This PDF file contains the front matter associated with SPIE Proceedings Volume XXXXX, including the Title Page, Copyright information, Table of Contents, and Conference Committee listing.

Photoacoustic tomography (PAT) is a noninvasive, high-resolution imaging modality, capable of providing functional and molecular information of various pathologies such as cancer. In most PAT systems, the effect of tissue heterogeneity (i.e. variations in acoustic properties such as speed of sound and acoustic attenuation) is neglected. This is due to the lack of information about acoustic properties of tissue and complexity of a model to compensate for these variations. We have been developing a full-ring PAT system consists of an omni-directional illumination and a ring-based acoustic detection. In this study, we investigate using a model-based method that employs light diffusion (Monte-carlo) and acoustic wave propagation (K-wave) to compensate for both optical and acoustic heterogeneity of the tissue and provide fully compensated (i.e. quantitative) PAT images for our full-ring PAT system. To demonstrate the feasibility of providing fully compensated PAT images, in silico studies were performed in which a heterogeneous breast-tissue-mimicking phantoms were computationally generated. The map includes optical (µa, µs, g) and acoustic properties (ρ, Cs) of the fatty, fibroglandular and breast lesions. The monte-carlo light diffusion model was first utilized to generate the fluence map and thus the initial photoacoustic pressure fields (P0) within the tissue. Following to the generation of P0, the propagation of acoustic waves through a heterogeneous medium was simulated using K-wave. Using a ring-geometry ultrasound transducers (N=256), the pressure waves were received and were utilized to reconstruct PAT images. Our results indicate the PAT improvement using acoustic and optical compensation and more importantly the feasibility of achieving “quantitative” PAT images upon compensating for tissue heterogeneity. 100 Word Abstract:

The development and investigation of PACT as an effective neuroimaging modality is highly warranted. Amajor challenge in transcranial PACT brain imaging is to compensate for aberrations in the measured datadue to the propagation of the photoacoustic wavefields through the skull. To properly account for these effects,image reconstruction methods in transcranial PACT require knowledge of the spatial distribution of the acousticparameters of the skull. However, estimating the spatial distribution of the acoustic parameters of the skullremains challenging. Inspired by our observation that information about the distribution of skull acousticparameters is encoded in PACT measurements, we propose to jointly reconstruct the initial pressure distributionand the spatial distributions of the acoustic parameters of the skull from PACT data alone. In this study, weimplement a joint image reconstruction algorithm to estimate both the initial pressure distribution as well asthe spatial distribution of the acoustic parameters of the skull for 3D transcranial PACT. The joint estimationof the initial pressure and spatial distributions of the acoustic parameters of the skull from PACT data alone isunstable. To overcome this instability, we propose to incorporate prior information about the acoustic propertiesof the skull from adjunct image data. The developed joint reconstruction algorithm is validated and investigatedthrough computer-simulation studies.

Among various types of cancer, breast cancer is considered to be the most common that affects thousands of women all over the world. Several imaging tools are being used for breast cancer detection and diagnosis. Mammography and B-mode ultrasound (US) are the primary screening tools for breast lesions. However, mammography is limited with low sensitivity especially in women with dense breasts, who appear to be at higher risk of breast cancer. Additionally, the B-mode US suffers from low specificity in the differential diagnosis of breast lesions. Therefore, it is clinically significant to develop screening techniques that could eliminate previous limitations. Photoacoustic (PA) has been showing potential for early stage detection and staging breast cancer due to its unique abilities to acquire functional and molecular information of the breast lesions. We have developed an optimized US and PA tomography system, which uses custom designed all reflective based optics to create an omnidirectional ring-shaped beam to illuminate a cross-section of the breast tissue and acquire thegenerated acoustic waves by using a full-ring US transducer. The developed PA tomography (PAT) system can potentially make a more uniform illumination of the breast tissue and more importantly enhance the imaging depth. In this study, development of the full-ring illumination and the results of our initial feasibility US/PA tests are presented.

Interest in ultrasound perfusion imaging has grown with the development of more sensitive algorithms to detect slow blood flow. Unfortunately, there are not many phantoms that can be used to evaluate these techniques. Some have used small linear tubes, while others have adapted dialysis cartridges. Here we propose a technique using conventional gelatin cast around a sacrificial polymer network. Specifically, we form a gelatin phantom, doped with graphite scatterers to mimic the diffuse scattering in soft tissue, around a polymer resin. The resin structure can be dissolved leaving behind a network of small randomly oriented channels that are connected to a large channel which is connected to a pump to perfuse blood mimicking fluid through the phantom. The phantoms were qualitatively demonstrated to show perfusion through visual confirmation and the speckle SNR, and speed of sound were calculated.

Treatment of hepatocellular carcinoma (HCC) with sorafenib, a multikinase inhibitor, results in decreased microvessel density associated with increased levels of tumor hypoxia. However, the response rate is relatively poor, and recently it has been shown that tumor hypoxia and perfusion have predictive correlations with HCC response to sorafenib. In this study, we have investigated the correlation of oxygen saturation (SO₂) and perfusion, estimated using photoacoustic-ultrasonic (PAUS) imaging, to the sorafenib treatment response in an orthotopic rat model of HCC. Following spectroscopic photoacoustic (sPA) imaging, microbubble contrast was introduced and harmonic imaging data were acquired for perfusion measurements. An FEM-based fluence correction model based on the diffusion approximation with empirically estimated tissue surface fluence and an SNR-based thresholding approach have been developed and validated on ex vivo and in vivo rat data to estimate SO₂ using sPA imaging. The SO₂ estimate has been obtained by solving an iterative minimization problem and then thresholded based on a pixel-wise empirically estimated SNR mask. For the treated cohort, the results show that the change in SO₂ during an oxygen challenge is positively correlated with disease progression, while it is negatively correlated for the untreated cohort. Additionally, perfusion was significantly decreased in the treated group compared to baseline pretreatment and untreated cohort measurements. The reduced treatment-mediated perfusion leads to lack of oxygen supply and thus reduced oxygen levels. This study shows the potential of PAUS estimation of SO₂ and perfusion to monitor and predict HCC sorafenib treatment response, ultimately leading to improved future treatment.

Tissue clutter caused by patient and sonographer hand motion makes perfusion ultrasound imaging difficult. We previously introduced an adaptive frequency and amplitude demodulation scheme to address this challenge. Our initial implementation used a conventional high-pass infinite impulse response (IIR) filter to attenuate the tissue signal after applying adaptive demodulation. However, other groups have shown that singular value decomposition (SVD) filtering is superior to conventional frequency domain filters. Here we evaluate the SVD filter both in comparison and in conjunction with our proposed adaptive demodulation technique. Blood-to-background SNRs were compared using power Doppler images made from single small vessel simulations with realistic tissue clutter. Additionally, filtering methods were qualitatively assessed using power Doppler images of a cut-in-half perfusion-mimicking phantom. Furthermore, in vivo power Doppler images were compared before and after muscle contraction. SVD filtering with adaptive demodulation resulted in a 7dB increase in simulated blood-to-background SNR compared to a conventional IIR filter and a 54.6% increase in power after in vivo muscle contraction compared to a 1.74% increase using a conventional IIR filter.

Image acquired during free breathing using contrast enhanced ultrasound (CEUS) hepatic perfusion imaging exhibits a periodic motion pattern. It needs to be compensated for if a further accurate quantification of the hepatic perfusion analysis is to be executed. A respiratory motion compensation strategy for CEUS imaging by using image clustering is proposed in this work. The proposed strategy separated the dual mode image to tissue image and contrast image firstly. Then, the image subsequences based on the tissue image are determined by using sparse subspace clustering (SSC) method. Finally, the motion compensated contrast images are acquired by using the position mapping. The strategy was tested on ten CEUS hepatic perfusion image sequences. Quantitative and visual comparisons demonstrate that the proposed strategy can compensate the misalignment of ultrasound hepatic perfusion image sequence caused by respiratory motion in free-breathing.

Introduction: To improve carotid artery stenting (CAS), more information about the functioning of the stent is needed. Therefore, a method that can image the flow near and around a stent is required. The aim of this study was to evaluate the performance of high-frame-rate contrast-enhanced ultrasound (HFR CEUS) in the presence of a stent. Methodology: HFR CEUS acquisitions of a carotid artery phantom, a silicone tube with diameter 8 mm, with and without a stent were acquired at transmit voltages of 2V, 4V and 10V using a Verasonics ultrasound system and C5-2 probe. Different concentrations of ultrasound contrast agent (UCA) were tested in a blood mimicking fluid (BMF). Particle image velocimetry (PIV) analysis was performed on Singular Value Decomposition (SVD) filtered images. Mean and peak velocities, and correlation coefficients were compared between stented and non-stented regions. Also, experimental results were compared with theoretical and numerical models. Results: The averaged experimental mean velocity (0.113 m/s) was significant lower than the theoretical and numerical mean velocity (0.129 m/s). The averaged experimental peak velocity (0.152 m/s) was significant lower than the theoretical and numerical peak velocity (0.259 m/s). Correlation coefficients and averaged mean velocity values were lower (difference of 0.022 m/s) in stented regions compared to non-stented regions. Conclusion: In vitro experiments showed an underestimation of mean and peak velocities in stented regions compared to non-stented regions. However, the microbubbles can be tracked efficiently and the expected laminar flow profile can be quantified using HFR CEUS near and around a stent.

Pneumonia is one of the most common acute respiratory infections among pediatric populations worldwide. Ultrasound is becoming increasingly important in the diagnosis of lung diseases as a more portable and safer alternative to X-ray imaging. In the current work, we present a new automatic system for detection of B-lines, one of the distinctive features of pneumonic ultrasound scans, using amplitude modulation-frequency modulation (AM-FM). Features are evaluated on 109 videos obtained from 100 pediatric patients using a Verasonics V1 scanner. Further, the results were compared to the ones obtained with a previously published spectral feature (SF) method. Sensitivities of 92% and 83% and specificities of 91% and 70% were obtained on zone-1 and zone-2 of the lungs, respectively. In contrast, the SF method provided sensitivities of 72% and 68% in zone-1 and zone-2, respectively, and specificities of 68% and 46% in zone-1 and zone-2, respectively. In addition, the AM-FM method allowed increasing the F1-score when compared to the SF method from 70% to 87% and from 61% to 78% in zone-1 and zone-2, respectively. The results suggest the proposed method may be useful for the computer assisted diagnosis of pneumonia.

High-speed acquisition of ultrasound volume data is needed for fetal cardiac diagnosis. The heartbeat of healthy fetus is 120-160 times per second. Therefore, the acquiring method based on plane wave compounding has been developed to achieve both volume rate and image quality. However, in the conventional plane wave compounding method, the sufficient image quality couldn’t be obtained when the acquiring speed is about 150 volumes per second. Compressed Sensing has been applied to improve the image quality and reduce the number of compounding in the previous work [1], but huge memory (458GB) is required. In this paper, we propose a method to acquire high quality volume image at high-speed with reasonable hardware resources. The basic concept is to improve the image quality of each plane wave image before compounding by simple signal processing. Generally, the acoustic noise of plane wave image generated by the diffraction depends on the steering angle. In our method, the acoustic noise is adaptively reduced depending on the steering angle, and the wavelet shrinkage [2] is used as a basic noise reduction algorithm. In the experimental results, the acoustic noise is reduced by 22dB with only 20MB memory usage for radio frequency simulation data. As a result, we achieved high-speed data acquisition of 167 volumes per second.

Harmonic imaging has been a breakthrough for the quality of clinical ultrasound imaging, greatly reducing the acoustic clutter that typically reduces in vivo image quality. The generation of the second harmonic signal by non-linear propagation is optimized for a focused transmission in which focal gain raises the fundamental pressure. However, the signal-to-noise ratio (SNR) of the harmonic backscattered signal is lower than for the fundamental frequency. We demonstrate the application of Retrospective Encoding For Conventional Ultrasound Sequences (REFoCUS), a framework for performing spatial decoding of existing pulse sequences irrespective of transducer or scan geometry, to improve transmit depth of field and SNR in harmonic imaging. Unlike other spatial coding methods, REFoCUS allows for maintaining a transmit focus and the corresponding harmonic generation. We demonstrate the ability to recover the effective transmit element sources that would linearly produce the observed harmonic fields, enabling individual transmit element processing. The technique is applied to in vivo liver and fetal targets to produce improved image quality away from the original transmit focal depth using the same data.

Collecting high quality volumetric ultrasound (US) data using freehand scanning is challenging. The quality of the final 3DUS image is highly related to the applied scanning protocol and the subsequently used reconstruction method. The protocol should ensure the sonographer collects sufficient data of satisfactory quality for an accurate reconstruction.

In this study we developed a real-time reconstruction method that provides visual feedback during scanning. The feedback indicates the areas, of which the sonographer should collect more data. The method was tested by acquiring US data of a breast phantom in a setup mimicking freehand scanning which consisted of a linear transducer mounted in a translation stage that also allowed rotation.

To reconstruct the volume in real-time on a target grid of 0.5x0.5x0.5mm, we applied a simplified Voxel Nearest Neighbor (VNN) method, i.e., only the closest to B-mode plane voxels were updated. Furthermore, voxels were updated only when their projection on the B-mode plane was closer to the transducer surface than in the previous scan planes. Interpolation was performed within the acquired volume to fill in the holes where sufficient data were available. Sub-volumes with insufficient data were visualized in the reconstructed volume (update rate 50 Hz). This visual feedback can guide the sonographer during freehand scanning to improve the quality of the reconstructed 3DUS images. Cross-sections of the reconstructed data were compared to the independently acquired B-mode images and confirmed that our real-time method of low computational complexity provided accurate volumetric ultrasound images.

Previous studies demonstrated that our aperture domain model image reconstruction (ADMIRE) beamforming algorithm mitigates some common ultrasound imaging artifacts, which may increase ultrasound's clinical utility and reliability. Specifically, ADMIRE can suppress clutter caused by reverberation, off-axis scattering and wavefront aberration. Along with this, we demonstrated that ADMIRE is robust to model-mismatch caused by gross sound speed deviation. These findings suggest that ADMIRE may be an effective tool to provide high quality images in real clinical applications. Many of our previous effort have occurred on research platforms, but it is thought that dedicated clinical systems have better front-end electronics and transducers compared to research oriented platforms. If this is true then it is important to perform in vivo evaluations using the highest quality data possible in order to appropriately characterize (and not overemphasize possible) algorithmic gains. To this end, we modified a Siemens ACUSON SC2000 ultrasound system to capture I/Q channel signals. We acquired channel data using a full synthetic receive sequence. We also acquired channel data in conjunction with pulse inversion sequencing to obtain harmonic images. In this study, we collected data from a tissue-mimicking phantom and a human subject's abdomen and liver. We reconstructed both fundamental and harmonic B-mode images before and after applying ADMIRE. We then measured contrast and contrast-to-noise ratio (CNR). When comparing in vivo images, ADMIRE using low and high degrees of freedom improves contrast by 12.2 ± 2.6 dB and 2.5 ± 0.5 dB, respectively, relative to fundamental delay-and-sum(DAS) B-mode, and boosts contrast by 8.7 ± 3.7 dB and 2.0 ± 0.7 dB, respectively, with harmonic B-mode images.

We developed a method that uses deep neural networks (DNNs) to suppress off-axis scattering in ultrasound images. This approach operates in the frequency domain and networks were trained using the simulated responses from individual point targets. The network inputs consisted of the separated in-phase and quadrature components observed across the aperture of the array. The output had the same structure as the input and an inverse short- time Fourier transform was used to convert the processed data back to the time domain. In this work, we examined the noise handling characteristics of the DNN beamformer and also the relation between final image quality and the loss function for training networks.

Waveform inversion for ultrasound computed tomography (USCT) is a promising imaging technique for breast cancer screening. However, the improved spatial resolution and the ability to constrain multiple parameters simultaneously demand substantial computational resources for the recurring simulations of the wave equation. Hence, it is crucial to use fast and accurate methods for numerical wave propagation, on the one hand, and to keep the number of required simulations as small as possible, on the other hand. We present an efficient strategy for acoustic waveform inversion that combines (i) a spectral-element continuous Galerkin method for solving the wave equation, (ii) conforming hexahedral mesh generation to discretize the scanning device, (iii) a randomized descent method based on mini-batches to reduce the computational cost for misfit and gradient computations, and (iv) a trust-region method using a quasi-Newton approximation of the Hessian to iteratively solve the inverse problem. This approach combines ideas and state-of-the-art methods from global-scale seismology, large-scale nonlinear optimization, and machine learning. Numerical examples for a synthetic phantom demonstrate the efficiency of the discretization, the effectiveness of the mini-batch approximation and the robustness of the trust-region method to reconstruct the acoustic properties of breast tissue with partial information.

Waveform inversion is a promising method for ultrasound computed tomography able to produce high-resolution images of human breast tissue. However, the computational complexity of waveform inversion remains a considerable challenge, and the costs per iteration are proportional to the number of emitting transducers. We propose a twofold strategy to accelerate the time-to-solution by identifying the optimal number and location of emitters using sequential optimal experimental design (SOED). SOED is a powerful tool to iteratively add the most informative transducer or remove redundant measurements, respectively. This approach simultaneously provides optimized transducer configurations and a cost-benefit curve that quantifies the information gain versus the computational cost.

First, we propose a method to identify the emitters that provide reconstructions with minimal expected uncertainties. Using a Bayesian approach, model uncertainties and resolution can be quantified with the trace of the posterior covariance. By linearizing the wave equation, we can compute the posterior covariance using the inverse of the Gauss-Newton approximation of the Hessian. Furthermore, this posterior is independent of the breast model and the experimental data, thus enabling pre-acquisition experimental optimization. Then, for the post-acquisition inversion, we present an approach to select a subsample of sources that accurately approximates the full gradient direction in each iteration. We control the convergence of the angular differences between consecutive gradient directions by randomly adding new emitters into the subsample.

We present synthetic studies in 2D and 3D that consider a ring-shaped and a semi-ellipsoidal scanning device, respectively. Numerical results suggest that the provided methods have the potential to identify redundancies from the corresponding cost-benefit curves. Furthermore, the gradient direction rapidly converges to the direction of the full gradient, which appears to be independent of the model and the emitter locations.

The distorted Born iterative method (DBI) is used to solve the inverse scattering problem in the ultrasound tomography with the objective of determining a scattering function that is related to the acoustical properties of the region of interest (ROI) from the disturbed waves measured by transducers outside the ROI. Since the method is iterative, we use Born approximation for the first estimate of the scattering function. The main problem with the DBI is that the linear system of the inverse scattering equations is ill-posed. To deal with that, we use two different algorithms and compare the relative errors and execution times. The first one is Truncated Total Least Squares (TTLS). The second one is Regularized Total Least Squares method (RTLS-Newton) where the parameters for regularization were found by solving a nonlinear system with Newton method. We simulated the data for the DBI method in a way that leads to the overdetermined system. The advantage of RTLS-Newton is that the computation of singular value decomposition for a matrix is avoided, so it is faster than TTLS, but it still solves the similar minimization problem. For the exact scattering function we used Modified Shepp-Logan phantom. For finding the Born approximation, RTLS-Newton is 10 times faster than TTLS. In addition, the relative error in L2-norm is smaller using RTLS-Newton than TTLS after 10 iterations of the DBI method and it takes less time.

The distorted Born iterative (DBI) method is a powerful approach for solving the inverse scattering problem for ultrasound tomographic imaging. This method iteratively solves the inverse problem for the scattering function and the forward problem for the inhomogeneous Green’s function and the total field. Because of the ill-posed system from the inverse problem, regularization methods are needed to obtain a smooth solution. The three methods compared are truncated total least squares (TTLS), conjugate gradient for least squares (CGLS), and Tikhonov regularization. This paper uses numerical simulations to compare these three approaches to regularization in terms of both quality of image reconstruction and speed. Noise from both transmitters and receivers is very common in real applications, and is considered in stimulation as well. The solutions are evaluated by residual error of scattering function of region of interest(ROI), convergence of total field solutions in all iteration steps, and accuracy of estimated Green’s functions. By comparing the result of reconstruction quality as well as the computational cost of the three methods under different ultrasound frequency, we prove that TTLS method has the lowest error in solving strongly ill-posed problems. CGLS consumes the shortest computational time but its error is higher than TTLS, but lower than Tikhonov regularization.

In Ultrasound computer tomography (USCT) Synthetic aperture focusing technique (SAFT) is often applied for reflectivity image reconstruction. Phase aberration correction is essential to cope with the large sound speed differences in water and the different human tissues. In this paper we compare two approaches for phase aberration correction: a straight ray approximation using the Bresenham algorithm (B-SAFT) and a bent ray approximating using a multi-stencil Fast Marching Method (FMM-SAFT). The analysis is carried out with simulated point scatterers and simulated phantoms to measure the effect on the image resolution and contrast. The method is additionally applied to experimental data. B-SAFT degrades the image resolution and contrast in cases of large sound speed differences of objects and if the reconstructed point is close to a boundary where a change in impedance is present. FMM-SAFT is able to recover the image quality in these cases if the sound speed distribution is known accurately and with high resolution. If these requirements cannot be met, B-SAFT proved to be more robust.

In breast imaging by ultrasound CT, ultrasound is refracted owing to the difference of the sound speed between the breast and background water. The sound speed of a dense breast is higher than that of the water, while that of a fatty breast is lower than that of the water. In this study, we developed an oil-gel-based phantom for mimicking the wave refraction from the fatty breast to the dense breast. An oil gel was generated by adding SEBS (Styrene-Ethylene/Butylenes-Styrene, 10 wt%) to paraffin oil. The oil-gel-based phantom has a cylindrical shape and contains rod shaped inclusions which can be filled with salty water (3.5%). When temperature increases, the sound speed of water increases, while that of the oil gel decreases; the sound speeds of the oil gel were higher than those of the water at less than 20°C, while the sound speeds of the oil gel were lower than those of water at higher than 20°C. By controlling the temperature, the oil-gel-based phantom was able to simulate the refraction from the fatty breast (1476 [m/s]) to the dense breast (1559 [m/s]). For 43 days, the variation of the sound speed and attenuation of the oil gel in the reconstructed images were 0.7[m/s] and 0.03[dB/MHz/cm], respectively. This phantom with high temporal stability is suitable for multi-center distribution and may be used for standardization of data acquisition and image reconstruction across centers.

We are developing an ultrasound computed tomography (USCT) system for early breast-cancer screening. USCT has great advantages over mammographies because of its lack of X-ray exposure and compression pain. USCT can show both the reflection boundary (structure) distribution and the sound speed (hardness) distribution in a subject, which is estimated from the time-of-flight (TOF) information of transmitted ultrasound waves on the basis of an X-ray CT algorithm. Considering the nature of ultrasound waves, improving the image quality generally increases the calculation burden. To achieve both high-quality images and high throughput, we developed an iterative refraction calibration method. The measured TOF sinogram was iteratively calibrated by the difference between the fastest wave arrival time and the arrival time of the wave along the geometrically shortest path in a section. This method was applied to the data of a gel phantom and a dog’s tumor extirpated at Tokyo University of Agriculture and Technology, which was measured by a USCT prototype with a 10 cm-diameter ring array. As a result, we achieved a calculation speed seven times faster than that of a conventional bent-ray reconstruction with the same contrast as that of a sound-speed image.

Breast ultrasound tomography imaging (BUTI) is a new ultrasound imaging technique developed in recent years. In contrast to traditional ultrasound, BUTI uses a ring transducer to surround an object in a water tank and transmits the ultrasound echo from each element sequentially while receiving all the reflective and transmission signals from all elements. The tomography image is reconstructed using a similar reconstruction technique like x-ray computed tomography (X-CT) but much complicate due to the echo travelled along the curve instead of straight line like X ray. In this paper, with the objective of developing a breast ultrasound screening product, in vitro- and in vivo evaluation experiments were performed before proceeding to formal clinic trials. For the in vitro evaluation, a Breast Ultrasound Needle Biopsy Phantom from Supertech, IN, USA, was scanned by BUTIS (Breast ultrasound tomography imaging system) developed in HUST (Huazhong University of Science and Technology, Wuhan, China), MRI and traditional ultrasound scanner. Their image qualities were compared. In addition, the spatial resolution was estimated by using a nylon wire phantom. The results demonstrated that the spatial resolution of BUTIS is over 180 μm, which is almost 1 order higher than the traditional ultrasounds with the same frequency transducer. The in vivo evaluation was composed of a human arm and leg, the breast of a pregnant goat as well as human breasts from a female volunteer. The experimental results demonstrated that BUTIS can not only obtain exceptionally high contrast and high resolution images of soft tissue like the breast in vivo both for animal or human volunteer, but it can also be used to scan the subject with bones inside such as human arms and legs, which seems impossible for traditional ultrasounds. It illustrated that BUTIS will become a new efficient ultrasound imaging technique with wide potential applications in clinics.

Ultrasound Computer Tomography is an exciting new technology mostly aimed at breast cancer imaging. Due to the complex interaction of ultrasound with human tissue, the large amount of raw data, and the large volumes of interest, both image acquisition and image reconstruction are challenging. Following the idea of open science, the long term goal of the USCT reference database is establishing open and easy to use data and code interfaces and stimulating the exchange of available reconstruction algorithms and raw data sets of different USCT devices. The database was established with freely available and open licensed USCT data for comparison of reconstruction algorithms, and will be maintained and updated. Additionally, the feedback about data and system architecture of the scientists working on reconstruction methods will be published to help to drive further development of the various measurement setups.

In recent years, many research studies have been carried out on ultrasound computed tomography (USCT) for its application prospect in early detection of breast cancer. The synthetic aperture focusing technique (SAFT) widely used for the USCT image reconstruction is highly compute-intensive. Speeding up and optimizing the reconstruction algorithm on the graphics processing units (GPUs) have been highly applied to medical ultrasound imaging field. In this paper, we focus on accelerating the processing speed of SAFT with the GPU, considering its high parallel computation ability. The main computational features of SAFT are discussed to show the degree of computation parallelism. On the basis of the compute unified device architecture (CUDA) programming model and the Single Instruction Multiple Threads (SIMT) model, the optimization of SAFT parallel computation is performed. The proposed method was verified with the radio-frequency (RF) data of the breast phantom and the pig heart in vitro captured by the USCT system developed in the Medical Ultrasound Laboratory. Experimental results show that a 1024×1024 image reconstruction with a single NVIDIA GTX-1050 GPU could be 25 times faster than that with a 3.20-GHz Intel Core-i5 processor without image quality loss. The results also imply that with the increase of the image pixels, the acceleration effect is more notable.

Nearly 20% of the United States’ population is affected by varicose veins at some point in their lives. Currently, ultrasound (US) imaging is used as clinical imaging modality to help surgeons visualize and place the ablation catheter within the diseased vein accurately. However, US imaging of catheters has limitations such as angular dependency, especially for treating perforating veins. In addition, the laser ablation procedure is often performed without any real-time temperature monitoring, which could lead to non-sufficient thermal dose or heat induced thrombosis. We propose using combined US and Photoacoustic (PA) imaging for accurate localization of the laser ablation fibers within the veins. More specifically, we proposed coupling both ablation CW laser and pulsed laser into a single ablation catheter to perform both ablation and PA localization and of the catheter and thermometry. Our studies clearly indicated that while US imaging visualizes the body of the catheter, PA signal is only generated at the interface between the fiber tip and the tissue. As a result, PA images of the catheter indicate the location of the tip of the catheter only, without any possibility of error and mislocation. We initially investigated and compared the utility of US and PA in tracking fiber tip in a set of vessel-mimicking phantoms. Our results indicated artifact-free and accurate detection of the fiber tip using PA in contrast to US. Using the PA signal temperature dependency, we also demonstrated the utility of PA for real-time monitoring of temperature increase during laser ablation procedures.

Spontaneous preterm birth (sPTB) occurs in about one in ten infants born in the United States and is leading to almost 1 million neonatal deaths worldwide [1]. Diagnostic imaging of cervix is mostly limited to using ultrasound (US) to measure cervical length and has shown a low specificity to determine the risk of sPTB [2-7]. Quantitative functional imaging modalities such as elastography (EL) and photoacoustic (PA) imaging, are commonly used in conjunction with US imaging to provide additional information on tissue compositions and function. We propose using an endocavity probe to acquire US, PA, and EL information of the cervical tissue. Specifically, spectroscopic PA (sPA) is proposed to provide information on cervical tissue such as total hemoglobin (blood perfusion), tissue oxygenation level, and more importantly the collagen-to-water ratio in tissue. Shear wave elastography (SWE) measurements of cervical tissue indicates the correlation between cervical ripening and lower tissue elasticity. Our custom-designed imaging system consists of an endovaginal US transducer (ATL C9-5) capable of performing high frame rate US and acoustic radiation force shear wave imaging, and an optimized fiber-optic light delivery system’s for PA imaging. Our experimental results indicate the system’s ability to measure the presence of different concentrations of hemoglobin in tissue-mimicking phantoms as well as accurate measurement of hemoglobin oxygen saturation (SO2). In another set of experiments, we demonstrated the feasibility of monitoring collagen-to-water ratio in tissues through monitoring changes in sPA signature between 1100 and 1650 nm. Monitoring the variations of collagen in cervical tissue can help to predict sPTB.

Due to the high rate of gynecologic cancers among females, obtaining structural, functional, and molecular information from reproductive organs can potentially reveal diseases at their early stages of development . In this study, we aimed to develop a miniaturized phased-array ultrasound (US) and photoacoustic (PA) endoscope for potential imaging gynecologic cancer. The developed endoscope is built around a phased-array US transducer coupled to a fiber optic light delivery system. In particular, the proposed endoscope consists of a 64-element phased array US transducer, coupled to a light delivery system that includes six fiber optics. The probe dimensions allow for utilizing this device for imaging various types of gynecologic cancers in which the probe can become close to the pathologic tissue. Given the small imaging aperture, adaptive beamforming was developed to reconstruct co-registered US and PA images in 90-degrees sector scan format. The developed endoscope was tested in a set of tissue-mimicking phantom studies to determine its characteristics and its ability to form form co-registered volumetric US and PA images. In addition, spectroscopic PA (sPA) imaging of biocompatible, folate conjugated dye was tested to demonstrate the possibility of using the developed endoscope in imaging PA molecular contrast agents.

Acute lymphoblastic leukemia (ALL) interacts with bone marrow cells, creating hypoxic niches that stabilize HIF-1α and promote chemotherapeutic resistance. Spectrosocopic photoacoustic (PA) imaging is a label-free, noninvasive technique that probes the in vivo oxygenation status of hemoglobin, resulting in a measurement of oxygen saturation (SO₂) and providing a surrogate measure of tissue hypoxia. This work investigates multispectral PA imaging to assess the SO₂ in the femoral bone marrow in mice. Preliminary work was performed to assess the capability of imaging through bone, followed by an oxygen challenge to determine the magnitude of systemic SO₂ changes measurable in wild type mice. Furthermore, a pilot study to compare SO₂ measured in a murine model of ALL versus in healthy controls was performed to investigate a correlation between SO₂ changes in the femoral bone marrow and disease progression. Study results show that femoral SO₂ can be measured with a variation less than 10% in wild type mice over multiple time-points. In the oxygen challenge, a 10% difference in systemic SO₂ was observed between 100% and 21% O₂ inhalation conditions. Additionally, leukemic mice demonstrate significantly more variation in femoral SO₂ over the length of the femur than control mice at day 14 post-inoculation, indicating that femoral SO₂ is affected by leukemic disease progression. This work demonstrates the feasibility of observing changes in leukemic disease progression through the measurement of SO₂ with spectroscopic PA imaging, which could help develop a more complete understanding of the interplay of the local microenvironment with leukemogenesis.

Mammography is not sufficiently effective for women with dense breast tissue – women who are at much higher risk for developing breast cancer. Consequently, many breast cancers go undetected at their treatable stage. Improved cancer detection and characterization for women with dense breast tissue is urgently needed. Our clinical study has shown that ultrasound tomography (UST) is an emerging technique that moves beyond B-mode imaging by its through transmission capabilities. Transmission ultrasound provides additional tissue parameters such as sound speed, attenuation, and through-transmission rendered tissue stiffness information. For women with dense breasts, these parameters can be used to assist in detecting malignant masses within glandular or fatty tissue and differentiating malignant and benign masses. This paper focuses on the use of waveform ultrasound sound speed imaging and tissue stiffness information generated using through-transmission data to characterize different breast tissues and breast masses. In-vivo examples will be given to assess its effectiveness.

Ultrasound tomography (UST) is an emerging breast imaging modality that produces quantitative volumetric measurements of breast density without using ionizing radiation. Waveform reconstructions of UST sound speed images produce higher resolution maps of density distributions and has been shown to better separate dense tissue from non-dense tissue than ray based reconstructions. Volpara produces automated measures of volumetric mammographic density. Women who underwent both a UST scan and had a Volpara reading of their mammographic breast density had their density measures compared. Waveform sound speed images were reconstructed from the UST raw data and these images were then separated into regions of dense and non-dense tissue using a k-means clustering algorithm. This allowed for quantitative volumetric measures of average breast density along with subregion density measures. After preliminary analysis, correlations between the UST density measures and Volpara density measures were strong. In particular, the waveform density measures showed slightly stronger correlations with Volpara than the previous ray-based reconstructions, especially for direct measures of dense tissue. Further analysis is still required but this potentially indicates that waveform sound speed images are able to more clearly separate dense and non-dense regions of breast tissue.

Both mammography and standard ultrasound (US) rely upon subjective criteria within the breast imaging reporting and data system (BI-RADS) to provide more uniform interpretation outcomes, as well as differentiation and risk stratification of associated abnormalities. In addition, the technical performance and professional interpretation of both tests suffer from machine and operator dependence. Breast MR has become the new gold standard for screening of high-risk women but has cost and access limitations in extending screening to the entire population. We have been developing a new technique for breast imaging that is based on ultrasound tomography which quantifies tissue characteristics while also producing 3-D images of breast anatomy. Results are presented from clinical studies that utilize this method. Informed consent was obtained from all patients, prospectively recruited in an IRB-approved protocol following HIPAA guidelines. Images were produced by tomographic algorithms for reflection, sound speed and attenuation. All images were reviewed by a board-certified radiologist who has more than 20 years of experience in breast imaging and US-technology development. In the first phase of the study, UST images were compared to multi-modal imaging to determine the appearance of lesions and breast parenchyma. In the second phase of the study, correlative comparisons with MR breast imaging were used to establish basic operational capabilities of the UST system including the identification and characterization of parenchymal patterns. Our study demonstrated a high degree of correlation of breast tissue structures relative to fat subtracted contrast enhanced MRI. With a scan duration of ~ 1-3 minutes, no significant motion artifacts were observed.

Arteriosclerosis, increasing future risks of cardiovascular events even in its early stages, occurs in the vessel walls where stimulation stress, known as wall shear stress (WSS), is constantly lower than 0.4 Pa (referred to as low-WSS vessels). For early arteriosclerosis detection, we previously proposed a WSS-measurement method and detected low-WSS vessels by comparing the threshold value with the WSS calculated at a given moment when the mean flow-velocity maximized, presuming that the WSS maximized simultaneously at all measurement points. However, in reality, the moments were different between the upstream and downstream of blood flows because of the pulse wave propagation, and this difference resulted in false identification of low-WSS vessels. The objective of this study is to precisely identify low-WSS vessels by detecting the maximum-WSS during heartbeat cycles for each measurement point. We propose a method for identifying low-WSS vessels by calculating the WSS distributions in every frame and comparing the maximum-WSS with the threshold value for each measurement point. To evaluate the method, we compared it with the conventional one while identifying low-WSS vessels in a carotid-artery phantom. The precision of the classifications was assessed by the agreement rate between the echographic method and the ground truth, which was classified by the maximum-WSS measured by particle image velocimetry (PIV). The results revealed that the classification by our method agreed with that by the PIV in 84% of cases, and that by the conventional method agreed in 64%. In conclusion, our method increases the precision of low-WSS vessel identification.

Publisher’s Note: This paper, originally published on 4/13/2018, was withdrawn per author request.

In this work, we describe a method for converting fat-water-separated magnetic resonance imaging (MRI) volumes to acoustic maps for ultrasound simulations. An acoustic map is a mapping of acoustic imaging parameters such as speed of sound and density to grid points in the ultrasound simulations. Tissues are segmented into five primary classes of tissue in the human abdominal wall (skin, fat, muscle, connective tissue, and non-tissue). This segmentation is achieved using an unsupervised machine learning algorithm, called soft k-means clustering, on a multi-scale feature representation of the MRI volumes. We describe an automated method for utilizing soft k-means weights to produce an acoustic map that achieves approximately 90% agreement with manual segmentation. Two-dimensional (2D) and three-dimensional (3D) nonlinear ultrasound simulations are conducted, demonstrating the utility of realistic 3D maps over previously-available 2D acoustic maps.

Attenuation imaging using spectral techniques such as the spectral log difference (SLD) method suffers from a severe trade-off between spatial resolution and estimation variance. Recently, the regularized spectral log difference (RSLD) method was proposed as a technique that extends such trade-off by incorporating spatial priors (i.e., total variation) in the inversion process. However, the reduction of the variance of attenuation images could also be accomplished by post-processing of the attenuation maps using noise reduction techniques. The main goal of this study is to determine which strategy (i.e., noise handling during or after the attenuation image reconstruction) provides attenuation maps of better quality, both with synthetic data and experimental data obtained from calibrated physical phantoms. The results suggest that the noise rejection mechanism of RSLD significantly outperforms post-processing SLD images by filtering, nearly doubling the contrast-to-noise ratio for comparable values of estimation bias.

In a pilot study, radiofrequency backscatter data was collected in the paravertebral (PV) spaces of 4 healthy individuals. Using the associated gray scale ultrasound and Doppler data as guidance, regions-of-interest (ROIs) were chosen to represent five tissue types found in and around the PV space – rib shadow, pleura, superior costotransverse ligament, intercostal vessel (artery or vein), and the PV space away from the vessel. ROI sizes of 1.0 mm, 1.5 mm, and 2.0 mm square were examined for auto-regressive (AR) orders of 10, 20, 30, and 40 and bandwidths of 3dB, 6dB, 20dB. Spectral estimations were performed for each ROI size, AR order, and bandwidth over the A-lines of the ultrasound radiofrequency data. The spectra were averaged and normalized using data collected from a tissue phantom. Eight spectral parameters – Y-intercept, slope, and mid-band fit of the regression line, maximum dB of the spectra, frequency at maximum dB, minimum dB of the spectra, frequency at minimum dB, and integrated backscatter were calculated for each spectral estimate and used to create ensembles of bagged tree classifiers. An ROI size of 2.0 mm, bandwidth of 20 dB, and AR order 10 had the lowest out-of-bag error at 0.315, and averaged across all tissue types, an accuracy of 89.15%, sensitivity of 0.70, specificity of 0.93, and Youden’s Index (YI) of 0.62. These results show that the identification of the five tissues types in radiofrequency backscatter from intercostal ultrasound is feasible.

Dilatation of the cerebral ventricles is a common condition in preterm neonates with intraventricular hemorrhage (IVH). Post Hemorrhagic Ventricular Dilatation (PHVD) can lead to lifelong neurological impairment caused by ischemic injury due to increased intracranial pressure, and without treatment can lead to death. Previously, we have developed and validated a 3D ultrasound (US) system to monitor the progression of ventricle volumes (VV) in IVH patients; however, many patients with severe PHVD have ventricles so large they cannot be imaged within a single 3D US image. This limits the utility of atlas based segmentation algorithms required to measure VV as parts of the ventricles are in separate 3D US images, and thus, an already challenging segmentation becomes increasingly difficult to solve. Without a more automated segmentation, the clinical utility of 3D US ventricle volumes cannot be fully realized due to the large number of images and patients required to validate the technique in a clinical trials. Here, we describe the initial results of an automated ‘stitching’ algorithm used to register and combine multiple 3D US images of the ventricles of patients with PHVD. Our registration results show that we were able to register these images with an average target registration error (TRE) of 4.25±1.95 mm.

Despite Ultrasound Tomography (USCT) not being ionizing and having low cost, it faces several challenges due to the physical nature of ultrasound interaction and propagation on the medium. One common strategy of diverse USCT-algorithms is to initialize with a-priori anatomical information of the region to be reconstructed, but the effect of this initialization is not clear. To contribute in this topic, our work presents, on simulated medium, a study of how the modification in this initialization (value and area) affects the USCT convergence and the generated image quality. In this study, the following were used: Distorted Born Iterative Method (DBIM) as the USCT-algorithm; a Matlab Toolbox (k-wave) for the forward problem; and the Algebraic Reconstruction Technique for the inverse problem. All simulated objects have equal density and attenuation, and the sound speed is varied between 1400 and 1680m/s. The initial attributes of Ob1, the largest object and with slowest velocity, are varied (c_ob1 ± [0; 40; 80; 120]m/s and Area_ob1 ± [0%; 15%; 25%]). In all simulations 64 transducers(100kHz) uniformly distributed around the medium were used. The results, measured by Relative-Residual-Error and the Normalized-Root-Mean-Square-Error show, for all investigated areas, that convergence is better when the initialization velocity is set with higher than expected speed. Additionally, the RRE is higher for augmented area than for decreased area. For reconstruction quality, the NRMSE is also higher for augmented initial area than for small area. The investigation, based on simulations, suggests that when the contour of cold object is inaccurate, it is preferable to use smaller area such as its internal contour as prior region.

Ultrasound is widely used for diagnosing cardiovascular diseases. However, estimates such as left ventricle volume currently require manual segmentation, which can be time consuming. In addition, cardiac ultrasound is often complicated by imaging artifacts such as shadowing and mirror images, making it difficult for simple intensity-based automated segmentation methods. In this work, we use convolutional neural networks (CNNs) to segment ultrasound images of rat hearts embedded in agar phantoms into four classes: background, myocardium, left ventricle cavity, and right ventricle cavity. We also explore how the inclusion of a single diseased heart changes the results in a small dataset. We found an average overall segmentation accuracy of 70.0% ± 7.3% when combining the healthy and diseased data, compared to 72.4% ± 6.6% for just the healthy hearts. This work suggests that including diseased hearts with healthy hearts in training data could improve segmentation results, while testing a diseased heart with a model trained on healthy hearts can produce accurate segmentation results for some classes but not others. More data are needed in order to improve the accuracy of the CNN based segmentation.

Despite the use of protective equipment, burns are a significant source of battlefield injury particularly for operators of military vehicles. Burn severity is classified by the depth of heat penetration which is dependent on skin thickness. Current ASTM values for skin thickness used in burn injury models are based on forearm estimates. However, variations in skin thickness with body location and posture may be critical to accurately estimate burn injury and develop thermal protective equipment. This study used ultrasound to quantify epidermis and dermis skin layer thicknesses at various locations and postures on a human body. Superficial ultrasound images of seventeen male military personnel were obtained using a 22MHz linear probe (LOGIQe, GE). Three images were taken at twelve different locations. Hand locations were scanned in a neutral posture as well as a clenched-fist posture akin to grasping a steering wheel while operating a military vehicle. Measurements of the epidermis and dermis were obtained at each location and mean results were taken. Measured values were compared to the ASTM standard using a one sample t-test. In general, measured epidermis and dermis layer thickness was significantly larger compared to the current standard. The effect of hand posture was determined using a two sample t-test. Dermis values significantly decreased with the clenched fist posture while the epidermis remained unchanged between the two postures. Obtaining in-vivo skin thicknesses across the body will allow for more accurate predictions of burn injury and more efficient thermal protective equipment.