions, this uniform weighting causes non-uniformity in the recon images and leads to cone-beam artifact. In this paper, we propose a 3-D view weighting technique in combination with FBP to combat this artifact. An anthropomorphic chest phantom was placed at supine position to enable the imaging of chest PA view. During a linear sweep of X-ray source, 41 X-ray images at different projection angles were acquired with the following protocol: 120kVp, 160mA, and 0.64mAs/exposure. To create the worst scenario for testing, we chose 60 degrees as the sweep angle in this exam. The data set was reconstructed with conventional CB-FBP and proposed algorithm under the same parameters: FOV = 40x40 cm^2, and slice thickness = 4mm. 3 recon slices were randomly selected for review with slice height = 10.5/14.5/17.5cm. Results were assessed qualitatively by human observers and quantitatively through ROI measurement. In each slice, three pre-defined ROIs (50x50 pixels)--ROI A and B are in artifact more pronounced area, and ROI C is in relatively artifact-free area--are extracted and measured. The non-uniformity error was defined as the ratio of MEAN(AVG(C-A), AVG(C-B)) / AVG(C). The average non-uniformity error over the three test images was 0.428 for without view weighting and only 0.041 for with view weighting.
Automatic generation of 3D coronary artery centerlines
Show abstract
A fully automated 3D centerline modeling algorithm for coronary arteries is presented. It utilizes a subset of standard rotational X-Ray angiography projections that correspond to a single cardiac phase. The projection selection is based on a simultaneously recorded electrocardiogram (ECG). The algorithm utilizes a region growing approach, which selects voxels in 3D space that most probably belong to the vascular structure. The local growing speed is controlled by a 3D response computation algorithm. This algorithm calculates a measure for the probability of a point in 3D to belong to a vessel or not.
Centerlines of all detected vessels are extracted from the 3D representation built during the region growing and linked in a hierarchical manner. The centerlines representing the most significant vessels are selected by a geometry-based weighting criterion.
The theoretically achievable accuracy of the algorithm is evaluated on simulated projections of a virtual heart phantom. It is capable of extracting coronary centerlines with an accuracy that is mainly limited by projection and volume quantization (0.25 mm). The algorithm needs at least 3 projections for modeling, while in the phantom study, 5 projections are sufficient to achieve the best possible accuracy. It is shown that the algorithm is reasonably insensitive to residual motion, which means that it is able to cope with inconsistencies within the projection data set caused by finite gating accuracy, respiration or irregular heart beats. Its practical feasibility is demonstrated on clinical cases showing automatically generated models of left and right coronary arteries (LCA/RCA).
Cone beam CT image quality measurements: PSF de-convolution
Show abstract
The Modulation Transfer Function (MTF) of any system is the frequency response to a delta
signal. Ideally, this response should be a step function for a flat panel detector (FPD); it should be one for
frequencies less than the Nyquist frequency and zero for frequencies above. Yet, this response is degraded
by several factors such as the veiling glare of the detector and the focal spot size. Consequently, the MTF
has been one of the physical characteristics that are commonly used to quantitatively measure the physical
performance of a system. One of the standard techniques to evaluate the MTF is by using an edge of a
metal with a high atomic number. For instance, it has been suggested by IEC 62220-1 to use an opaque
edge to evaluate the MTF of FPDs. In a previous study, it was found that different metals yield slightly
different evaluation of the MTF. The effects of these slightly different MTFs on image quality were
investigated. The evaluation of the MTFs of a PaxScan 4030CB and PaxScan 2520 from a previous study
were used in this study. A ball, a cylindrical water phantom, a breast phantom, a living mouse and three
breasts of patients from a pilot study were analyzed for improvements in image quality after PSF deconvolution
post-processing. The results of this study suggest that the detector's MTF de-convolution
post-processing achieved a CNR's increase while it also enhanced the edges and uniformity.
Point spread function based classification of regions for linear digital tomosynthesis
Show abstract
In digital tomosynthesis, one of the limitations is the presence of out-of-plane blur due to the limited angle acquisition. The point spread function (PSF) characterizes blur in the imaging volume, and is shift-variant in tomosynthesis. The purpose of this research is to classify the tomosynthesis imaging volume into four different categories based on PSF-driven focus criteria. We considered linear tomosynthesis geometry and simple back projection algorithm for reconstruction. The three-dimensional PSF at every pixel in the imaging volume was determined. Intensity profiles were computed for every pixel by integrating the PSF-weighted intensities contained within the line segment defined by the PSF, at each slice. Classification rules based on these intensity profiles were used to categorize image regions. At background and low-frequency pixels, the derived intensity profiles were flat curves with relatively low and high maximum intensities respectively. At in-focus pixels, the maximum intensity of the profiles coincided with the PSF-weighted intensity of the pixel. At out-of-focus pixels, the PSF-weighted intensity of the pixel was always less than the maximum intensity of the profile. We validated our method using human observer classified regions as gold standard. Based on the computed and manual classifications, the mean sensitivity and specificity of the algorithm were 77+/-8.44% and 91+/-4.13% respectively (t=-0.64, p=0.56, DF=4). Such a classification algorithm may assist in mitigating out-of-focus blur from tomosynthesis image slices.
A practical correction of scatter-related artifacts in SPECT reconstruction
Show abstract
We have observed that an expectation maximization (EM) algorithm applied to SPECT reconstruction may produce hotspot
artifacts of varying intensity. Our hypothesis was that scatter caused these artifacts. To test this assumption, we
studied the performance of forward- and back-projection procedures in the EM algorithm for simulated and experimental
SPECT data. First, synthetic scatter-free projections and projections with only one scattered photon in each view were
created for a simulated simple object, and reconstructed with a fully 3D ordered-subsets EM (OSEM) algorithm. Then,
Monte Carlo simulated brain SPECT (with no scatter and with scatter present), a mini-Defrise phantom, and patient
SPECT were reconstructed. We confirmed our hypothesis: hot-spot artifacts appeared only in the reconstruction from
noisy projections but not in the reconstruction from scatter-free projections. We investigated a practical and simple
method, critical path-length control (CPLC), for suppression of the hot-spot artifacts. To this end we performed
reconstructions with or without CPLC and quantitatively evaluated the results including estimation of accuracy, bias,
contrast-to-noise ratio, and uniformity. We found that the OSEM-with-CPLC method significantly reduced hot-spot
artifacts, and yielded a similar or improved image quality. We conclude that the CPLC method provides a useful yet
simple tool to reduce scatter-related hot-spot artifacts.
Statistics of MR signals: revisited
Show abstract
A common view in MRI research is that the object variability of MR signals
is negligible. With this recognition, the signal components of MR signals
are treated as constant and the noise components are studied as random
processes. Because signal components of MR signals represent a collective
behavior of a huge mount of spins, a statistical investigation may provide
a better understanding of MR signals. The work reported in this paper first
investigates statistics of the thermal equilibrium bulk magnetization (TEMM)
which is the quantity to be imaged - one that can be measured and actually
observed in MRI. Then it investigates statistics of the transverse precessing
bulk magnetization (TPMM) which introduce an electromagnetic force in the
receiver coil of MRI. Finally this study investigates statistics of signal
components of three MR signals at the different stages of MR signal detection
module: Free Induction Decay (FID), Phase Sensitive Detection (PSD), and
Analog-to-Digital Conversion (ADC), sequentially. k-space sample is a
reformatted ADC signal. The study derives and proves stochastic models for
TEMM, TPMM, FID, PSD, and ADC signals, also proposes and justifies stochastic
models for homogeneous and inhomogeneous samples. The study shows that under
the normal conditions and the ordinary settings, magnetizations can be
characterized as spatially deterministic processes with Probability one,
and MR signals - signal component plus noise component - can be characterized
as temporal Gaussian random processes with the means of signal components
and the variances of noise components. These means are expressed in closed
forms in terms of parameters of MR imaging system and the samples. The derived
statistical properties of MR signals will serve as the basis for evaluating
performances of imaging system and studying statistics of the MR image.
A posteriori respiratory motion compensation for PET imaging
Show abstract
This paper deals with enhancing the formation of PET images. Physiological motion, such as breathing, may cause significant alteration of image quality. Correction methods include gated acquisitions that significantly increase the acquisition time. In this paper we propose an original method for reducing respiratory motion artefacts in PET images. It is based on synchronous acquisition of PET and CT data with a spirometer. CT images are acquired at each step of a subdivided respiratory cycle, and registered to estimate the body transformations. Then PET data is indirectly registered and corrected for attenuation before reconstructing a PET image with enhanced quality. This method has been validated using a specific phantom experimentation. Results show that the method brings improved accuracy in tumour volume representation. In addition, the PET imaging clinical protocol is unchanged: our method does not increase the acquisition time nor constrain the patient breathing.
Image reconstruction from sparse data samples along spiral trajectories in MRI
Show abstract
We present a method for obtaining accurate image reconstruction from sparsely sampled magnetic resonance
imaging (MRI) data obtained along spiral trajectories in Fourier space. This method minimizes the total
variation (TV) of the estimated image, subject to the constraint that the Fourier transform of the image
matches the known samples in Fourier space. Using this method, we demonstrate accurate image reconstruction
from sparse Fourier samples. We also show that the algorithm is reasonably robust to the effects
of measurement noise. Reconstruction from such sparse sampling should reduce scan times, improving scan
quality through reduction of motion-related artifacts and allowing more rapid evaluation of time-critical
conditions such as stroke. Although our results are discussed in the context of two-dimensional MRI, they
are directly applicable to higher dimensional imaging and to other sampling patterns in Fourier space.
Poster Session: Reconstruction
Extended volume coverage in helical cone-beam CT by using PI-line-based BPF algorithm
Show abstract
We compared data requirements of filtered-backprojection (FBP) and backprojection-filtration (BPF) algorithms
based on PI-lines in helical cone-beam CT. Since the filtration process in FBP algorithm needs all the projection data
of PI-lines for each view, the required detector size should be bigger than the size that can cover Tam-Danielsson
(T-D) window to avoid data truncation. BPF algorithm, however, requires the projection data only within the T-D
window, which means smaller detector size can be used to reconstruct the same image than that in FBP. In other
words, a longer helical pitch can be obtained by using BPF algorithm without any truncation artifacts when a fixed
detector size is given. The purpose of the work is to demonstrate numerically that extended volume coverage in
helical cone-beam CT by using PI-line-based BPF algorithm can be achieved.
Implementation of a cone-beam backprojection algorithm on the cell broadband engine processor
Show abstract
Tomographic image reconstruction is computationally very demanding. In all cases the backprojection represents the performance bottleneck due to the high operational count and due to the high demand put on the memory subsystem. In the past, solving this problem has lead to the implementation of specific architectures, connecting Application Specific Integrated Circuits (ASICs) or Field Programmable Gate Arrays (FPGAs) to memory through dedicated high speed busses. More recently, there have also been attempt to use Graphic Processing Units (GPUs) to perform the backprojection step.
Originally aimed at the gaming market, IBM, Toshiba and Sony have introduced the Cell Broadband Engine (CBE) processor, often considered as a multicomputer on a chip. Clocked at 3 GHz, the Cell allows for a theoretical performance of 192 GFlops and a peak data transfer rate over the internal bus of 200 GB/s. This performance indeed makes the Cell a very attractive architecture for implementing tomographic image reconstruction algorithms.
In this study, we investigate the relative performance of a perspective backprojection algorithm when implemented on a standard PC and on the Cell processor. We compare these results to the performance achievable with FPGAs based boards and high end GPUs.
The cone-beam backprojection performance was assessed by backprojecting a full circle scan of 512 projections of 1024x1024 pixels into a volume of size 512x512x512 voxels. It took 3.2 minutes on the PC (single CPU) and is as fast as 13.6 seconds on the Cell.
Image reconstruction from rebinned helical cone-beam projection data
Show abstract
In this work, we introduced an algorithm for image reconstruction in helical cone-beam CT based upon the backprojection-filtration (BPF) algorithm. This algorithm is a backprojection-filtration-type algorithm that reconstructs images from rebinned data. It retains the properties of the original BPF algorithm in that it requires minimum data and can reconstruct ROI images from truncated data. More importantly, due to the elimination of the spatially-variant weighting factor in the backprojection, it may improve the noise properties in reconstructed images. We have performed computer-simulation studies to investigate the ROI-image reconstruction and noise properties of this algorithm, and the quantitative results verify and demonstrate the proposed algorithm.
Implementation of the FDK algorithm for cone-beam CT on the cell broadband engine architecture
Show abstract
In most of today's commercially available cone-beam CT scanners, the well known FDK method is used for solving
the 3D reconstruction task. The computational complexity of this algorithm prohibits its use for many medical
applications without hardware acceleration. The brand-new Cell Broadband Engine Architecture (CBEA) with
its high level of parallelism is a cost-efficient processor for performing the FDK reconstruction according to the
medical requirements. The programming scheme, however, is quite different to any standard personal computer
hardware. In this paper, we present an innovative implementation of the most time-consuming parts of the
FDK algorithm: filtering and back-projection. We also explain the required transformations to parallelize the
algorithm for the CBEA. Our software framework allows to compute the filtering and back-projection in parallel,
making it possible to do an on-the-fly-reconstruction. The achieved results demonstrate that a complete FDK
reconstruction is computed with the CBEA in less than seven seconds for a standard clinical scenario. Given the
fact that scan times are usually much higher, we conclude that reconstruction is finished right after the end of
data acquisition. This enables us to present the reconstructed volume to the physician in real-time, immediately
after the last projection image has been acquired by the scanning device.
Fast arbitrary-slice CT reconstruction with GPUs
Show abstract
Volume data reconstruction from cone-beam projections can be time consuming, typically reconstructing parallel axial
slices. We present a system for performing fast, arbitrary-orientation CT slice reconstruction using commodity graphics
processing units (GPUs). For 1024x1024 voxel slice reconstruction we have achieved a 89-fold performance
improvement over a CPU implementation with comparable image quality.
Hybrid helical image reconstruction in volumetric CT using ray-wise weighted cone beam filtered backprojection algorithm to significantly improve dose efficiency
Show abstract
In clinical applications, CT images are acquired in both helical and axial scans. In general, a helical scan can provide
better image quality and faster patient throughput, and thus is performed more frequently in clinic. However, the first
and last images in a helical scan are usually prescribed at the locations that are half helical turn indented from the
starting and ending points of the scan. Hence, due to image location indention, the dose efficiency of helical scanning is
usually not as good as that of axial scan in which no image location indention occurs. With increasing detector
dimension along z-direction, the dose efficiency difference between helical and axial scans is becoming more significant,
particularly at relatively small helical pitches. To match the dose efficiency of helical scan with axial scan as much as
possible, a hybrid helical CB-FBP algorithm is presented to reconstruct helical images beyond the conventional indented
image zone. The hybrid algorithm is actually a combination of the ray-wise 3D weighted CB-FBP algorithms that are
previously proposed for helical and axial CB image reconstructions, in which the ray-wise 3D weighting becomes
dependent on both helical pitch and image plane location. Phantom study shows that the conventional indented image
zone in helical scan can be extended substantially by using the presented algorithm. Consequently, the dose efficiency in
volumetric CT in helical scan can be improved significantly, especially in clinical applications where the helical pitch is
relatively low. With increasing dose discrepancy between helical and axial scans, the presented algorithm will become
more attractive in clinical applications.
Simultaneous reconstruction of activity and attenuation in PET: crosstalk measurements
Show abstract
Quantitative tomographic reconstruction based on positron emission (PET) requires attenuation correction on projections. The objective of this work is to analyze an algorithm for attenuation estimation which is used simultaneously with activity reconstruction (PET-AA) based only on emission data. The focus is on measurement of crosstalk effects on activity due to non-uniform attenuation. The derived algorithms are based on maximization of likelihood, but using minorization function approach. The activity estimator turned out to be the same as EM-ML for attenuated projections. On the other hand, the proposed algorithm for attenuation is simple and, basically, corrects the previous estimation of attenuation in a voxel by the ratio between the backprojection of estimated projections and backprojection of measured projections. Preliminary assessments were carried out via tridimensional phantoms with non-uniform attenuation emulating 3D PET with and without noise. Individually, the algorithms for attenuation and activity presented very good results with negligible crosstalk and global reconstruction error less than 2% for noiseless projections. For the case of simultaneous estimation, even without noise, the crosstalks in estimating activity and attenuation are significant (-0.223 ± 0.051) yielding global errors for activity between 10% and 20%. With noise, the error increased to 36% and the crosstalk to (-0.688 ± 0.139). The proposed algorithm for attenuation estimation presents encouraging behavior and the results confirm the necessity of methodologies to deal with crosstalk effects in PET-AA.
A fast 3D reconstruction algorithm for inverse-geometry CT based on an exact PET rebinning algorithm
Show abstract
Inverse-geometry CT (IGCT) employs a large area x-ray source array opposite a small area detector array.
The system is expected to provide sub-second volumetric imaging with isotropic resolution and no cone-beam
effects. Due to the large amount of data, it is desirable to have an exact 3D reconstruction algorithm that is
fast. Currently known IGCT algorithms are either slow, due to 3D backprojection, and/or require a reprojection
step, or are inexact. Defrise et al. developed an exact Fourier rebinning algorithm (FORE-J) for 3D PET. This
algorithm first rebins the 3D PET data into in-plane sinograms and then reconstructs the series of axial slices
using any 2D method. FORE-J is fast, exact, and efficiently uses all of the acquired PET data. We modified this
algorithm to adapt it to the IGCT geometry. Experiments were performed using a numerical "Defrise" phantom
consisting of high-intensity discs spaced in z to assess the accuracy of the modified algorithm as well as highlight
any cone-beam effects. A noise simulation was performed to analyze the noise properties of FORE-J and the
modified algorithm. The modified algorithm is very fast and slightly more accurate than the original algorithm
with a very small noise penalty in the central axial slices.
Implementation of the circle-and-line algorithm for 256-detector row CT
Show abstract
In this work we apply the circle-and-line acquisition for the 256-detector row medical CT scanner. Reconstruction is based on the exact algorithm of the FBP type suggested recently by one of the co-authors. We derived equations for the cylindrical detector, common for medical CT scanners. To minimize hardware development efforts we use ramp-based reconstruction of the circle data. The line data provides an additional term that corrects the cone beam artifacts that are caused by the incompleteness of the circular trajectory. We illustrate feasibility of our approach using simulated data and real scanned data of the anthropomorphic phantom and evaluate stability of reconstruction to motion and misalignments during the scan. The additional patient dose from the line scan is relatively low compared to the circle scan. The proposed algorithm allows cone beam artifact-free reconstruction with large cone angle.
Pre-calculation of the image quality of the simultaneous iterative reconstruction technique
Show abstract
Iterative reconstruction methods possess many advantages over analytical reconstruction methods especially if constraints can be used to regularize the reconstruction. However the main problem of iterative reconstruction algorithms is to decide when to stop the iteration.
For the Simultaneous Iterative Reconstruction Technique (SIRT) without constraints we derived a mathematical formula with which the quality of the reconstruction after a given number of iterations can be calculated. The image quality is expressed here by a special filter kernel for a FBP reconstruction which creates images with the same sharpness and noise properties as SIRT.
Further on the formula can be used to analyze the numerical stability of a certain implementation of SIRT.
Experiments show the validity of these "iteration-equivalent"-kernels with respect to sharpness and noise properties of the reconstructed images.
Hardware acceleration vs. algorithmic acceleration: can GPU-based processing beat complexity optimization for CT?
Show abstract
Three-dimensional computed tomography (CT) is a compute-intensive process, due to the large amounts of source and
destination data, and this limits the speed at which a reconstruction can be obtained. There are two main approaches to
cope with this problem: (i) lowering the overall computational complexity via algorithmic means, and/or (ii) running CT
on specialized high-performance hardware. Since the latter requires considerable capital investment into rather inflexible
hardware, the former option is all one has typically available in a traditional CPU-based computing environment.
However, the emergence of programmable commodity graphics hardware (GPUs) has changed this situation in a
decisive way. In this paper, we show that GPUs represent a commodity high-performance parallel architecture that
resonates very well with the computational structure and operations inherent to CT. Using formal arguments as well as
experiments we demonstrate that GPU-based 'brute-force' CT (i.e., CT at regular complexity) can be significantly faster
than CPU-based as well as GPU-based CT with optimal complexity, at least for practical data sizes. Therefore, the
answer to the title question: "Can GPU-based processing beat complexity optimization for CT?" is "Absolutely!"
GPU accelerated CT reconstruction for clinical use: quality driven performance
Show abstract
We present performance and quality analysis of GPU accelerated FDK filtered backprojection for cone beam computed tomography (CBCT) reconstruction. Our implementation of the FDK CT reconstruction algorithm does not compromise fidelity at any stage and yields a result that is within 1 HU of a reference C++ implementation. Our streaming implementation is able to perform reconstruction as the images are acquired; it addresses low latency as well as fast throughput, which are key considerations for a "real-time" design. Further, it is scaleable to multiple GPUs for increased performance. The implementation does not place any constraints on image acquisition; it works effectively for arbitrary angular coverage with arbitrary angular spacing. As such, this GPU accelerated CT reconstruction solution may easily be used with scanners that are already deployed. We are able to reconstruct a 512 x 512 x 340 volume from 625 projections, each sized 1024 x 768, in less than 50 seconds. The quoted 50 second timing encompasses the entire reconstruction using bilinear interpolation and includes filtering on the CPU, uploading the filtered projections to the GPU, and also downloading the reconstructed volume from GPU memory to system RAM.
A fast and high-quality cone beam reconstruction pipeline using the GPU
Show abstract
Cone beam scanners have evolved rapidly in the past years. Increasing sampling resolution of the projection
images and the desire to reconstruct high resolution output volumes increases both the memory consumption
and the processing time considerably. In order to keep the processing time down new strategies for memory
management are required as well as new algorithmic implementations of the reconstruction pipeline. In this
paper, we present a fast and high-quality cone beam reconstruction pipeline using the Graphics Processing Unit
(GPU). This pipeline includes the backprojection process and also pre-filtering and post-filtering stages. In
particular, we focus on a subset of five stages, but more stages can be integrated easily. In the pre-filtering
stage, we first reduce the amount of noise in the acquired projection images by a non-linear curvature-based
smoothing algorithm. Then, we apply a high-pass filter as required by the inverse Radon transform. Next,
the backprojection pass reconstructs a raw 3D volume. In post-processing, we first filter the volume by a ring
artifact removal. Then, we remove cupping artifacts by our novel uniformity correction algorithm. We present
the algorithm in detail. In order to execute the pipeline as quickly as possible we take advantage of GPUs that
have proven to be very fast parallel processors for numerical problems. Unfortunately, both the projection images
and the reconstruction volume are too large to fit into 512 MB of GPU memory. Therefore, we present an efficient
memory management strategy that minimizes the bus transfer between main memory and GPU memory. Our
results show a 4 times performance gain over a highly optimized CPU implementation using SSE2/3 commands.
At the same time, the image quality is comparable to the CPU results with an average per pixel difference of
10-5.
Improvement of the OPED algorithm by means of introducing an integration into the evaluation process
Show abstract
The tomographic method based on the orthogonal polynomial expansion on disc (OPED) was presented at SPIE
conference of Medical Imaging 2006. We could show already some advantages compared to FBP as it is commonly used
in today's CT systems. However, OPED did show for some specific cases some noise in the reconstructed images and
even artefacts, mainly an aliasing. We have found that the OPED algorithm can be essentially improved by integrating
the polynomial over the whole area belonging to the pixel instead of assigning to the whole pixel the polynomial value
calculated just for one point of this pixel (typically bottom left). This advantageous implementation is effective in view
of reduction of the aliasing artefacts and noise without affecting the resolution. This can be fulfilled effectively for
OPED due to its simple structure.
The Juggler algorithm: a hybrid deformable image registration algorithm for adaptive radiotherapy
Show abstract
Fast deformable registration can potentially facilitate the clinical implementation of adaptive radiation therapy (ART), which allows for daily organ deformations not accounted for in radiotherapy treatment planning, which typically utilizes a static organ model, to be incorporated into the fractionated treatment. Existing deformable registration algorithms typically utilize a specific diffusion model, and require a large number of iterations to achieve convergence. This limits the online applications of deformable image registration for clinical radiotherapy, such as daily patient setup variations involving organ deformation, where high registration precision is required. We propose a hybrid algorithm, the "Juggler", based on a multi-diffusion model to achieve fast convergence. The Juggler achieves fast convergence by applying two different diffusion models: i) one being optimized quickly for matching high gradient features, i.e. bony anatomies; and ii) the other being optimized for further matching low gradient features, i.e. soft tissue. The regulation of these 2 competing criteria is achieved using a threshold of a similarity measure, such as cross correlation or mutual information. A multi-resolution scheme was applied for faster convergence involving large deformations. Comparisons of the Juggler algorithm were carried out with demons method, accelerated demons method, and free-form deformable registration using 4D CT lung imaging from 5 patients. Based on comparisons of difference images and similarity measure computations, the Juggler produced a superior registration result. It achieved the desired convergence within 30 iterations, and typically required <90sec to register two 3D image sets of size 256×256×40 using a 3.2 GHz PC. This hybrid registration strategy successfully incorporates the benefits of different diffusion models into a single unified model.
Metal artifacts correction in cone-beam CT bone imaging
Show abstract
Cone-beam CT (CBCT) technique is needed by orthopaedists in their new studies to monitor bone volume growth
and blood vessel growth of structural bone grafts used in reconstruction surgery. However, titanium plate and screws,
which are commonly used to connect bone grafts to host bones, can cause severe streaking artifacts and shading
artifact in the reconstructed images due to their high attenuation of x-rays. These metal artifacts will distort the
information of the bone and cause difficulties when measuring bone volume growth and the inside blood vessel
growth. To solve this problem and help orthopaedists quantitatively record the growth of bone grafts, we present a
three-dimensional metal artifact correction technique to correct the streaking artifacts generated by titanium implants.
In this project not only the artifacts need to be corrected but also the correct information of the bone is required in the
image for the quantitative measurements. Both phantom studies and animal studies were conducted to test this
correction method. Images without metal correction and images with metal correction were compared together, as well
as the reference bone images acquired without metal. It's shown the streaking and shading artifacts were greatly
reduced after metal correction. The accuracy of bone volume measurements was also greatly increased by 79% for
phantom studies and 53% for animal studies.
Sinogram restoration for ultra-low-dose x-ray multi-slice helical CT by nonparametric regression
Show abstract
During the last decade, x-ray computed tomography (CT) has been applied to screen large asymptomatic smoking and nonsmoking populations for early lung cancer detection. Because a larger population will be involved in such screening exams, more and more attention has been paid to studying low-dose, even ultra-low-dose x-ray CT. However, reducing CT radiation exposure will increase noise level in the sinogram, thereby degrading the quality of reconstructed CT images as well as causing more streak artifacts near the apices of the lung. Thus, how to reduce the noise levels and streak artifacts in the low-dose CT images is becoming a meaningful topic.
Since multi-slice helical CT has replaced conventional stop-and-shoot CT in many clinical applications, this research mainly focused on the noise reduction issue in multi-slice helical CT. The experiment data were provided by Siemens SOMATOM Sensation 16-Slice helical CT. It included both conventional CT data acquired under 120 kvp voltage and 119 mA current and ultra-low-dose CT data acquired under 120 kvp and 10 mA protocols. All other settings are the same as that of conventional CT. In this paper, a nonparametric smoothing method with thin plate smoothing splines and the roughness penalty was proposed to restore the ultra-low-dose CT raw data. Each projection frame was firstly divided into blocks, and then the 2D data in each block was fitted to a thin-plate smoothing splines' surface via minimizing a roughness-penalized least squares objective function. By doing so, the noise in each ultra-low-dose CT projection was reduced by leveraging the information contained not only within each individual projection profile, but also among nearby profiles. Finally the restored ultra-low-dose projection data were fed into standard filtered back projection (FBP) algorithm to reconstruct CT images. The rebuilt results as well as the comparison between proposed approach and traditional method were given in the results and discussions section, and showed effectiveness of proposed thin-plate based nonparametric regression method.
Two-dimensional reconstruction algorithm of an inverse-geometry volumetric CT system
Show abstract
An inverse-geometry volumetric CT (IGCT) system uses a large source array opposite a smaller detector array.
Conventional 2D IGCT reconstruction is performed by using gridding. We describe a 2D IGCT reconstruction algorithm
without gridding. The IGCT raw data can be viewed as being composed of many fan beams, each with a detector at its
focus. Each projection is undersampled but the missing samples are provided by other views. In order to get high spatial
resolution, zeros are inserted between acquired projection samples in each fan beam, and reconstruction is performed
using a direct fan beam reconstruction algorithm. Initial IGCT reconstruction results showed ringing artifacts caused by
fact that the rho samples in the ensemble of views are not equally spaced. We present a new method for correcting the
errors that reduces the artifacts to below one Hounsfield Unit
FFT and cone-beam CT reconstruction on graphics hardware
Show abstract
Graphics processing units (GPUs) are increasingly used for general purpose calculations. Their pipelined architecture
can be exploited to accelerate various parallelizable algorithms. Medical imaging applications are
inherently well suited to benefit from the development of GPU-based computational platforms. We evaluate in
this work the potential of GPUs to improve the execution speed of two common medical imaging tasks, namely
Fourier transforms and tomographic reconstructions. A two-dimensional fast Fourier transform (FFT) algorithm
was GPU-implemented and compared, in terms of execution speed, to two popular CPU-based FFT routines.
Similarly, the Feldkamp, David and Kress (FDK) algorithm for cone-beam tomographic reconstruction was implemented
on the GPU and its performance compared to a CPU version. Different reconstruction strategies
were employed to assess the performance of various GPU memory layouts. For the specific hardware used, GPU
implementations of the FFT were up to 20 times faster than their CPU counterparts, but slower than highly
optimized CPU versions of the algorithm. Tomographic reconstructions were faster on the GPU by a factor up to
30, allowing 2563 voxel reconstructions of 256 projections in about 20 seconds. Overall, GPUs are an attractive
alternative to other imaging-dedicated computing hardware like application-specific integrated circuits (ASICs)
and field programmable gate arrays (FPGAs) in terms of cost, simplicity and versatility. With the development
of simpler language extensions and programming interfaces, GPUs are likely to become essential tools in medical
imaging.
Ray-wise weighted helical cone beam filtered backprojection algorithm for image reconstruction under moderate cone angle
Show abstract
With an accelerated pace the CT technology has achieved the latest milestone - cone beam volumetric CT with 40mm
detector coverage. To obtain an optimized image reconstruction solution for future cone beam VCT systems, the raywise
weighted helical CB-FBP algorithm, which was proposed by us to reconstruct image under cone angles up to 4.25°,
is optimized and evaluated in this study to verify its imaging performance for image reconstruction under larger cone
angles up to 8.5°. The ray-wise weighted helical CB-FBP algorithm proposed by us possesses two important features: (a)
tangential filtering that is naturally implemented via row-wise fan-to-parallel rebinning to maintain spatial resolution
along patient's longitudinal direction; and (b) 3D weighting that is a ray-wise optimization process to obtain image
quality controllability. By using computer-simulated phantoms, such as the helical body and humanoid head phantoms, it
has been shown that the ray-wise weighted helical CB-FBP algorithm can provide a well balanced imaging performance
over helical pitches while a large field of view (FOV) can be maintained. It is the optimized ray-wise weighting that
enables the proposed CB-FBP algorithm performs well at larger cone angle. Based on the experimental evaluation, it is
believed that the ray-wise weighted helical CB-FBP algorithm can be a candidate solution for image reconstruction in
future cone beam VCT systems with detectors corresponding to larger cone angles up to 8.5° (~ 80 mm detector z coverage).
ECG gated circular cone-beam multi-cycle short-scan reconstruction algorithm
Show abstract
In this contribution, the results of a phantom study for in-stent restenosis imaging with ECG gated continuous
circular acquisition and reconstruction are summarized. Different rotation speeds and angular ranges are used
to enable high resolution 3D and 4D reconstruction of objects covered by the cone at a high temporal resolution.
Though the detector coverage of today's CT scanners is not large enough to irradiate the complete human heart,
the coverage is sufficient to image smaller objects like conventional stents. We applied the proposed method to
the visualization of an in-stent re-stenoses phantom covered by a clinical stent, attached to a dynamic heart
phantom. The method delivers images of stents in vitro at an excellent visibility and is able to rule out in-stent
occlusions.
Application of fast radon transform to CT scanners: difficulties and solutions
Show abstract
As a tomographic reconstruction algorithm, the recently proposed "Fast Radon Transform" (FRT) has some computational advantages. To prove its practical importance the technical difficulties associated with its application to fan-beam CT scanners as well as Spiral/Helical CT system are solved here. Some techniques are described to convert the actual fan-beam data or the spiral/helical CT data to parallel-beam data required for the FRT algorithm in order to reconstruct the CT images. Simulation results are presented to validate the complete method.
A ray-tracing backprojection algorithm for cone beam CT
Show abstract
We have developed a ray-tracing backprojection (RTB) to back-project all the detector pixels into
the image domain of cone beam CT (CBCT). The underlying mathematic framework is the FDK
reconstruction. In this method, every ray recorded by the flat panel detector is traced back into the
image space. In each voxel of the imaging domain, all the rays contributing to the formation of
the CT image are summed together weighted by each rays' intersection length with the voxel. The
RTB is similar to a reverse process of x-ray transmission imaging, as opposed to the conventional
voxel-driven backprojection (VDB). In the RTB, we avoided interpolation and pixel binning
approximations, achieved better spatial resolution and eliminated some image artifacts. We have
successfully applied the RTB in phantom studies on the Varian On Board Imager CBCT. The images of the Catphan CTP404 module show more accurate representation of the oblique ramps in the measurement of slice thickness, and more accurate determination of slice thickness with the RTB than with VDB. The RTB also shows higher spatial resolution than the VDB in the studies of a high contrast resolution phantom.
Hardware-accelerated cone-beam reconstruction on a mobile C-arm
Michael Churchill,
Gordon Pope,
Jeffrey Penman,
et al.
Show abstract
The three-dimensional image reconstruction process used in interventional CT imaging is computationally demanding. Implementation on general-purpose computational platforms requires a substantial time, which is undesirable during time-critical surgical and minimally invasive procedures. Field Programmable Gate Arrays (FPGA)s and Graphics Processing Units (GPU)s have been studied as a platform to accelerate 3-D imaging. FPGA and GPU devices offer a reprogrammable hardware architecture, configurable for pipelining and high levels of parallel processing to increase computational throughput, as well as the benefits of being off-the-shelf and effective 'performance-to-watt' solutions. The main focus of this paper is on the backprojection step of the image reconstruction process, since it is the most computationally intensive part. Using the popular Feldkamp-Davis-Kress (FDK) cone-beam algorithm, our studies indicate the entire 2563 image reconstruction process can be accelerated to real or near real-time (i.e. immediately after a finished scan of 15-30 seconds duration) on a mobile X-ray C-arm system using available resources on built-in FPGA board. High resolution 5123 image backprojection can be also accomplished within the same scanning time on a high-end GPU board comprising up to 128 streaming processors.
Implementation and evaluation of 4D cone beam CT (CBCT) reconstruction
Show abstract
Tumor angiogenesis is the process by which new blood vessels are formed from the existing vessels in a tumor to
promote tumor growth. Tumor angiogenesis has important implications in the diagnosis and treatment of various
solid tumors. Flat panel detector based cone beam CT opens up a new way for detection of tumors, and tumor
angiogenesis associated with functional CBCT has the potential to provide more information than traditional
functional CT due to more overall coverage during the same scanning period and the reconstruction being isotropic
resulting in a more accurate 3D volume intensity measurement. A functional study was conducted by using CBCT
to determine the degree of the enhancement within the tumor after injecting the contrast agent intravenously. For
typical doses of contrast material, the amount of enhancement is proportional to the concentration of this material
within the region of interest. A series of images obtained at one location over time allows generation of timeattenuation
data from which a number of semi-quantitative parameters, such as enhancement rate, can be determined.
Computer simulations prove the superiority of half scan over full scan in terms of more accurately delineating the
time-intensity curve, and all the simulation parameter settings are based on the actual CBCT prototype. An
experiment study was conducted on our prototype CBCT system, and a full and half scan scheme is used to
determine the time-intensity curve within the ROI of the mouse. The CBCT has an x-ray tube, a gantry with slip
ring technology, and a 40x30 cm Varian Paxscan 4030CB real time FPD.
A statistical approach to high-quality CT reconstruction at low radiation doses for real-time guidance and navigation
Show abstract
The advent of 64-slice computed tomography (CT) with high-speed scanning makes CT a highly attractive and powerful tool for navigating image-guided procedures. For interactive navigation, scanning will need to be performed over extended time periods or even continuously. However, continuous CT is likely to expose the patient and the physician to potentially unsafe levels of radiation. Before CT can be used appropriately for navigational purposes, the dose problem must be solved. Simple dose reduction is not adequate, because it degrades image quality. This problem can be overcome if the traditional filtered back-projection (FBP) reconstruction is replaced with the maximum likelihood expectation maximization (MLEM) approach. MLEM is more accurate in that it incorporates Poisson statistics of the noisy projection data, especially at low doses. Our study shows that MLEM reconstruction is able to reduce x-ray dose from 200 to 11 mAs (the lowest dose-simulator setting in the present study) without significant image degradation. Taking advantage of modern CT scanners and specialized hardware, it may be possible to perform continuous CT scanning at acceptable radiation doses for intraoperative visualization and navigation.
Missing data estimation for fully 3D spiral CT image reconstruction
Show abstract
It is often the case in tomography that a scanner is unable to collect a full set of projection data. Reconstruction
algorithms that are not set up to handle this type of problem can lead to artifacts in the reconstructed images
because the assumptions regarding the size of the image space and/or data space are violated. In this study,
we apply two recently developed geometry-independent methods to fully 3D multi-slice spiral CT image reconstruction.
The methods build upon an existing statistical iterative reconstruction algorithm developed by our
group. The first method reconstructs images without the missing data, and the second method seeks to jointly
estimate the missing data and attenuation image. We extend the existing results for the 2D fan-beam geometry
to multi-slice spiral CT in an effort to investigate some challenges in 3D, such as the long object problem. Unlike
the original formulation of the reconstruction algorithms, a regularization term was added to the objective
function in this work. To handle the large number of computations required by fully 3D reconstructions, we
have developed an optimized parallel implementation of our iterative reconstruction algorithm. Using simulated
and clinical datasets, we demonstrate the effectiveness of the missing data approaches in improving the quality
of slices that have experienced truncation in either the transverse or longitudinal direction.
Fast variance predictions for 3D cone-beam CT with quadratic regularization
Show abstract
Fast and accurate variance/covariance predictions are useful for analyzing the statistical characteristics of the reconstructed
images and may aid regularization parameters selection. The existing methods, the matrix-based method and its DFT approximations,
are impractical for realistic data size in X-ray CT. We have previously addressed this problem in 2D fan-beam
CT by proposing "analytical" approaches, the simplest of which requires computation equivalent to one backprojection and
some summations. This paper extends these approaches to 3D step-and-shoot "cylindrical" cone-beam CT.
Iterative extended field of view reconstruction
Show abstract
Incomplete data due to the object extent beyond the scanning field of view (SFOV) is a common
problem in computed tomography. In these cases, there are parts of the object to be reconstructed
for which only incomplete projections of less than 180o are available. Applying iterative algorithms
like algebraic reconstruction technique (ART) or simultaneous algebraic reconstruction Technique
(SART) onto the problem of truncated projections can not produce a satisfying solution unless
special constraints are used. To regularize the reconstruction algorithm, we extend iterative
reconstruction algorithms by introducing information regarding the statistics of the attenuation
values of the reconstructed object in terms of the log likelihood function of attenuation values. This
information can be taken from the regions of the image still inside the SFOV but close to the region
where the object exceeds the SFOV. The information can be utilized in an algebraic reconstruction
method by adding a constraint term to the cost function that shall be minimized.
Experiments show that for not severely truncated projections, as they are common for CT
applications, including this information yields good estimates about the object.
Iterative reconstruction for multi-source inverse geometry CT: a feasibility study
Show abstract
In a 3rd generation CT system, a single source projects the entire field of view (FOV) onto a large detector opposite to
the source. In multi-source inverse geometry CT imaging, a multitude of sources sequentially project complementary
parts of the FOV on a much smaller detector. These sources may be distributed in both the trans-axial and axial
directions and jointly cover the entire FOV. Multi-source CT has several important advantages, including large axial
coverage, improved dose-efficiency, and improved spatial resolution. One of the challenges of this concept is to ensure
that no artifacts emerge in the reconstructed images where the sampling switches from one source to the next. This
work studies iterative reconstruction for multi-source imaging and focuses on the appearance of such artifacts. For that
purpose, phantom data are simulated using a realistic multi-source CT geometry, iteratively reconstructed and inspected
for artifact content. More realistic experiments using rebinned clinical datasets (emulating a multi-source CT system)
have also been performed. The results confirm the feasibility of artifact-free multi-source CT imaging in both full-scan
and half-scan situations.