This PDF file contains the front matter associated with SPIE Proceedings Volume 10137, including the Title Page, Copyright information, Table of Contents, Introduction (if any), and Conference Committee listing.

Hepatocellular carcinoma (HCC) is one of the most important leading causes of cancer-related deaths worldwide. In this study, we evaluated the efficacy of sorafenib on hepatocellular carcinoma through bioluminescence tomography (BLT) based on Micro-CT/BLT multi-modal system. Initially, the human hepatocellular carcinoma cell line HepG2-Red-FLuc, which was transfected with luciferase gene, was cultured. And then, the orthotopic liver tumor mouse model was established on 4~5 weeks old athymic male Balb/c nude mice by inoculating the HepG2-Red-FLuc cell suspension into the liver lobe under isoflurane anesthesia. 15~20 days after tumor cells implantation, the mice were divided into two groups including the sorafenib treatment group and the control group. The mice in the treatment group were treated with sorafenib with dosage of 62 mg/kg/day by oral gavage for continuous 14 days, and the mice in the control group were treated with sterile water at equal volume. The tumor growth and drug treatment efficacy were dynamically monitored through BLT. The results in this study showed that the growth of liver cancer can be dynamically monitored from very early stage, and also the sorafenib treatment efficacy can be reliably and objectively assessed using BLT imaging method. Our experimental result demonstrated sorafenib can inhibit the tumor growth effectively. BLT enabled the non-invasive and reliable assessment of anti-neoplastic drug efficacy on liver cancer.

Dynamic fluorescence molecular tomography (DFMT) is a promising method for the quantitative evaluation of the metabolic process of fluorescent agents in body. However, the resolution is limited due to the ill-posed nature of fluorescence molecular tomography (FMT) and the high absorption and scattering of the fluorescent light in biological tissues. In this paper, the resolution of DFMT is improved by multispectral excitation method. Firstly, the imaged object with varied fluorescent concentrations at different time points is excited by several excitation lights with different wavelengths, and the fluorescent images are collected. Secondly, the individual FMT images at different time points are respectively reconstructed, and independent component analysis (ICA) is employed to decompose the fluorescent targets. The independent components (ICs) and corresponding spectrum courses (SCs) which obtained from ICA represent the spatial structures and spectral variations of the fluorescent targets, respectively. Thirdly, the ICs and SCs are combined to quantitatively recover the concentrations of individual fluorescent targets. Finally, the metabolic parameters and DFMT images are obtained by fitting the FMT images of each fluorescent targets at different time points into a two compartment model. Numerical simulations are carried out to validate the feasibility of the proposed method. The results demonstrate that the resolution of DFMT is significantly improved. The metabolic curves can be correctly recovered even when the edge-edge-distance of the fluorescent targets is less than 0.1 cm.

X-ray luminescence computed tomography (XLCT) is a hybrid molecular imaging modality that uses high energy x-ray photons to excite nanophosphors (e.g. Europium doped Gadolinium Oxysulfide – GOS: Eu³⁺) emitting optical photons to be measured by a sensitive detector for image reconstruction. XLCT has potentials to combine both the merits of x-ray imaging (high spatial resolution) and optical imaging (high sensitivity), which makes XLCT an attractive imaging modality to image nanophosphor targets deeply embedded in turbid media. In this study, we have evaluated the sensitivity of XLCT with phantom experiments by scanning targets of different phosphor concentrations at different depths. Cylindrical phantoms embedded with a cylindrical target with varying concentrations of GOS: Eu³⁺ (27.6 mM, 2.76 mM, 276 μM, and 27.6 μM) were scanned inside our lab made XLCT imaging system for varying scanning depths (6, 11, 16, and 21 mm). We found that XLCT is capable of imaging targets of very low concentrations (27.6 μM or 0.01 mg/mL) at significant depths, such as 21 mm. Our results demonstrate that there is also little variation in the reconstructed target size for different imaging depths for XLCT. We have for the first time, compared the sensitivity of XLCT with that of traditional computed tomography (CT) for phosphor targets. We found that XLCT’s use of x-ray induced photons provides much higher measurement sensitivity and contrast compared to CT which provides image contrast solely based on x-ray attenuation.

Developing reconstruction methods for diffuse optical imaging requires accurate modeling of photon propagation, including boundary conditions arising due to refractive index mismatch as photons propagate from the tissue to air. For this purpose, we developed an analytical Neumann-series radiative transport equation (RTE)-based approach. Each Neumann series term models different scattering, absorption, and boundary-reflection events. The reflection is modeled using the Fresnel equation. We use this approach to design a gradient-descent-based analytical reconstruction algorithm for a three-dimensional (3D) setup of a diffuse optical imaging (DOI) system. The algorithm was implemented for a three-dimensional DOI system consisting of a laser source, cuboidal scattering medium (refractive index > 1), and a pixelated detector at one cuboid face. In simulation experiments, the refractive index of the scattering medium was varied to test the robustness of the reconstruction algorithm over a wide range of refractive index mismatches. The experiments were repeated over multiple noise realizations. Results showed that by using the proposed algorithm, the photon propagation was modeled more accurately. These results demonstrated the importance of modeling boundary conditions in the photon-propagation model.

X-ray luminescence computed tomography (XLCT) is an emerging hybrid imaging modality that can provide functional and anatomical images at the same time. Traditional narrow beam XLCT can achieve high spatial resolution as well as high sensitivity. However, by treating the CCD camera as a single pixel detector, this kind of scheme resembles the first generation of CT scanner which results in a long scanning time and a high radiation dose. Although cone beam or fan beam XLCT has the ability to mitigate this problem with an optical propagation model introduced, image quality is affected because the inverse problem is ill-conditioned. Much effort has been done to improve the image quality through hardware improvements or by developing new reconstruction techniques for XLCT. The objective of this work is to further enhance the already reconstructed image by introducing anatomical information through retrospective processing. The deblurring process used a spatially variant point spread function (PSF) model and a joint entropy based anatomical prior derived from a CT image acquired using the same XLCT system. A numerical experiment was conducted with a real mouse CT image from the Digimouse phantom used as the anatomical prior. The resultant images of bone and lung regions showed sharp edges and good consistency with the CT image. Activity error was reduced by 52.3% even for nanophosphor lesion size as small as 0.8mm.

Micogravity, as experienced during prolonged space flight, presents a problem for space exploration. Animal models, specifically tadpoles, with altered connections of the vestibular ear allow the examination of the effects of microgravity and can be quantitatively monitored through tadpole swimming behavior. We describe an image analysis framework for performing automated quantification of tadpole swimming behavior. Speckle reducing anisotropic diffusion is used to smooth tadpole image signals by diffusing noise while retaining edges. A narrow band level set approach is used for sharp tracking of the tadpole body. The use of level set method for interface tracking provides an inherent advantage of using level set based image segmentation algorithm (active contouring). Active contour segmentation is followed by two-dimensional skeletonization, which allows the automated quantification of tadpole deflection angles, and subsequently tadpole escape (or C-start) response times. Evaluation of the image analysis methodology was performed by comparing the automated quantifications of deflection angles to manual assessments (obtained using a standard grading scheme), and produced a high correlation (r² = 0.99) indicating high reliability and accuracy of the proposed method. The methods presented form an important element of objective quantification of the escape response of the tadpole vestibular system to mechanical and biochemical manipulations, and can ultimately contribute to a better understanding of the effects of altered gravity perception on humans.

Optical coherence tomography (OCT) has been widely applied in the examination and diagnosis of corneal diseases, but the information directly achieved from the OCT images by manual inspection is limited. We propose an automatic processing method to assist ophthalmologists in locating the boundaries in corneal OCT images and analyzing the recovery of corneal wounds after treatment from longitudinal OCT images. It includes the following steps: preprocessing, epithelium and endothelium boundary segmentation and correction, wound detection, corneal boundary fitting and wound analysis. The method was tested on a data set with longitudinal corneal OCT images from 20 subjects. Each subject has five images acquired after corneal operation over a period of time. The segmentation and classification accuracy of the proposed algorithm is high and can be used for analyzing wound recovery after corneal surgery.

According to statistics, lung disease is among the leading causes of death worldwide. As such, many research groups are developing powerful tools for understanding, diagnosis and treatment of various lung diseases. Recently, biomechanical modeling has emerged as an effective tool for better understanding of human physiology, disease diagnosis and computer assisted medical intervention. Mechanical properties of lung tissue are important requirements for methods developed for lung disease diagnosis and medical intervention. As such, the main objective of this study is to develop an effective tool for estimating the mechanical properties of normal and pathological lung parenchyma tissue based on its microstructure. For this purpose, a micromechanical model of the lung tissue was developed using finite element (FE) method, and the model was demonstrated to have application in estimating the mechanical properties of lung alveolar wall. The proposed model was developed by assembling truncated octahedron tissue units resembling the alveoli. A compression test was simulated using finite element method on the created geometry and the hyper-elastic parameters of the alveoli wall were calculated using reported alveolar wall stress-strain data and an inverse optimization framework. Preliminary results indicate that the proposed model can be potentially used to reconstruct microstructural images of lung tissue using macro-scale tissue response for normal and different pathological conditions. Such images can be used for effective diagnosis of lung diseases such as Chronic Obstructive Pulmonary Disease (COPD).

Objectives: Our aim was to develop a clinically-practical and physiologically-relevant approach for regional structure-function measurements of the lung using Fourier decomposition of free-breathing pulmonary magnetic resonance imaging (FDMRI). Methods: Ten patients with chronic obstructive pulmonary disease provided written informed consent to a study protocol approved by Health Canada and completed pulmonary function tests, 1H/hyperpolarized noble gas and free-breathing pulmonary magnetic resonance imaging (MRI) during a single 2-hour visit. Free-breathing ¹H MRI was simultaneously segmented using a multi-region coupled continuous max-flow approach by exploring primal/dual analysis and convex optimization techniques. The segmented free-breathing 1H MRI lung was registered using deformable registration approach that was developed using dual and convex optimization methods to compensate for respiratory/cardiac motion. Fourier decomposition of the co-registered lung was used to generate pulmonary functional information that was quantified as ventilation-defect-percent (VDP). The pipeline was implemented on a GPU for speed-up. Lung segmentation accuracy was measured by comparing algorithm and manual lung masks using Dice-similarity-coefficient (DSC). FD-VDP was compared to ³He-VDP using Pearson correlation coefficient and Bland-Altman analysis. The reproducibility of our algorithm was measured using coefficient of variation (CoV) and intraclass correlation coefficient (ICC) for DSC and FD-VDP. Results: The pipeline yielded a whole lung DSC of 95.7±1.7% and FD-VDP that were correlated with ³He-VDP (r = 0.81, p = 0.004). CoV (ICC) were 0.4% (0.98) and 4.1% (0.98) for whole lung DSC and FD-VDP, respectively. The proposed approach requires ~45 min for parallel implementation with minimal user interaction. Conclusion: The proposed approach provides a clinically-practical pipeline to generate regional pulmonary structure-function measurements using free-breathing pulmonary ¹HMRI with promising potential for widespread clinical translation.

Segmentation is one of the most important steps in automated medical diagnosis applications, which affects the accuracy of the overall system. In this paper, we propose a semi-automated segmentation method for extracting lung lesions from thoracic PET/CT images by combining low level processing and active contour techniques. The lesions are first segmented in PET images which are first converted to standardised uptake values (SUVs). The segmented PET images then serve as an initial contour for subsequent active contour segmentation of corresponding CT images. To evaluate its accuracy, the Jaccard Index (JI) was used as a measure of the accuracy of the segmented lesion compared to alternative segmentations from the QIN lung CT segmentation challenge, which is possible by registering the whole body PET/CT images to the corresponding thoracic CT images. The results show that our proposed technique has acceptable accuracy in lung lesion segmentation with JI values of around 0.8, especially when considering the variability of the alternative segmentations.

The lower jawbone (or mandible), is due to its exposure to complex biomechanical forces the largest and strongest facial bone in humans. In this publication, an algorithmic evaluation of lower jawbone segmentation with a cellular automata algorithm called GrowCut is presented. For an evaluation, the algorithmic segmentation results were compared with slice-by-slice segmentations from two specialized physicians, which is considered to assess the given ground truth. As a result, pure manual slice-by-slice outlining took on average 39 minutes (minimum 35 minutes and maximum 46 minutes). This stands in strong contrast to an algorithmic segmentation which needed only about one minute for an initialization, hence needing just a fraction of the manual contouring time. At the same time, the algorithmic segmentations could achieve an acceptable Dice Similarity Score (DSC) of nearly ninety percent when compared to the ground truth slice-by-slice segmentations generated by the physicians. This stands in direct comparison to somewhat above ninety percent Dice Score between the two manual segmentations of the jawbones. In summary, this contribution shows that an algorithmic GrowCut segmentation can be an alternative to the very time consuming manual slice-by-slice outlining in the clinical practice.

Purpose: Anatomical metrics of the tibiofemoral joint support assessment of joint stability and surgical planning. We propose an automated, atlas-based algorithm to streamline the measurements in 3D images of the joint and reduce userdependence of the metrics arising from manual identification of the anatomical landmarks.

Methods: The method is initialized with coarse registrations of a set of atlas images to the fixed input image. The initial registrations are then refined separately for the tibia and femur and the best matching atlas is selected. Finally, the anatomical landmarks of the best matching atlas are transformed onto the input image by deforming a surface model of the atlas to fit the shape of the tibial plateau in the input image (a mesh-to-volume registration). We apply the method to weight-bearing volumetric images of the knee obtained from 23 subjects using an extremity cone-beam CT system. Results of the automated algorithm were compared to an expert radiologist for measurements of Static Alignment (SA), Medial Tibial Slope (MTS) and Lateral Tibial Slope (LTS).

Results: Intra-reader variability as high as ~10% for LTS and 7% for MTS (ratio of standard deviation to the mean in repeated measurements) was found for expert radiologist, illustrating the potential benefits of an automated approach in improving the precision of the metrics. The proposed method achieved excellent registration of the atlas mesh to the input volumes. The resulting automated measurements yielded high correlations with expert radiologist, as indicated by correlation coefficients of 0.72 for MTS, 0.8 for LTS, and 0.89 for SA.

Conclusions: The automated method for measurement of anatomical metrics of the tibiofemoral joint achieves high correlation with expert radiologist without the need for time consuming and error prone manual selection of landmarks.

Mechanotransduction – the process by which cells sense and respond to mechanical stimuli – is essential for several physiological processes including skeletal homeostasis. Mammalian cells are thought to be sensitive to different modes of mechanical stimuli, including vibration and fluid shear. To better understand the mechanisms underlying the early stages of mechanotransduction, we describe the development of devices for mechanostimulation (by vibration and fluid shear) of live cells that can be integrated with real-time optical microscopy. The integrated system can deliver up to 3 Pa of fluid shear simultaneous with high-frequency sinusoidal vibrations up to 1 g. Stimuli can be applied simultaneously or independently to cells during real-time microscopic imaging. A custom microfluidic chamber was prepared from polydimethylsiloxane on a glass-bottom cell culture dish. Fluid flow was applied with a syringe pump to induce shear stress. This device is compatible with a custom-designed motion control vibration system. A voice coil actuates the system that is suspended on linear air bushings. Accelerations produced by the system were monitored with an on-board accelerometer. Displacement was validated optically using particle tracking digital high-speed imaging (1200 frames per second). During operation at nominally 45 Hz and 0.3 g, displacements were observed to be within 3.56% of the expected value. MC3T3-E1 osteoblast like cells were seeded into the microfluidic device and loaded with the calcium sensitive fluorescent probe fura-2, then mounted onto the dual-modal mechanostimulation platform. Cells were then imaged and monitored for fluorescence emission. In summary, we have developed a system to deliver physiologically relevant vibrations and fluid shear to live cells during real-time imaging and photometry. Monitoring the behavior of live cells loaded with appropriate fluorescent probes will enable characterization of the signals activated during the initial stages of mechanotransduction.

This paper presents a novel computational method for micromechanical modeling of adipose tissue. The model can be regarded as the first step for developing an inversion based framework that uses adipose stiffness data obtained from elastography to determine its microstructural alterations. Such information can be used as biomarkers for diseases associated with adipose tissue microstructure alteration (e.g. adipose tissue fibrosis and inflammation in obesity). In contrast to previous studies, the presented model follows a multiphase structure which accounts for both solid and fluid components as well as their mechanical interaction. In the model, the lipid droplets and extracellular matrix were considered as the fluid and solid phase, respectively. As such, the fluid-structure interaction (FSI) problem was solved using finite element method. In order to gain insight into how microstructural characteristics influence the macro scale mechanical properties of the adipose tissue, a compression mechanical test was simulated using the FSI model and its results were fitted to corresponding experimental data. The simulation procedure was performed for adipocytes in healthy conditions while the stiffness of extracellular matrix in normal adipose tissue was found by varying it systematically within an optimization process until the simulation response agreed with experimental data. Results obtained in this study are encouraging and show the capability of the proposed model to capture adipose tissue macroscale mechanical behavior based on its microstructure under health and different pathological conditions.

Accurate measurement of long limb alignment is an essential stage of the pre-operative planning of realignment surgery. This alignment is quantified according to the hip-knee-ankle (HKA) angle of the mechanical axis of the lower extremity and is measured based on a full-length weight-bearing X-ray or standard computed radiography (CR) image of the patient in standing position. Due to the limited field-of-view of the traditionally employed digital X-ray imaging systems, several sector images are required to capture the posture of a standing individual. These sector images need to then be stitched" together to reconstruct the standing posture. To eliminate user-induced variability and time constraints associated with the traditional manual "stitching" protocol, we have created an image processing application to automate the stitching process, when there are no reliable external markers available in the images, by only relying on the most reliable anatomical content of the image. The application starts with a rough segmentation of the tibia and the sector images are then registered by evaluating the DICE coefficient between the edges of these corresponding bones along the medial edge. The identified translations are then used to register the original sector images into the standing panorama image. To test the robustness of our method, we randomly selected 40 datasets from a variant database consisting of nearly 100 patient X-ray images acquired for patient screening as part of a multi-site clinical trial. The resulting horizontal and vertical translation values from the automated registration were compared to the homologous translations recorded during the manual panorama generation conducted by a knowledgeable X-ray imaging technician. The mean and standard deviation of the differences for the horizontal translation parameters was -0:27±1:14 mm and 0:31±1:86 mm for the left and right tibia, respectively. The vertical translation differences for the left and right tibia were 1:05±5:24 mm and 1:32±4:77 mm, respectively. For these differences, the expert radiologist reported no difference in the hip-knee-ankle angular assessment.

Modern 3D printing technology can fabricate vascular phantoms based on an actual human patient with a high degree of precision facilitating a realistic simulation environment for an intervention. We present two experimental setups using 3D printed patient-specific neurovasculature to simulate different disease anatomies.

To simulate the human neurovasculature in the Circle of Willis, patient-based phantoms with aneurysms were 3D printed using a Objet Eden 260V printer. Anthropomorphic head phantoms and a human skull combined with acrylic plates simulated human head bone anatomy and x-ray attenuation. For dynamic studies the 3D printed phantom was connected to a pulsatile flow loop with the anthropomorphic phantom underneath. By combining different 3D printed phantoms and the anthropomorphic phantoms, different patient pathologies can be simulated. For static studies a 3D printed neurovascular phantom was embedded inside a human skull and used as a positional reference for treatment devices such as stents. To simulate tissue attenuation acrylic layers were added. Different combinations can simulate different patient treatment procedures.

The Complementary-Metal-Oxide-Semiconductor (CMOS) based High Resolution Fluoroscope (HRF) with 75μm pixels offers an advantage over the state-of-the-art 200 μm pixel Flat Panel Detector (FPD) due to higher Nyquist frequency and better DQE performance. Whether this advantage is clinically useful during an actual clinical neurovascular intervention can be addressed by qualitatively evaluating images from a cohort of various cases performed using both detectors. The above-mentioned method can offer a realistic substitute for an actual clinical procedure. Also a large cohort of cases can be generated and used for a HRF clinical utility determination study.

Blood flow rate is a critical parameter for diagnosing dialysis access function during fistulography where a flow rate of 600 ml/min in arteriovenous graft or 400-500 ml/min in arteriovenous fistula is considered the clinical threshold for fully functioning access. In this study, a flow rate computational model for calculating intra-access flow to evaluate dialysis access patency was developed and validated in an in vitro set up using digital subtraction angiography. Flow rates were computed by tracking the bolus through two regions of interest using cross correlation (XCOR) and mean arrival time (MAT) algorithms, and correlated versus an in-line transonic flow meter measurement. The mean difference (mean ± standard deviation) between XCOR and in-line flow measurements for in vitro setup at 3, 6, 7.5 and 10 frames/s was 118±63; 37±59; 31±31; and 46±57 ml/min respectively while for MAT method it was 86±56; 57±72; 35±85; and 19±129 ml/min respectively. The result of this investigation will be helpful for selecting candidate algorithms while blood flow computational tool is developed for clinical application.

Cerebral vessel segmentation is an important step in image analysis for brain function and brain disease studies. To extract all the cerebrovascular patterns, including arteries and capillaries, some filter-based methods are used to segment vessels. However, the design of accurate and robust vessel segmentation algorithms is still challenging, due to the variety and complexity of images, especially in cerebral blood vessel segmentation. In this work, we addressed a problem of automatic and robust segmentation of cerebral micro-vessels structures in cerebrovascular images acquired by light-sheet microscope for mouse. To segment micro-vessels in large-scale image data, we proposed a convolutional neural networks (CNNs) architecture trained by 1.58 million pixels with manual label. Three convolutional layers and one fully connected layer were used in the CNNs model. We extracted a patch of size 32x32 pixels in each acquired brain vessel image as training data set to feed into CNNs for classification. This network was trained to output the probability that the center pixel of input patch belongs to vessel structures. To build the CNNs architecture, a series of mouse brain vascular images acquired from a commercial light sheet fluorescence microscopy (LSFM) system were used for training the model. The experimental results demonstrated that our approach is a promising method for effectively segmenting micro-vessels structures in cerebrovascular images with vessel-dense, nonuniform gray-level and long-scale contrast regions.

Recent developments in MRI enable the acquisition of image sequences with high spatio-temporal resolution. Cardiac motion can be captured without gating and triggering. Image size and contrast relations differ from conventional cardiac MRI cine sequences requiring new adapted analysis methods. We suggest a novel segmentation approach utilizing contrast invariant polar scanning techniques. It has been tested with 20 datasets of arrhythmia patients. The results do not differ significantly more between automatic and manual segmentations than between observers. This indicates that the presented solution could enable clinical applications of real-time MRI for the examination of arrhythmic cardiac motion in the future.

In this paper we propose an estimation method of extracellular matrix directions of the heart. Myofiber are surrounded by the myocardial cell sheets whose directions have strong correspondence between heart failure. Estimation of the myocardial cell sheet directions is difficult since they are very thin. Therefore, we estimate the extracellular matrices which are touching to the sheets as if piled up. First, we perform a segmentation of the extracellular matrices by using the Hessian analysis. Each extracellular matrix region has sheet-like shape. We estimate the direction of each extracellular matrix region by the principal component analysis (PCA). In our experiments, mean inclination angles of two normal canine hearts were 50.6 and 46.2 degrees, while the angle of a failing canine heart was 57.4 degrees. This results well fit the anatomical knowledge that failing hearts tend to have vertical myocardical cell sheets.

Previous researches have shown that type 2 diabetes mellitus (T2DM) is associated with an increased risk of cognitive impairment. Early detection of brain abnormalities at the preclinical stage can be useful for developing preventive interventions to abate cognitive decline. We aimed to investigate the whole-brain resting-state functional connectivity (RSFC) patterns of T2DM patients between 90 regions of interest (ROIs) based on the RS-fMRI data, which can be used to test the feasibility of identifying T2DM patients with cognitive impairment from other T2DM patients. 74 patients were recruited in this study and multivariate pattern analysis was utilized to assess the prediction performance. Elastic net was firstly used to select the key features for prediction, and then a linear discrimination model was constructed. 23 RSFCs were selected and it achieved the performance with classification accuracy of 90.54% and areas under the receiver operating characteristic curve (AUC) of 0.944 using ten-fold cross-validation. The results provide strong evidence that functional interactions of brain regions undergo notable alterations between T2DM patients with cognitive impairment or not. By analyzing the RSFCs that were selected as key features, we found that most of them involved the frontal or temporal. We speculated that cognitive impairment in T2DM patients mainly impacted these two lobes. Overall, the present study indicated that RSFCs undergo notable alterations associated with the cognitive impairment in T2DM patients, and it is possible to predicted cognitive impairment early with RSFCs.

Craniosynostosis, the premature fusion of one or more cranial sutures, leads to grossly abnormal head shapes and pressure elevations within the brain caused by these deformities. To date, accepted treatments for craniosynostosis involve improving surgical skull shape aesthetics. However, the relationship between improved head shape and brain structure after surgery has not been yet established. Typically, clinical standard care involves the collection of diagnostic medical computed tomography (CT) imaging to evaluate the fused sutures and plan the surgical treatment. CT is known to provide very good reconstructions of the hard tissues in the skull but it fails to acquire good soft brain tissue contrast. This study intends to use magnetic resonance imaging to evaluate brain structure in a small dataset of sagittal craniosynostosis patients and thus quantify the effects of surgical intervention in overall brain structure. Very importantly, these effects are to be contrasted with normative shape, volume and brain structure databases. The work presented here wants to address gaps in clinical knowledge in craniosynostosis focusing on understanding the changes in brain volume and shape secondary to surgery, and compare those with normally developing children. This initial pilot study has the potential to add significant quality to the surgical care of a vulnerable patient population in whom we currently have limited understanding of brain developmental outcomes.

It’s widely known that HIV infection would cause white matter integrity impairments. Nevertheless, it is still unclear that how the white matter anatomical structural connections are affected by HIV infection. In the current study, we employed a multivariate pattern analysis to explore the HIV-related white matter connections alterations. Forty antiretroviraltherapy- naïve HIV patients and thirty healthy controls were enrolled. Firstly, an Automatic Anatomical Label (AAL) atlas based white matter structural network, a 90 × 90 FA-weighted matrix, was constructed for each subject. Then, the white matter connections deprived from the structural network were entered into a lasso-logistic regression model to perform HIV-control group classification. Using leave one out cross validation, a classification accuracy (ACC) of 90% (P=0.002) and areas under the receiver operating characteristic curve (AUC) of 0.96 was obtained by the classification model. This result indicated that the white matter anatomical structural connections contributed greatly to HIV-control group classification, providing solid evidence that the white matter connections were affected by HIV infection. Specially, 11 white matter connections were selected in the classification model, mainly crossing the regions of frontal lobe, Cingulum, Hippocampus, and Thalamus, which were reported to be damaged in previous HIV studies. This might suggest that the white matter connections adjacent to the HIV-related impaired regions were prone to be damaged.

The triple network model provides a common framework for understanding affective and neurocognitive dysfunctions across multiple disorders, including central executive network (CEN), default mode network (DMN), and salience network (SN). Considering the effect of traumatic experience on post-traumatic stress disorder (PTSD), this study aims to explore the alteration of triple network connectivity in a specific PTSD induced by a single prolonged trauma exposure. With arterial spin labeling sequence, three networks were identified using independent component analysis in 10 PTSD patients and 10 healthy survivors, who experienced the same coal mining flood disaster. In PTSD patients, decreased connectivity was identified in left middle frontal gyrus of CEN, left precuneus and bilateral superior frontal gyrus of DMN, and right anterior insula of SN. The decreased connectivity in left middle frontal gyrus was identified to associate with clinical severity. These results indicated the decreased triple network connectivity, which not only supported the proposal of the triple network model, but also prompted possible neurobiology mechanism of cognitive dysfunction for this kind of PTSD.

Cloud based Interactive Visualization of Tractography Brain Connectome (CIVILITY) is an interactive visualization tool of brain connectome in the cloud. This application submits tasks to remote computing grids were the CIVILITY-tractography pipeline is deployed. The application will list the running tasks for the user and once a task is completed the brain connectome is visualized using Hierarchical Edge Bundling. The analysis pipeline uses FSL tools (bedpostx and probtrackx2) to generate a triangular matrix indicating the connectivity strength between different regions in the brain. This work is motivated by medical applications in which expensive computational tasks such as brain connectivity is needed and to provide a state of the art visualization tool of Brain Connectome.

This work does not contribute any novelty with respect to the visualization methodology, is rather a new resource for the neuroimaging community. This work is submitted to the SPIE Biomedical Applications in Molecular, Structural, and Functional Imaging conference. The source code of this application is available in NITRC∗

Recent advancement in MRI established multi-parametric imaging for in vivo characterization of pathologic changes in brain cancer, which is expected to play a role in imaging biomarker development. Diffusion Tensor Imaging (DTI) is a prime example, which has been deployed for assessment of therapeutic response via analysis of apparent diffusion coefficient (ADC) / mean diffusivity (MD) values. They have been speculated to reflect apoptosis/necrosis. As newer medical imaging emerges, it is essential to verify that apparent abnormal features in imaging correlate with histopathology. Furthermore, the feasibility of imaging correlation with molecular profile should be explored in order to enhance the potential of biomedical imaging as a reliable biomarker. We focus on glioblastoma, which is an aggressive brain cancer. Despite the increased number of studies involving DTI in glioblastoma; however, little has been explored to bridge the gap between the molecular biomarkers and DTI data. Due to spatial heterogeneity in, MRI signals, pathologic change and protein expression, precise correlation is required between DTI, pathology and proteomics data in a histoanatomically identical manner. The challenge is obtaining an identical plane from in vivo imaging data that exactly matches with histopathology section. Thus, we propose to incorporate ex vivo tissue imaging to bridge between in vivo imaging data and histopathology. With ex vivo scan of removed tissue, it is feasible to use high-field 7T MRI scanner, which can achieve microscopic resolution. Once histology section showing the identical plane, it is feasible to correlate protein expression by a unique technology, “multiplex tissue immunoblotting”.

The ability of fast and single-neuron resolution imaging of neural activities enables light-sheet fluorescence microscopy (LSFM) as a powerful imaging technique in functional neural connection applications. The state-of-art LSFM imaging system can record the neuronal activities of entire brain for small animal, such as zebrafish or C. elegans at single-neuron resolution. However, the stimulated and spontaneous movements in animal brain result in inconsistent neuron positions during recording process. It is time consuming to register the acquired large-scale images with conventional method. In this work, we address the problem of fast registration of neural positions in stacks of LSFM images. This is necessary to register brain structures and activities. To achieve fast registration of neural activities, we present a rigid registration architecture by implementation of Graphics Processing Unit (GPU). In this approach, the image stacks were preprocessed on GPU by mean stretching to reduce the computation effort. The present image was registered to the previous image stack that considered as reference. A fast Fourier transform (FFT) algorithm was used for calculating the shift of the image stack. The calculations for image registration were performed in different threads while the preparation functionality was refactored and called only once by the master thread. We implemented our registration algorithm on NVIDIA Quadro K4200 GPU under Compute Unified Device Architecture (CUDA) programming environment. The experimental results showed that the registration computation can speed-up to 550ms for a full high-resolution brain image. Our approach also has potential to be used for other dynamic image registrations in biomedical applications.

Patients’ responses to a drug differ at the cellular level. Here, we present an image-based cell phenotypic feature quantification method for predicting the responses of patient-derived glioblastoma cells to a particular drug. We used high-content imaging to understand the features of patient-derived cancer cells. A 3D spheroid culture formation resembles the in vivo environment more closely than 2D adherent cultures do, and it allows for the observation of cellular aggregate characteristics. However, cell analysis at the individual level is more challenging. In this paper, we demonstrate image-based phenotypic screening of the nuclei of patient-derived cancer cells. We first stitched the images of each well of the 384-well plate with the same state. We then used intensity information to detect the colonies. The nuclear intensity and morphological characteristics were used for the segmentation of individual nuclei. Next, we calculated the position of each nucleus that is appeal of the spatial pattern of cells in the well environment. Finally, we compared the results obtained using 3D spheroid culture cells with those obtained using 2D adherent culture cells from the same patient being treated with the same drugs. This technique could be applied for image-based phenotypic screening of cells to determine the patient’s response to the drug.

The derivation of quantitative information from images to make quantitative radiology (QR) clinically practical continues to face a major image analysis hurdle because of image segmentation challenges. This paper presents a novel approach to disease quantification (DQ) via positron emission tomography/computed tomography (PET/CT) images that explores how to decouple DQ methods from explicit dependence on object segmentation through the use of only object recognition results to quantify disease burden. The concept of an object-dependent disease map is introduced to express disease severity without performing explicit delineation and partial volume correction of either objects or lesions. The parameters of the disease map are estimated from a set of training image data sets. The idea is illustrated on 20 lung lesions and 20 liver lesions derived from ¹⁸F-2-fluoro-2-deoxy-D-glucose (FDG)-PET/CT scans of patients with various types of cancers and also on 20 NEMA PET/CT phantom data sets. Our preliminary results show that, on phantom data sets, “disease burden” can be estimated to within 2% of known absolute true activity. Notwithstanding the difficulty in establishing true quantification on patient PET images, our results achieve 8% deviation from “true” estimates, with slightly larger deviations for small and diffuse lesions where establishing ground truth becomes really questionable, and smaller deviations for larger lesions where ground truth set up becomes more reliable. We are currently exploring extensions of the approach to include fully automated body-wide DQ, extensions to just CT or magnetic resonance imaging (MRI) alone, to PET/CT performed with radiotracers other than FDG, and other functional forms of disease maps.

This paper presents an end-to-end framework for automatically detecting and segmenting blood cells including normal red blood cells (RBCs), connected RBCs, abnormal RBCs (i.e. tear drop, burr cell, helmet, etc.) and white blood cells (WBCs). Our proposed system contains several components to solve different problems regarding RBCs and WBCs. We first design a novel blood cell color representation which is able to emphasize the RBCs and WBCs in separate channels. Template matching technique is then employed to individually detect RBCs and WBCs in our proposed representation. In order to automatically segment the RBCs and nuclei from WBCs, we develop an adaptive level set-based segmentation method which makes use of both local and global information. The detected and segmented RBCs, however, can be a single RBC, a connected RBC or an abnormal RBC. Therefore, we first separate and reconstruct RBCs from the connected RBCs by our suggested modified template matching. Shape matching by inner distance is later used to classify the abnormal RBCs from the normal RBCs. Our proposed method has been tested and evaluated on different images from ALL-IDB,¹⁰ WebPath,²⁴ UPMC,²³ Flicker datasets, and the one used by Mohamed et al.¹⁴ The precision and recall of RBCs detection are 98.43% and 94.99% respectively, whereas those of WBCs detection are 99.12% and 99.12%. The F-measure of our proposed WBCs segmentation gets up to 95.8%.

Binary morphology has innumerable applications in biomedical imaging, from segmentation to denoising. However, it suffers from inherently low precision. This is primarily because binary morphology is a binary technique, where each image voxel is all-or-nothing included or excluded. Many desirable structuring element shapes, especially circles or spheres, are poorly approximated on regular grids. Making things worse, common workflows involving multiple binary morphology iterations, such as opening or closing, compound this error. Also, small structuring elements often cannot be applied to 3D anisotropic image volumes. This work describes an extension to the theory of binary morphology, dubbed partial volume morphology or PVM, which allows the structuring element and/or image to hold fractional gray values to account for partial volumes. Partial volume morphology enables arbitrarily shaped structuring elements to be used, regardless of the underlying image resolution, with arbitrary precision. This technique also extends to 3D anisotropic volumes, allowing high precision morphological operations in anisotropic datasets heretofore impossible with binary morphology. This technique can be applied to a binary segmentation, where it provides subtle improvements and eliminates precision error in the intermediate steps of a multiple-operation workflow. Additionally, PVM is particularly suited for use on ‘soft’ segmentated data, where the partial volume contribution or probability at each point can be found. With segmentation and structuring elements both partial volume aware, partial volume morphology reaches its full potential as a high precision analytical tool. An open source reference implementation in Python, pvmpy, is provided.

Super resolution (SR) is to produce a higher resolution image from one or a sequence of low resolution images of a scene. It is essential in medical image analysis as a zooming of a specific area of interest is often required. This paper presents a new multi-frame super resolution (SR) method that is robust to both global and local motion. One of major challenges in multi-frame SR is concurrent global and local motion emergent in the sequence of low resolution images. It poses difficulties in aligning the low resolution images, resulting in artifacts or blurred pixels in the computed high resolution image. We solve the problem via a series of new methods. We first align the upscaled images from bicubic interpolation, and analyze the pixel distribution for the presence of local motion. If local motion is identified, we conduct the local image registration using dense SIFT features. Based on the local registration of images, we analyze pixel locations whose cross-frame variation is high and adaptively select subset of frame pixels in those locations. The adaptive selection of frame pixels is based on a clustering analysis of luminance values of pixels aligned at the same position, such that noise and motion biases are excluded. At the end, a median filter is applied for the selected pixels at each pixel location for super resolution image. We conduct experiments for multi-frame SR, where the proposed method delivers favorable results, especially better than state-of-the-art in dealing with concurrent local and global motions across frames.

Common vessel centerline extraction methods rely on the computation of a measure providing the likeness of the local appearance of the data to a curvilinear tube-like structure. The most popular techniques rely on empirically designed (hand crafted) measurements as the widely used Hessian vesselness, the recent oriented flux tubeness or filters (e.g. the Gaussian matched filter) that are developed to respond to local features, without exploiting any context information nor the rich structural information embedded in the data. At variance with the previously proposed methods, we propose a completely data-driven approach for learning a vesselness measure from expert-annotated dataset. For each data point (voxel or pixel), we extract the intensity values in a neighborhood region, and estimate the discriminative convolutional kernel yielding a positive response for vessel data and negative response for non-vessel data. The process is iterated within a boosting framework, providing a set of linear filters, whose combined response is the learned vesselness measure. We show the results of the general-use proposed method on the DRIVE retinal images dataset, comparing its performance against the hessian-based vesselness, oriented flux antisymmetry tubeness, and vesselness learned with a probabilistic boosting tree or with a regression tree. We demonstrate the superiority of our approach that yields a vessel detection accuracy of 0.95, with respect to 0.92 (hessian), 0.90 (oriented flux) and 0.85 (boosting tree).

Recent research has shown that deep learning methods have performed well on supervised machine learning, image classification tasks. The purpose of this study is to apply deep learning methods to classify brain images with different tumor types: meningioma, glioma, and pituitary. A dataset was publicly released containing 3,064 T1-weighted contrast enhanced MRI (CE-MRI) brain images from 233 patients with either meningioma, glioma, or pituitary tumors split across axial, coronal, or sagittal planes. This research focuses on the 989 axial images from 191 patients in order to avoid confusing the neural networks with three different planes containing the same diagnosis. Two types of neural networks were used in classification: fully connected and convolutional neural networks. Within these two categories, further tests were computed via the augmentation of the original 512×512 axial images. Training neural networks over the axial data has proven to be accurate in its classifications with an average five-fold cross validation of 91.43% on the best trained neural network. This result demonstrates that a more general method (i.e. deep learning) can outperform specialized methods that require image dilation and ring-forming subregions on tumors.

Extraction of blood vessels of the organ is a challenging task in the area of medical image processing. It is really difficult to get accurate vessel segmentation results even with manually labeling by human being. The difficulty of vessels segmentation is the complicated structure of blood vessels and its large variations that make them hard to recognize. In this paper, we present deep artificial neural network architecture to automatically segment the hepatic vessels from computed tomography (CT) image. We proposed novel deep neural network (DNN) architecture for vessel segmentation from a medical CT volume, which consists of three deep convolution neural networks to extract features from difference planes of CT data. The three networks have share features at the first convolution layer but will separately learn their own features in the second layer. All three networks will join again at the top layer. To validate effectiveness and efficiency of our proposed method, we conduct experiments on 12 CT volumes which training data are randomly generate from 5 CT volumes and 7 using for test. Our network can yield an average dice coefficient 0.830, while 3D deep convolution neural network can yield around 0.7 and multi-scale can yield only 0.6.

Voxel-based analysis is widely used in quantitative analysis of structural brain magnetic resonance imaging (MRI) and automated disease detection, such as Alzheimer's disease (AD). However, noise at the voxel level may cause low sensitivity to AD-induced structural abnormalities. This can be addressed with the use of a whole brain structural segmentation approach which greatly reduces the dimension of features (the number of voxels). In this paper, we propose an automatic AD diagnosis system that combines such whole brain segmen- tations with advanced machine learning methods. We used a multi-atlas segmentation technique to parcellate T1-weighted images into 54 distinct brain regions and extract their structural volumes to serve as the features for principal-component-analysis-based dimension reduction and support-vector-machine-based classification. The relationship between the number of retained principal components (PCs) and the diagnosis accuracy was systematically evaluated, in a leave-one-out fashion, based on 28 AD subjects and 23 age-matched healthy subjects. Our approach yielded pretty good classification results with 96.08% overall accuracy being achieved using the three foremost PCs. In addition, our approach yielded 96.43% specificity, 100% sensitivity, and 0.9891 area under the receiver operating characteristic curve.

Adult Spinal Deformity is a prominent medical issue with about 68% of the elderly population suffering from the disease.¹ Detailed biomechanical assessment is needed both in the presurgical planning of structural spinal deformity as well as in early functional biomechanical compensation in ambulatory spinal pain patients. When considering automation of this process, we have to look at photographic intervertebral disc detection technique as a way to produce a detailed model of the spine with appropriate measurements required to make efficient and accurate decisions on patient care. Deep convolutional neural network (CNN) has given remarkable results in object recognition tasks in recent years. However, massive training data, computational resources and long training time is needed for both training a deep network from scratch or finetuning a network. Using pretrained model as feature extractor has shown promising result for moderate sized medical data.2 However, most work have extracted features from the last layer and little has been explored in terms of the number of convolutional layers needed for best performance. In this work we trained Support Vector Machine (SVM) classifiers on different layers of CaffeNet³ features to show that deeper the better concept does not hold for task such as intervertebral disc detection. Furthermore, our experimental results show the potential of using very small training data, such as 15 annotated medical images in our experiment, to yield satisfactory classification performance with accuracy up to 97.2%.

This project assessed the effectiveness of using two different detectors to obtain dual-energy (DE) micro-CT data for the carrying out of material decomposition. A micro-CT coupled to either a complementary metal-oxide semiconductor (CMOS) or an electron multiplying CCD (EMCCD) detector was used to acquire image data of a 3D-printed phantom with channels filled with different materials. At any instance, materials such as iohexol contrast agent, water, and platinum were selected to make up the scanned object. DE micro-CT data was acquired, and slices of the scanned object were differentiated by material makeup. The success of the decomposition was assessed quantitatively through the computation of percentage normalized root-mean-square error (%NRMSE). Our results indicate a successful decomposition of iohexol for both detectors (%NRMSE values of 1.8 for EMCCD, 2.4 for CMOS), as well as platinum (%NRMSE value of 4.7). The CMOS detector performed material decomposition on air and water on average with 7 times more %NRMSE, possibly due to the decreased sensitivity of the CMOS system. Material decomposition showed the potential to differentiate between materials such as the iohexol and platinum, perhaps opening the door for its use in the neurovascular anatomical region. Work supported by Toshiba America Medical Systems, and partially supported by NIH grant 2R01EB002873.

We are developing ultrahigh spatial resolution (FWHM < 2 μm) high-brilliance x-ray source for rapid in vivo tomographic microvasculature imaging micro-CT angiography (μCTA) in small animal models using optimized contrast agent. It exploits Laser Wakefield Accelerator (LWFA) betatron x-ray emission phenomenon. Ultrashort high-intensity laser pulse interacting with a supersonic gas jet produces an ion cavity (“bubble”) in the plasma in the wake of the laser pulse. Electrons that are injected into this bubble gain energy, perform wiggler-like oscillations and generate burst of incoherent x-rays with characteristic duration time comparable to the laser pulse duration, continuous synchrotron-like spectral distribution that might extend to hundreds keV, very high brilliance, very small focal spot and highly directional emission in the cone-beam geometry. Such LWFA betatron x-ray source created in our lab produced 10²¹ –10²³ photons⋅ shot^-1⋅mrad^-2⋅mm^-2/0.1%bw with mean critical energy in the12–30 keV range. X-ray source size for a single laser shot was FWHM=1.7 μm; x-ray beam divergence 20–30 mrad, and effective focal spot size for multiple shots FWHM= ~2 μm. Projection images of simple phantoms and complex biological objects including insects and mice were obtained in single laser shots. We conclude that ultrahigh spatial resolution μCTA (FWHM ~2 μm) requiring thousands of projection images could be accomplished using LWFA betatron x-ray radiation in approximately 40 s with our existing 220 TW laser and sub seconds with next generation of ultrafast lasers and x-ray detectors, as opposed to several hours required using conventional microfocal x-ray tubes. Thus, sub second ultrahigh resolution in vivo microtomographic microvasculature imaging (in both absorption and phase contrast mode) in small animal models of cancer and vascular diseases will be feasible with LWFA betatron x-ray source.

Attaching alpha-emitting radionuclides to cancer-targeting agents increases the anti-tumor effects of targeted cancer therapies. The success of alpha therapy for treating bone metastases has increased interest in using targeted alpha therapy (TAT) to treat a broad spectrum of metastatic cancers. Estimating radiation doses to targeted tumors, including small (<250 μm) clusters of cancer cells, and to non-targeted tissues is critical in the pre-clinical development of TATs. However, accurate quantification of heterogeneous distributions of alpha-emitters in small metastases is not possible with existing pre-clinical in-vivo imaging systems. Ex-vivo digital autoradiography using a scintillator in combination with an image intensifier and a charged coupled device (CCD) has gained interest for pre-clinical ex-vivo alpha particle imaging. We present a simulation-based analysis of the fundamental spatial resolution limits of digital autoradiography systems. Spatial resolution was quantified in terms of the modulation transfer function (MTF) and Wagner's equivalent aperture. We modeled systems operating in either particle-counting (PC) or energy-integrating (EI) mode using a cascaded systems approach that accounts for: 1) the stopping power of alpha particles; 2) the distance alpha particles travel within the scintillator; 3) optical blur, and; 4) binning in detector elements. We applied our analysis to imaging of astatine-211 using an LYSO scintillator with thickness ranging from 10 μm to 20 μm. Our analysis demonstrates that when these systems are operated in particle-counting mode with a centroid-calculation algorithm, the effective apertures of ~35 μm can be achieved, which suggests that digital autoradiography may enable quantifying the uptake of alpha emitters in tumors consisting of a few cancer cells. Future work will investigate the image noise and energy-resolution properties of digital autoradiography systems.

Breast cancer is the most common cancer in women in the world. Breast-conserving surgery (BCS) is a standard surgical treatment for breast cancer with the key objective of removing breast tissue, maintaining a negative surgical margin and providing a good cosmetic outcome. A positive surgical margin, meaning the presence of cancerous tissues on the surface of the breast specimen after surgery, is associated with local recurrence after therapy. In this study, we investigate a new imaging modality based on Cerenkov luminescence imaging (CLI) for the purpose of detecting positive surgical margins during BCS. We develop Monte Carlo (MC) simulations using the Geant4 nuclear physics simulation toolbox to study the spectrum of photons emitted given 18F-FDG and breast tissue properties. The resulting simulation spectra show that the CLI signal contains information that may be used to estimate whether the cancerous cells are at a depth of less than 1 mm or greater than 1 mm given appropriate imaging system design and sensitivity. The simulation spectra also show that when the source is located within 1 mm of the surface, the tissue parameters are not relevant to the model as the spectra do not vary significantly. At larger depths, however, the spectral information varies significantly with breast optical parameters, having implications for further studies and system design. While promising, further studies are needed to quantify the CLI response to more accurately incorporate tissue specific parameters and patient specific anatomical details.

A comparative study of different bipolar current injection methods viz. Adjacent method, Cross method and Opposite method used in Electrical Impedance Tomography(EIT) is reported in this paper. Different electrode configurations are considered for current injection and voltage measurement to identify the one which yields better signal strength. Sensitivity of different current injection methods to inhomogeneity at different locations is examined. The effect of conductivity contrast on boundary voltages is studied by varying the conductivity of the inhomogeneity from 0.01mS/cm to 9.1mS/cm. Ill-posedness of the inverse problem is analyzed in terms of condition number for the aforementioned methods. Reconstruction of two closely placed inhomogeneities is done using Levenberg Marquardt method for different current injection methods to compare the resolution and corresponding error voltage is determined. Experiments are conducted using agar phantoms to validate some of the results obtained from the simulations. Based on the simulation studies and experimental validations, cross method is found to be the optimal current injection method to attain better data acquisition and image reconstruction.

Associating schizophrenia with disrupted functional connectivity is a central idea in schizophrenia research. However, identifying neuroimaging-based features that can serve as reliable “statistical biomarkers” of the disease remains a challenging open problem. We argue that generalization accuracy and stability of candidate features (“biomarkers”) must be used as additional criteria on top of standard significance tests in order to discover more robust biomarkers. Generalization accuracy refers to the utility of biomarkers for making predictions about individuals, for example discriminating between patients and controls, in novel datasets. Feature stability refers to the reproducibility of the candidate features across different datasets. Here, we extracted functional connectivity network features from fMRI data at both high-resolution (voxel-level) and a spatially down-sampled lower-resolution (“supervoxel” level). At the supervoxel level, we used whole-brain network links, while at the voxel level, due to the intractably large number of features, we sampled a subset of them. We compared statistical significance, stability and discriminative utility of both feature types in a multi-site fMRI dataset, composed of schizophrenia patients and healthy controls. For both feature types, a considerable fraction of features showed significant differences between the two groups. Also, both feature types were similarly stable across multiple data subsets. However, the whole-brain supervoxel functional connectivity features showed a higher cross-validation classification accuracy of 78.7% vs. 72.4% for the voxel-level features. Cross-site variability and heterogeneity in the patient samples in the multi-site FBIRN dataset made the task more challenging compared to single-site studies. The use of the above methodology in combination with the fully data-driven approach using the whole brain information have the potential to shed light on “biomarker discovery” in schizophrenia.

Clinically Isolated Syndrome (CIS) is often considered to be the first neurological episode associated with Multiple sclerosis (MS). At an early stage the inflammatory demyelination occurring in the CNS can manifest as a change in neuronal metabolism, with multiple asymptomatic white matter lesions detected in clinical MRI. Such damage may induce topological changes of brain networks, which can be captured by advanced functional MRI (fMRI) analysis techniques. We test this hypothesis by capturing the effective relationships of 90 brain regions, defined in the Automated Anatomic Labeling (AAL) atlas, using a large-scale Granger Causality (lsGC) framework. The resulting networks are then characterized using graph-theoretic measures that quantify various network topology properties at a global as well as at a local level. We study for differences in these properties in network graphs obtained for 18 subjects (10 male and 8 female, 9 with CIS and 9 healthy controls). Global network properties captured trending differences with modularity and clustering coefficient (p<0.1). Additionally, local network properties, such as local efficiency and the strength of connections, captured statistically significant (p<0.01) differences in some regions of the inferior frontal and parietal lobe. We conclude that multivariate analysis of fMRI time-series can reveal interesting information about changes occurring in the brain in early stages of MS.

Infection of the brain by the Human Immunodeficiency Virus (HIV) causes irreversible damage to the synaptic connections resulting in cognitive impairment. Patients with HIV infection, showing signs of impairment in multiple cognitive domains, as assessed by neuropsychological testing, are said to exhibit symptoms of HIV Associated Neurocognitive Disorder (HAND). In this study, we use resting-state functional MRI (fMRI) data to distinguish between healthy subjects and subjects with symptoms of HAND. To this end, we first establish a measure of interaction between pairs of regional time-series by quantifying their non-linear functional connectivity using Mutual Connectivity Analysis (MCA). Subsequently, we use a classifier to distinguish patterns of interaction between healthy and diseased individuals. Our results, quantified as the mean Area under the ROC curve (AUC) over 75 iterations, indicate that, using fMRI data, we can discriminate between the two cohorts well (AUC > 0.8). Specifically, we find that MCA (mean AUC = 0.89) based connectivity features perform significantly better (p < 0.05) when compared to cross-correlation (mean AUC = 0.82) at the classification task. A higher AUC using our approach suggests that such a nonlinear approach is better able to capture connectivity changes between brain regions and has potential for the development of novel neuro-imaging biomarkers.

The falx cerebri and tentorium cerebelli are dural structures found in the brain. Due to the roles both structures play in constraining brain motion, the falx and tentorium must be identified and included in finite element models of the head to accurately predict brain dynamics during injury events. To date there has been very little research work on automatically segmenting these two structures, which is understandable given that their 1) thin structure challenges the resolution limits of in vivo 3D imaging, and 2) contrast with respect to surrounding tissue is low in standard magnetic resonance imaging. An automatic segmentation algorithm to find the falx and tentorium which uses the results of a multi-atlas segmentation and cortical reconstruction algorithm is proposed. Gray matter labels are used to find the location of the falx and tentorium. The proposed algorithm is applied to five datasets with manual delineations. 3D visualizations of the final results are provided, and Hausdorff distance (HD) and mean surface distance (MSD) is calculated to quantify the accuracy of the proposed method. For the falx, the mean HD is 43.84 voxels and the mean MSD is 2.78 voxels, with the largest errors occurring at the frontal inferior falx boundary. For the tentorium, the mean HD is 14.50 voxels and mean MSD is 1.38 voxels.

Tourette syndrome (TS) is a childhood-onset neurobehavioral disorder. At present, the topological disruptions of the whole brain white matter (WM) structural networks remain poorly understood in TS children. Considering the unique position of the topologically central role of densely interconnected brain hubs, namely the rich club regions, therefore, we aimed to investigate whether the rich club regions and their related connections would be particularly vulnerable in early TS children. In our study, we used diffusion tractography and graph theoretical analyses to explore the rich club structures in 44 TS children and 48 healthy children. The structural networks of TS children exhibited significantly increased normalized rich club coefficient, suggesting that TS is characterized by increased structural integrity of this centrally embedded rich club backbone, potentially resulting in increased global communication capacity. In addition, TS children showed a reorganization of rich club regions, as well as significantly increased density and decreased number in feeder connections. Furthermore, the increased rich club coefficients and feeder connections density of TS children were significantly positively correlated to tic severity, indicating that TS may be characterized by a selective alteration of the structural connectivity of the rich club regions, tending to have higher bridging with non-rich club regions, which may increase the integration among tic-related brain circuits with more excitability but less inhibition for information exchanges between highly centered brain regions and peripheral areas. In all, our results suggest the disrupted rich club organization in early TS children and provide structural insights into the brain networks.

Multi-atlas label fusion is an accurate but time-consuming method of labeling the human brain. Using an intermediate image as a registration target can allow researchers to reduce time constraints by storing the deformations required of the atlas images. In this paper, we investigate the effect of registration through an intermediate template image on multi-atlas label fusion and propose a novel registration technique to counteract the negative effects of through-template registration. We show that overall computation time can be decreased dramatically with minimal impact on final label accuracy and time can be exchanged for improved results in a predictable manner. We see almost complete recovery of Dice similarity over a simple through-template registration using the corrected method and still maintain a 3-4 times speed increase. Further, we evaluate the effectiveness of this method on brains of patients with normal-pressure hydrocephalus, where abnormal brain shape presents labeling difficulties, specifically the ventricular labels. Our correction method creates substantially better ventricular labeling than traditional methods and maintains the speed increase seen in healthy subjects.

Noninvasive analysis of motion has important uses as qualitative markers for organ function and to validate biomechanical computer simulations relative to experimental observations. Tagged MRI is considered the gold standard for noninvasive tissue motion estimation in the heart, and this has inspired multiple studies focusing on other organs, including the brain under mild acceleration and the tongue during speech. As with other motion estimation approaches, using tagged MRI to measure 3D motion includes several preprocessing steps that affect the quality and accuracy of estimation. Benchmarks, or test suites, are datasets of known geometries and displacements that act as tools to tune tracking parameters or to compare different motion estimation approaches. Because motion estimation was originally developed to study the heart, existing test suites focus on cardiac motion. However, many fundamental differences exist between the heart and other organs, such that parameter tuning (or other optimization) with respect to a cardiac database may not be appropriate. Therefore, the objective of this research was to design and construct motion benchmarks by adopting an "image synthesis" test suite to study brain deformation due to mild rotational accelerations, and a benchmark to model motion of the tongue during speech. To obtain a realistic representation of mechanical behavior, kinematics were obtained from finite-element (FE) models. These results were combined with an approximation of the acquisition process of tagged MRI (including tag generation, slice thickness, and inconsistent motion repetition). To demonstrate an application of the presented methodology, the effect of motion inconsistency on synthetic measurements of head- brain rotation and deformation was evaluated. The results indicated that acquisition inconsistency is roughly proportional to head rotation estimation error. Furthermore, when evaluating non-rigid deformation, the results suggest that inconsistent motion can yield "ghost" shear strains, which are a function of slice acquisition viability as opposed to a true physical deformation.

Spectral-domain optical coherence tomography (SDOCT), in addition to its routine clinical use in the diagnosis of ocular diseases, has begun to fund increasing use in animal studies. Animal models are frequently used to study disease mechanisms as well as to test drug efficacy. In particular, SDOCT provides the ability to study animals longitudinally and non-invasively over long periods of time. However, the lack of anatomical landmarks makes the longitudinal scan acquisition prone to inconsistencies in orientation. Here, we propose a method for the automated registration of mouse SDOCT volumes. The method begins by accurately segmenting the blood vessels and the optic nerve head region in the scans using a pixel classification approach. The segmented vessel maps from follow-up scans were registered using an iterative closest point (ICP) algorithm to the baseline scan to allow for the accurate longitudinal tracking of thickness changes. Eighteen SDOCT volumes from a light damage model study were used to train a random forest utilized in the pixel classification step. The area under the curve (AUC) in a leave-one-out study for the retinal blood vessels and the optic nerve head (ONH) was found to be 0.93 and 0.98, respectively. The complete proposed framework, the retinal vasculature segmentation and the ICP registration, was applied to a secondary set of scans obtained from a light damage model. A qualitative assessment of the registration showed no registration failures.

The temporal focusing two-photon microscope (TFM) is developed to perform depth resolved wide field fluorescence imaging by capturing frames sequentially. However, due to strong nonignorable noises and diffraction rings surrounding particles, further researches are extremely formidable without a precise particle localization technique. In this paper, we developed a fully-automated scheme to locate particles positions with high noise tolerance. Our scheme includes the following procedures: noise reduction using a hybrid Kalman filter method, particle segmentation based on a multiscale kernel graph cuts global and local segmentation algorithm, and a kinematic estimation based particle tracking method. Both isolated and partial-overlapped particles can be accurately identified with removal of unrelated pixels. Based on our quantitative analysis, 96.22% isolated particles and 84.19% partial-overlapped particles were successfully detected.

Reliable retinal layer thickness measurements using optical coherence tomography (OCT) are important to track the subtle retinal changes in longitudinal studies. A total of 10 eyes (5 healthy subjects, 40±13 years old) were enrolled to study the inter-session repeatability and identify the pitfalls affecting the reliabilities. Each eye was scanned using spectral domain OCT (Spectralis SDOCT, Heidelberg Engineering) for 3 sessions with 30 seconds rest in between. The first and second sessions were scanned independently and the third one was scanned with the first one as the baseline visit. Each session consisted of a confocal scanning laser ophthalmoscopy (cSLO) image and 61 B-scans of 496×768 pixels. The first, second and third sessions were named as baseline, unregistered and registered sessions; respectively. Seven retinal layers labeled as RNFL, GCL+IPL, INL, OPL, ONL, IS and OS were segmented using a custom software (OCTRIMA3D) and measured in the ETDRS grid. Inter-session standard deviation (σ), coefficient of repeatability (CR) and coefficient of variations (COV) were calculated to quantify the repeatability. Paired t-test of COVs was used to compare the repeatability and the level of significance was set at 5%. We obtained that values of the CR <5 μm and COV of 5%, were revealed only in the outer layers. The values of COV were not significantly different (p<0.05) in the unregistered scanning session. Our results show that the rotations in the unregistered scanning sessions do not cause significant change in repeatability.

Diffuse optical tomography (DOT) is a promising noninvasive imaging modality capable of providing the functional characteristics (oxygen saturation and hemodynamic states) of thick biological tissue by quantifying the optical parameters. The parameter recovery problem in DOT is a nonlinear, ill-posed and ill conditioned inverse problem. The non-linear iterative methods are usually employed for image reconstruction in DOT by utilizing Tikhonov based regularization approach. These methods employ l2-norm based regularization where the constant regularization parameter is determined either empirically or generalized cross validation methods or L curve method. The reconstructed images look smoother or noisy depending on the chosen value of the regularization constant. Moreover the edges information of the inclusions appeared to be blurred in such constant regularization methods. In this study we proposed a method to retrieve and utilized a non-zero support (possible tumor location) to generate a spatially varying regularization map. The inclusions locations were determined by considering the imaging problem as a multiple measurements vector (MMV) problem. Based on the recovered inclusion positions spatially regularization map was generated to be used in non-linear image reconstruction framework. The results retrieved with such spatially varying priors shows slightly improved image reconstruction in terms of better contrast recovery, reduction in background noise and preservation of edge information of inclusions compared with the constant regularization approach.

Laboratory optical imaging systems for fluorescence and bioluminescence imaging have become widely available for research applications. These systems use an ultra-sensitive CCD camera to produce quantitative measurements of very low light intensity, detecting signals from small-animal models labeled with optical fluorophores or luminescent emitters. Commercially available systems typically provide quantitative measurements of light output, in units of radiance (photons s^-1 cm^-2 SR^-1) or intensity (photons s^-1 cm^-2). One limitation to current systems is that there is often no provision for routine quality assurance and performance evaluation. We describe such a quality assurance system, based on an LED-illuminated thin-film transistor (TFT) liquid-crystal display module. The light intensity is controlled by pulse-width modulation of the backlight, producing radiance values ranging from 1.8 x 10⁶ photons s^-1 cm^-2 SR^-1 to 4.2 x 10¹³ photons s^-1 cm^-2 SR^-1. The lowest light intensity values are produced by very short backlight pulses (i.e. approximately 10 μs), repeated every 300 s. This very low duty cycle is appropriate for laboratory optical imaging systems, which typically operate with long-duration exposures (up to 5 minutes). The low-intensity light source provides a stable, traceable radiance standard that can be used for routine quality assurance of laboratory optical imaging systems.

Three dimensional segmentation of macular optical coherence tomography (OCT) data of subjects with retinitis pigmentosa (RP) is a challenging problem due to the disappearance of the photoreceptor layers, which causes algorithms developed for segmentation of healthy data to perform poorly on RP patients. In this work, we present enhancements to a previously developed graph-based OCT segmentation pipeline to enable processing of RP data. The algorithm segments eight retinal layers in RP data by relaxing constraints on the thickness and smoothness of each layer learned from healthy data. Following from prior work, a random forest classifier is first trained on the RP data to estimate boundary probabilities, which are used by a graph search algorithm to find the optimal set of nine surfaces that fit the data. Due to the intensity disparity between normal layers of healthy controls and layers in various stages of degeneration in RP patients, an additional intensity normalization step is introduced. Leave-one-out validation on data acquired from nine subjects showed an average overall boundary error of 4.22 μm as compared to 6.02 μm using the original algorithm.

To aid in diagnosis and treatment of upper airway obstructive disorders (UAOD), we propose anatomic Optical Coherence Tomography (aOCT) for endoscopic imaging of the upper airway lumen with high speed and resolution. aOCT and CT scans are performed sequentially on in vivo swine to compare dynamic airway imaging data. The aOCT system is capable of capturing the dynamic deformation of the airway during respiration. This may lead to methods for airway elastography and aid in our understanding of dynamic collapse in UAOD.

Photoacoustic imaging is a molecular cum functional imaging modality based on differential optical absorption of the incident laser pulse by the endogeneous tissue chromophores. Several numerical simulations and finite element models have been developed in the past to describe and study Photoacoustic (PA) signal generation principles and study the effect of variation in PA parameters. Most of these simulation work concentrate on analyzing extracted 1D PA signals and each of them mostly describe only few of the building blocks of a Photoacoustic Tomography (PAT) imaging system. Papers describing simulation of the entire PAT system in one simulation platform, along with reconstruction is seemingly rare. This study attempts to describe how a commercially available Finite Element software (COMSOL^(R)), can serve as a single platform for simulating PAT that couples the electromagnetic, thermodynamic and acoustic pressure physics involved in PA phenomena. Further, an array of detector elements placed at the boundary in the FE model can provide acoustic pressure data that can be exported to Matlab^(R) to perform tomographic image reconstruction. The performance of two most commonly used image reconstruction techniques; namely, Filtered Backprojection (FBP) and Synthetic Aperture (SA) beamforming are compared. Results obtained showed that the lateral resolution obtained using FBP vs. SA largely depends on the aperture parameters. FBP reconstruction was able to provide a slightly better lateral resolution for smaller aperture while SA worked better for larger aperture. This interesting effect is currently being investigated further. Computationally FBP was faster, but it had artifacts along the spherical shell on which the data is projected.

Ultrasound has become highly popular to monitor atherosclerosis, by scanning the carotid artery. The screening involves measuring the thickness of the vessel wall and diameter of the lumen. An automatic segmentation of the vessel lumen, can enable the determination of lumen diameter. This paper presents a fully automatic segmentation algorithm, for robustly segmenting the vessel lumen in longitudinal B-mode ultrasound images. The automatic segmentation is performed using a combination of B-mode and power Doppler images. The proposed algorithm includes a series of preprocessing steps, and performs a vessel segmentation by use of the marker-controlled watershed transform. The ultrasound images used in the study were acquired using the bk3000 ultrasound scanner (BK Ultrasound, Herlev, Denmark) with two transducers "8L2 Linear" and "10L2w Wide Linear" (BK Ultrasound, Herlev, Denmark). The algorithm was evaluated empirically and applied to a dataset of in-vivo 1770 images recorded from 8 healthy subjects. The segmentation results were compared to manual delineation performed by two experienced users. The results showed a sensitivity and specificity of 90.41±11.2 % and 97.93±5.7% (mean±standard deviation), respectively. The amount of overlap of segmentation and manual segmentation, was measured by the Dice similarity coefficient, which was 91.25±11.6%. The empirical results demonstrated the feasibility of segmenting the vessel lumen in ultrasound scans using a fully automatic algorithm.

Dynamic chest radiography combined with computer analysis allows quantitative analyses on pulmonary function and rib motion. The accuracy of kinematic analysis is directly linked to diagnostic accuracy, and thus body motion compensation is a major concern. Our purpose in this study was to develop a computerized method to reduce a breathing body motion in dynamic chest radiographs. Dynamic chest radiographs of 56 patients were obtained using a dynamic flat-panel detector. The images were divided into a 1 cm-square and the squares on body counter were used to detect the body motion. Velocity vector was measured using cross-correlation method on the body counter and the body motion was then determined on the basis of the summation of motion vector. The body motion was then compensated by shifting the images based on the measured vector. By using our method, the body motion was accurately detected by the order of a few pixels in clinical cases, mean 82.5% in right and left directions. In addition, our method detected slight body motion which was not able to be identified by human observations. We confirmed our method effectively worked in kinetic analysis of rib motion. The present method would be useful for the reduction of a breathing body motion in dynamic chest radiography.

The density of lung tissue changes as demonstrated on imagery is dependent on the relative increases and decreases in the volume of air and lung vessels per unit volume of lung. Therefore, a time-series analysis of lung texture can be used to evaluate relative pulmonary function. This study was performed to assess a time-series analysis of lung texture on dynamic chest radiographs during respiration, and to demonstrate its usefulness in the diagnosis of pulmonary impairments. Sequential chest radiographs of 30 patients were obtained using a dynamic flat-panel detector (FPD; 100 kV, 0.2 mAs/pulse, 15 frames/s, SID = 2.0 m; Prototype, Konica Minolta). Imaging was performed during respiration, and 210 images were obtained over 14 seconds. Commercial bone suppression image-processing software (Clear Read Bone Suppression; Riverain Technologies, Miamisburg, Ohio, USA) was applied to the sequential chest radiographs to create corresponding bone suppression images. Average pixel values, standard deviation (SD), kurtosis, and skewness were calculated based on a density histogram analysis in lung regions. Regions of interest (ROIs) were manually located in the lungs, and the same ROIs were traced by the template matching technique during respiration. Average pixel value effectively differentiated regions with ventilatory defects and normal lung tissue. The average pixel values in normal areas changed dynamically in synchronization with the respiratory phase, whereas those in regions of ventilatory defects indicated reduced variations in pixel value. There were no significant differences between ventilatory defects and normal lung tissue in the other parameters. We confirmed that time-series analysis of lung texture was useful for the evaluation of pulmonary function in dynamic chest radiography during respiration. Pulmonary impairments were detected as reduced changes in pixel value. This technique is a simple, cost-effective diagnostic tool for the evaluation of regional pulmonary function.

This paper presents an accurate extraction method of spherical shaped membrane structures in the eyeball from MR volumes. In ophthalmic surgery, operation field is limited to a small region. Patient specific surgical simulation is useful to reduce complications. Understanding of tissue structure in the eyeball of a patient is required to achieve patient specific surgical simulations. Previous extraction methods of tissue structure in the eyeball use optical coherence tomography (OCT) images. Although OCT images have high resolution, imaging regions are limited to very small. Global structure extraction of the eyeball is difficult from OCT images. We propose an extraction method of spherical shaped membrane structures including the sclerotic coat, choroid, and retina. This method is applied to a T2 weighted MR volume of the head region. MR volume can capture tissue structure of whole eyeball. Because we use MR volumes, out method extracts whole membrane structures in the eyeball. We roughly extract membrane structures by applying a sheet structure enhancement filter. The rough extraction result includes parts of the membrane structures. Then, we apply the Hough transform to extract a sphere structure from the voxels set of the rough extraction result. The Hough transform finds a sphere structure from the rough extraction result. An experimental result using a T2 weighted MR volume of the head region showed that the proposed method can extract spherical shaped membrane structures accurately.

Osteoporosis is associated with an increased risk of low-trauma fractures. Segmentation of trabecular bone (TB) is essential to assess TB microstructure, which is a key determinant of bone strength and fracture risk. Here, we present a new method for TB segmentation for in vivo CT imaging. The method uses Hessian matrix-guided anisotropic diffusion to improve local separability of trabecular structures, followed by a new multi-scale morphological reconstruction algorithm for TB segmentation. High sensitivity (0.93), specificity (0.93), and accuracy (0.92) were observed for the new method based on regional manual thresholding on in vivo CT images. Mechanical tests have shown that TB segmentation using the new method improved the ability of derived TB spacing measure for predicting actual bone strength (R²=0.83).

Osteoporosis is associated with increased fracture risk. Recent advancement in the area of in vivo imaging allows segmentation of trabecular bone (TB) microstructures, which is a known key determinant of bone strength and fracture risk. An accurate biomechanical modelling of TB micro-architecture provides a comprehensive summary measure of bone strength and fracture risk. In this paper, a new direct TB biomechanical modelling method using nonlinear manifold-based volumetric reconstruction of trabecular network is presented. It is accomplished in two sequential modules. The first module reconstructs a nonlinear manifold-based volumetric representation of TB networks from three-dimensional digital images. Specifically, it starts with the fuzzy digital segmentation of a TB network, and computes its surface and curve skeletons. An individual trabecula is identified as a topological segment in the curve skeleton. Using geometric analysis, smoothing and optimization techniques, the algorithm generates smooth, curved, and continuous representations of individual trabeculae glued at their junctions. Also, the method generates a geometrically consistent TB volume at junctions. In the second module, a direct computational biomechanical stress-strain analysis is applied on the reconstructed TB volume to predict mechanical measures. The accuracy of the method was examined using micro-CT imaging of cadaveric distal tibia specimens (N = 12). A high linear correlation (r = 0.95) between TB volume computed using the new manifold-modelling algorithm and that directly derived from the voxel-based micro-CT images was observed. Young’s modulus (YM) was computed using direct mechanical analysis on the TB manifold-model over a cubical volume of interest (VOI), and its correlation with the YM, computed using micro-CT based conventional finite-element analysis over the same VOI, was examined. A moderate linear correlation (r = 0.77) was observed between the two YM measures. This preliminary results show the accuracy of the new nonlinear manifold modelling algorithm for TB, and demonstrate the feasibility of a new direct mechanical strain-strain analysis on a nonlinear manifold model of a highly complex biological structure.

Diverse cardiac conditions such as myocardial infarction and hypertension can lead to diastolic dysfunction as a prevalent cardiac condition. Diastolic dysfunctions can be diagnosed through different adverse mechanisms such as abnormal left ventricle (LV) relaxation, filling, and diastolic stiffness. This paper is geared towards evaluating diastolic stiffness and measuring the LV blood pressure non-invasively. Diastolic stiffness is an important parameter which can be exploited for more accurate diagnosis of diastolic dysfunction. For this purpose, a finite element (FE) LV mechanical model, which works based on a novel composite material model of the cardiac tissue, was utilized. Here, this model was tested for inversion-based applications where it was applied for estimating the cardiac tissue passive stiffness mechanical properties as well as diastolic LV blood pressure. To this end, the model was applied to simulate diastolic inflation of the human LV. The start-diastolic LV geometry was obtained from MR image data segmentation of a healthy human volunteer. The obtained LV geometry was discretized into a FE mesh before FE simulation was conducted. The LV tissue stiffness and diastolic LV blood pressure were adjusted through optimization to achieve the best match between the calculated LV geometry and the one obtained from imaging data. The performance of the LV mechanical simulations using the optimal values of tissue stiffness and blood pressure was validated by comparing the geometrical parameters of the dilated LV model as well as the stress and strain distributions through the LV model with available measurements reported on the LV dilation.

We propose a novel method for nonrigid registration of coronary arteries within frames of a fluoroscopic X-ray angiogram sequence with propagated deformation field. The aim is to remove the motion of coronary arteries in order to simplify further registration of the 3D vessel structure obtained from computed tomography angiography, with the x-ray sequence. The Proposed methodology comprises two stages: propagated adjacent pairwise nonrigid registration, and, sequence-wise fixed frame nonrigid registration. In the first stage, a propagated nonrigid transformation reduces the disparity search range for each frame sequentially. In the second stage, nonrigid registration is applied for all frames with a fixed target frame, thus generating a motion-aligned sequence. Experimental evaluation conducted on a set of 7 fluoroscopic angiograms resulted in reduced target registration error, compared to previous methods, showing the effectiveness of the proposed methodology.

Electrical Impedance Spectroscopy (EIS) has emerged as a non-invasive imaging modality to detect and quantify functional or electrical properties related to the suspicious tumors in cancer screening, diagnosis and prognosis assessment. A constraint on EIS systems is that the current excitation system suffers from the effects of stray capacitance having a major impact on the hardware subsystem as the EIS is an ill-posed inverse problem which depends on the noise level in EIS measured data and regularization parameter in the reconstruction algorithm. There is high complexity in the design of stable current sources, with stray capacitance reducing the output impedance and bandwidth of the system. To confront this, we have designed an EIS current source which eliminates the effect of stray capacitance and other impacts of the capacitance via a variable inductance. In this paper, we present a combination of operational CCII based on a generalized impedance converter (OCCII-GIC) with a current source. The aim of this study is to use the EIS system as a biomedical imaging technique, which is effective in the early detection of breast cancer. This article begins with the theoretical description of the EIS structure, current source topologies and proposes a current conveyor in application of a Gyrator to eliminate the current source limitations and its development followed by simulation and experimental results. We demonstrated that the new design could achieve a high output impedance over a 3MHz frequency bandwidth when compared to other types of GIC circuits combined with an improved Howland topology.

Most premature infants born at less than 30 weeks gestation are exposed to periods of mild intermittent hypoxia (IH) associated with apnea of prematurity and periodic breathing. In adults, IH associated with sleep apnea causes neurochemical and structural alterations in the brain. However, it is unknown whether IH in the premature infant leads to neurodevelopmental impairment. Quantification of biochemical markers that can precisely identify infants at risk of adverse neurodevelopmental outcome is essential. In vivo 1H magnetic resonance spectroscopy (¹H MRS) facilitates the quantification of metabolites from distinct regions of the developing brain. We report the changes in metabolite profiles in the brainstem and hippocampal regions of developing rat brains, resulting from exposure to IH. Rat pups were chosen for study because there is rapid postnatal hippocampal development that occurs during the first 4 weeks in the developing rat brain, which corresponds to the first 2–3 postnatal years of development in humans. The brainstem was examined because of our interest in respiratory control disorders in the newborn and because of brainstem gliosis described in infants who succumb to Sudden Infant Death Syndrome (SIDS). Metabolite profiles were compared between hypoxia treated rat pups (n = 9) and normoxic controls (n = 6). Metabolite profiles were acquired using the Point-RESolved spectroscopy (PRESS) MRS sequence and were quantified using the TARQUIN software. There was a significant difference in the concentrations of creatine (p = 0.031), total creatine (creatine + phosphocreatine) (p = 0.028), and total choline (p = 0.001) in the brainstem, and glycine (p = 0.031) in the hippocampal region. The changes are consistent with altered cellular bioenergetics and metabolism associated with hypoxic insult.

Multiple system atrophy (MSA) is a rare, non-curable, progressive neurodegenerative disorder that affects nervous system and movement, poses a considerable diagnostic challenge to medical researchers. Corpus callosum (CC) being the largest white matter structure in brain, enabling inter-hemispheric communication, quantification of callosal atrophy may provide vital information at the earliest possible stages. The main objective is to identify the differences in CC structure for this disease, based on quantitative analysis on the pattern of callosal atrophy. We report results of quantification of structural changes in regional anatomical thickness, area and length of CC between patient-groups with MSA with respect to healthy controls. The method utilizes isolating and parcellating the mid-sagittal CC into 100 segments along the length - measuring the width of each segment. It also measures areas within geometrically defined five callosal compartments of the well-known Witelson, and Hofer-Frahma schemes. For quantification, statistical tests are performed on these different callosal measurements. From the statistical analysis, it is concluded that compared to healthy controls, width is reduced drastically throughout CC for MSA group and as well as changes in area and length are also significant for MSA. The study is further extended to check if any significant difference in thickness is found between the two variations of MSA, Parkinsonian MSA and Cerebellar MSA group, using the same methodology. However area and length of this two sub-MSA group, no substantial difference is obtained. The study is performed on twenty subjects for each control and MSA group, who had T1-weighted MRI.

The objective of this research is to correlate the brown fat activity (BAT) with the metabolic activity of thyroid in patients with fibromyalgia syndrome (FS). For the development of the research, it was select a database containing 132 patients of a thermography clinic, male and female, with age over 18 years old; where the images selected were anteroposterior orthostasis top and anteroposterior in cervical extension. In the program Flir Report, it was possible to demarcate the region of the left and right interscapular and thyroid of each patient by getting the respective temperatures, in addition to view the hyper-radiation (“signal of mantle”) in the interscapular. Temperature was organized in table format, and statistical analysis was performed in the program Microcal Origin 6.0. As conclusion, it was found that the greater the metabolic activity of thyroid in patients with fibromyalgia, the greater will be the metabolic rate of brown fat (BAT).

Arterial spin labeling (ASL) provides a noninvasive measurement of cerebral blood flow (CBF). Due to relatively low spatial resolution, the accuracy of CBF measurement is affected by the partial volume (PV) effect. To obtain accurate CBF estimation, the contribution of each tissue type in the mixture is desirable. In general, this can be obtained according to the registration of ASL and structural image in current ASL studies. This approach can obtain probability of each tissue type inside each voxel, but it also introduces error, which include error of registration algorithm and imaging itself error in scanning of ASL and structural image. Therefore, estimation of mixture percentage directly from ASL data is greatly needed. Under the assumption that ASL signal followed the Gaussian distribution and each tissue type is independent, a maximum a posteriori expectation-maximization (MAP-EM) approach was formulated to estimate the contribution of each tissue type to the observed perfusion signal at each voxel. Considering the sensitivity of MAP-EM to the initialization, an approximately accurate initialization was obtain using 3D Fuzzy c-means method. Our preliminary results demonstrated that the GM and WM pattern across the perfusion image can be sufficiently visualized by the voxel-wise tissue mixtures, which may be promising for the diagnosis of various brain diseases.

Lung delineation via dynamic 4D thoracic magnetic resonance imaging (MRI) is necessary for quantitative image analysis for studying pediatric respiratory diseases such as thoracic insufficiency syndrome (TIS). This task is very challenging because of the often-extreme malformations of the thorax in TIS, lack of signal from bone and connective tissues resulting in inadequate image quality, abnormal thoracic dynamics, and the inability of the patients to cooperate with the protocol needed to get good quality images. We propose an interactive fuzzy connectedness approach as a potential practical solution to this difficult problem. Manual segmentation is too labor intensive especially due to the 4D nature of the data and can lead to low repeatability of the segmentation results. Registration-based approaches are somewhat inefficient and may produce inaccurate results due to accumulated registration errors and inadequate boundary information. The proposed approach works in a manner resembling the Iterative Livewire tool but uses iterative relative fuzzy connectedness (IRFC) as the delineation engine. Seeds needed by IRFC are set manually and are propagated from slice-to-slice, decreasing the needed human labor, and then a fuzzy connectedness map is automatically calculated almost instantaneously. If the segmentation is acceptable, the user selects “next” slice. Otherwise, the seeds are refined and the process continues. Although human interaction is needed, an advantage of the method is the high level of efficient user-control on the process and non-necessity to refine the results. Dynamic MRI sequences from 5 pediatric TIS patients involving 39 3D spatial volumes are used to evaluate the proposed approach. The method is compared to two other IRFC strategies with a higher level of automation. The proposed method yields an overall true positive and false positive volume fraction of 0.91 and 0.03, respectively, and Hausdorff boundary distance of 2 mm.

OCT skin images suffer from artifacts. Speckle is the main artifact while the other one is called background noise. In this study, we propose an algorithm that significantly reduces the background noise before applying a speckle reduction method. The results show that the diagnostically relevant features in the images become clearer after applying the proposed method. We used sub-pixel weighted median filtering for speckle reduction. The results from background noise removal in combination with the proposed speckle reduction algorithm show a significant improvement in the clarity of diagnostically relevant features in in-vivo human skin images.

Optical Coherence Tomography (OCT) has a great potential to elicit clinically useful information from tissues due to its high axial and transversal resolution. In practice, an OCT setup cannot reach to its theoretical resolution due to imperfections of its components, which make its images blurry. The blurriness is different alongside regions of image; thus, they cannot be modeled by a unique point spread function (PSF). In this paper, we investigate the use of solid phantoms to estimate the PSF of each sub-region of imaging system. We then utilize Lucy-Richardson, Hybr and total variation (TV) based iterative deconvolution methods for mitigating occurred spatially variant blurriness. It is shown that the TV based method will suppress the so-called speckle noise in OCT images better than the two other approaches. The performance of proposed algorithm is tested on various samples, including several skin tissues besides the test image blurred with synthetic PSF-map, demonstrating qualitatively and quantitatively the advantage of TV based deconvolution method using spatially-variant PSF for enhancing image quality.

Intravoxel incoherent motion (IVIM) magnetic resonance imaging is an emerging non-invasive technique that has been recently applied to quantify in vivo global placental perfusion. We propose a robust semi-automated method for segmenting the placenta into fetal and maternal compartments from IVIM data, using a multi-label image segmentation algorithm called ‘GrowCut’. Placental IVIM data were acquired on a 1.5T scanner from 16 healthy pregnant women between 21-37 gestational weeks. The voxel-wise perfusion fraction was then estimated after non-rigid image registration. The seed regions of the fetal and maternal compartments were determined using structural T2-weighted reference images, and improved progressively through an iterative process of the GrowCut algorithm to accurately encompass fetal and maternal compartments. We demonstrated that the placental perfusion fraction decreased in both fetal (-0.010/week) and maternal compartments (-0.013/week) while their relative difference (f_fetal-f_maternal) gradually increased with advancing gestational age (+0.003/week, p=0.065). Our preliminary results show that the proposed method was effective in distinguishing placental compartments using IVIM.

Functional Analysis of Diffusion Tensor Tract Statistics (FADTTS) is a toolbox for analysis of white matter (WM) fiber tracts. It allows associating diffusion properties along major WM bundles with a set of covariates of interest, such as age, diagnostic status and gender, and the structure of the variability of these WM tract properties.

However, to use this toolbox, a user must have an intermediate knowledge in scripting languages (MATLAB). FADTTSter was created to overcome this issue and make the statistical analysis accessible to any non-technical researcher. FADTTSter is actively being used by researchers at the University of North Carolina. FADTTSter guides non-technical users through a series of steps including quality control of subjects and fibers in order to setup the necessary parameters to run FADTTS.

Additionally, FADTTSter implements interactive charts for FADTTS’ outputs. This interactive chart enhances the researcher experience and facilitates the analysis of the results. FADTTSter’s motivation is to improve usability and provide a new analysis tool to the community that complements FADTTS.

Ultimately, by enabling FADTTS to a broader audience, FADTTSter seeks to accelerate hypothesis testing in neuroimaging studies involving heterogeneous clinical data and diffusion tensor imaging.

This work is submitted to the Biomedical Applications in Molecular, Structural, and Functional Imaging conference. The source code of this application is available in NITRC.

This study aimed to investigate the VX2 tumor growth in rabbit liver using T2-weighted imaging (T2WI) and dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI). Five New Zealand white (NZW) rabbits were implanted with VX2 cell suspension in liver. Afterwards, MRI was performed 7, 14, 21 and 28 days after tumor implantation. A 1.5T clinical MRI scanner was used to perform scans. After 3-plane localizer, T1 weighted imaging (T1WI), T2WI, and DCE-MRI using a three-dimensional gradient echo pulse sequence was performed. After 4 pre-contrast images were acquired, each rabbit was injected i.v. with 0.1 mmol/kg Dotarem. The total scan time after Dotarem administration was 30 minutes. All acquired images were analyzed using ImageJ software. Several regions of interest were selected from the rims of tumor, liver, and muscle. The enhancement ratio (ER) was calculated by dividing the MR signal after Dotarem injection to the MR signal before Dotarem injection. The maximum ER (ER_max) value of tumor for each rabbit was observed right after the Dotarem injection. The T2W MR signal intensities (T2W_SI) and the ER_max values obtained 7, 14, 21 and 28 days after tumor implantation were analyzed with a linear regression algorithm. Both T2W_SI and ER_max of tumors increased with time. The changes for T2W_SI and ER_max of tumors between 7 and 28 days after tumor implantation were 32.66% and 18.14%, respectively. T2W_SI is more sensitive than ER_max for monitoring the growth of VX2 tumor in a rabbit liver model.

Idiopathic pulmonary fibrosis (IPF) is a chronic fibrotic lung disease that develops in adults without any known cause. It is an interstitial lung disease in which the lung tissue becomes scarred and stiffens, ultimately leading to respiratory failure. This disease currently has no cure with limited treatment options, leading to an average survival time of 3-5 years after diagnosis. In this paper we employ a mathematical model simulating the lung parenchyma as hexagons with elastic forces applied to connecting vertices and opposing vertices. Using an image registration algorithm, we obtain trajectories of 4D-CT scans of a healthy patient, and one suffering from IPF. Converting the image trajectories into a hexagonal lattice, we fit the model parameters to match the respiratory motion seen for both patients across multiple image slices. We found the model could decently describe the healthy lung slices, with a minimum average error between corresponding vertices to be 1.66 mm. For the fibrotic lung slices the model was less accurate, maintaining a higher average error across all slices. Using the optimized parameters, we apply the forces predicted from the model using the image trajectory positions for each phase. Although the error is large, the spring constant values determined for the fibrotic patient were not as high as we expected, and more often than not determined to be lower than corresponding healthy lung slices. However, the net force distribution for some of those slices was still found to be greater than the healthy lung counterparts. Other modifications to the model, including additional directional components and which vertices were receiving with the limited sample size available, a clear distinction between the healthy and fibrotic lung cannot yet be made by this model.

Abdominal aortic aneurysm (AAA) is known as a leading cause of death in the United States. AAA is an abnormal dilation of the aorta, which usually occurs below the renal arteries and causes an expansion at least 1.5 times its normal diameter. It has been shown that biomechanical parameters of the aortic tissue coupled with a set of specific geometric parameters characterizing the vessel expansion, affect the risk of aneurysm rupture. Here, we developed a numerical framework that incorporates both biomechanical and geometrical factors to study the behavior of abdominal aortic aneurysm. Our workflow enables the extraction of the aneurysm geometry from both clinical quality, as well as low-resolution MR images. We used a two-parameter, hyper-elastic, isotropic, incompressible material to model the vessel tissue. Our numerical model was tested using both synthetic and mouse data and we evaluated the effects of the geometrical and biomechanical properties on the developed peak wall stress. In addition, we performed several parameter sensitivity studies to investigate the effect of different factors affecting the AAA and its behavior and rupture. Lastly, relationships between different geometrical and biomechanical parameters and peak wall stress were determined. These studies help us better understand vessel tissue response to various loading, geometry and biomechanics conditions, and we plan to further correlate these findings with the pathophysiological conditions from a patient population diagnosed with abdominal aortic aneurysms.

Optical Coherence Optical coherence tomography is a powerful high-resolution imaging method with a broad biomedical application. Nonetheless, OCT images suffer from a multiplicative artefacts so-called speckle, a result of coherent imaging of system. Digital filters become ubiquitous means for speckle reduction. Addressing the fact that there still a room for despeckling in OCT, we proposed an intelligent speckle reduction framework based on OCT tissue morphological, textural and optical features that through a trained network selects the winner filter in which adaptively suppress the speckle noise while preserve structural information of OCT signal. These parameters are calculated for different steps of the procedure to be used in designed Artificial Neural Network decider that select the best denoising technique for each segment of the image. Results of training shows the dominant filter is BM3D from the last category.

The conjunctiva is a densely vascularized tissue of the eye that provides an opportunity for imaging of human microcirculation. In the current study, automated fine structure analysis of conjunctival microvasculature images was performed to discriminate stages of diabetic retinopathy (DR). The study population consisted of one group of nondiabetic control subjects (NC) and 3 groups of diabetic subjects, with no clinical DR (NDR), non-proliferative DR (NPDR), or proliferative DR (PDR). Ordinary least square regression and Fisher linear discriminant analyses were performed to automatically discriminate images between group pairs of subjects. Human observers who were masked to the grouping of subjects performed image discrimination between group pairs. Over 80% and 70% of images of subjects with clinical and non-clinical DR were correctly discriminated by the automated method, respectively. The discrimination rates of the automated method were higher than human observers. The fine structure analysis of conjunctival microvasculature images provided discrimination of DR stages and can be potentially useful for DR screening and monitoring.

We consider the problem of identifying frames in a cardiac ultrasound video associated with left ventricular chamber end-systolic (ES, contraction) and end-diastolic (ED, expansion) phases of the cardiac cycle. Our procedure involves a simple application of non-negative matrix factorization (NMF) to a series of frames of a video from a single patient. Rank-2 NMF is performed to compute two end-members. The end members are shown to be close representations of the actual heart morphology at the end of each phase of the heart function. Moreover, the entire time series can be represented as a linear combination of these two end-member states thus providing a very low dimensional representation of the time dynamics of the heart. Unlike previous work, our methods do not require any electrocardiogram (ECG) information in order to select the end-diastolic frame. Results are presented for a data set of 99 patients including both healthy and diseased examples.

Convergent evidence has been collected to support that Alzheimer’s disease (AD) is associated with reduction in hippocampal volume based on anatomical magnetic resonance imaging (MRI) and impaired functional connectivity based on functional MRI. Radiomics texture analysis has been previously successfully used to identify MRI biomarkers of several diseases, including AD, mild cognitive impairment and multiple sclerosis. In this study, our goal was to determine if MRI hippocampal textures, including the intensity, shape, texture and wavelet features, could be served as an MRI biomarker of AD. For this purpose, the texture marker was trained and evaluated from MRI data of 48 AD and 39 normal samples. The result highlights the presence of hippocampal texture abnormalities in AD, and the possibility that texture may serve as a neuroimaging biomarker for AD.

Hyperspectral imaging (HSI) is an emerging imaging modality that can provide a noninvasive tool for cancer detection and image-guided surgery. HSI acquires high-resolution images at hundreds of spectral bands, providing big data to differentiating different types of tissue. We proposed a deep learning based method for the detection of head and neck cancer with hyperspectral images. Since the deep learning algorithm can learn the feature hierarchically, the learned features are more discriminative and concise than the handcrafted features. In this study, we adopt convolutional neural networks (CNN) to learn the deep feature of pixels for classifying each pixel into tumor or normal tissue. We evaluated our proposed classification method on the dataset containing hyperspectral images from 12 tumor-bearing mice. Experimental results show that our method achieved an average accuracy of 91.36%. The preliminary study demonstrated that our deep learning method can be applied to hyperspectral images for detecting head and neck tumors in animal models.

Osteoarthritis (OA) of temporomandibular joints (TMJ) occurs in about 40% of the patients who present TMJ disorders. Despite its prevalence, OA diagnosis and treatment remain controversial since there are no clear symptoms of the disease, especially in early stages. Quantitative tools based on 3D imaging of the TMJ condyle have the potential to help characterize TMJ OA changes. The goals of the tools proposed in this study are to ultimately develop robust imaging markers for diagnosis and assessment of treatment efficacy. This work proposes to identify differences among asymptomatic controls and different clinical phenotypes of TMJ OA by means of Statistical Shape Modeling (SSM), obtained via clinical expert consensus. From three different grouping schemes (with 3, 5 and 7 groups), our best results reveal that that the majority (74.5%) of the classifications occur in agreement with the groups assigned by consensus between our clinical experts. Our findings suggest the existence of different disease-based phenotypic morphologies in TMJ OA. Our preliminary findings with statistical shape modeling based biomarkers may provide a quantitative staging of the disease. The methodology used in this study is included in an open source image analysis toolbox, to ensure reproducibility and appropriate distribution and dissemination of the solution proposed.

Compressed sensing (CS) can accelerate magnetic resonance spectroscopic imaging (MRSI), facilitating its widespread clinical integration. The objective of this study was to assess the effect of different undersampling strategy on CS-MRSI reconstruction quality. Phantom data were acquired on a Philips 3 T Ingenia scanner. Four types of undersampling masks, corresponding to each strategy, namely, low resolution, variable density, iterative design, and a priori were simulated in Matlab and retrospectively applied to the test 1X MRSI data to generate undersampled datasets corresponding to the 2X – 5X, and 7X accelerations for each type of mask. Reconstruction parameters were kept the same in each case(all masks and accelerations) to ensure that any resulting differences can be attributed to the type of mask being employed. The reconstructed datasets from each mask were statistically compared with the reference 1X, and assessed using metrics like the root mean square error and metabolite ratios. Simulation results indicate that both the a priori and variable density undersampling masks maintain high fidelity with the 1X up to five-fold acceleration. The low resolution mask based reconstructions showed statistically significant differences from the 1X with the reconstruction failing at 3X, while the iterative design reconstructions maintained fidelity with the 1X till 4X acceleration. In summary, a pilot study was conducted to identify an optimal sampling mask in CS-MRSI. Simulation results demonstrate that the a priori and variable density masks can provide statistically similar results to the fully sampled reference. Future work would involve implementing these two masks prospectively on a clinical scanner.

Recent studies have demonstrated the difficulties to replicate scientific findings and/or experiments published in past.¹ The effects seen in the replicated experiments were smaller than previously reported. Some of the explanations for these findings include the complexity of the experimental design and the pressure on researches to report positive findings. The International Committee of Medical Journal Editors (ICMJE) suggests that every study considered for publication must submit a plan to share the de-identified patient data no later than 6 months after publication. There is a growing demand to enhance the management of clinical data, facilitate data sharing across institutions and also to keep track of the data from previous experiments. The ultimate goal is to assure the reproducibility of experiments in the future. This paper describes Shiny-tooth, a web based application created to improve clinical data acquisition during the clinical trial; data federation of such data as well as morphological data derived from medical images; Currently, this application is being used to store clinical data from an osteoarthritis (OA) study. This work is submitted to the SPIE Biomedical Applications in Molecular, Structural, and Functional Imaging conference.

Near infrared spectroscopy (NIRS) is an optical imaging technique that allows real-time measurements of Oxy and Deoxy-hemoglobin concentrations in human body tissue. In functional NIRS (fNIRS), this technique is used to study cortical activation in response to changes in neural activity. However, analysis of activation regions using NIRS is a challenging task in the field of medical image analysis and despite existing solutions, no homogeneous analysis method has yet been determined. For that reason, the aim of our present study is to report the use of an optical flow method for the analysis of cortical activation using near-infrared spectroscopy signals. We used real fNIRS data recorded from a noxious stimulation experiment as base of our implementation. To compute the optical flow algorithm, we first arrange NIRS signals (Oxy-hemoglobin) following our 24 channels (12 channels per hemisphere) head-probe configuration to create image-like samples. We then used two consecutive fNIRS samples per hemisphere as input frames for the optical flow algorithm, making one computation per hemisphere. The output from these two computations is the velocity field representing cortical activation from each hemisphere. The experimental results showed that the radial structure of flow vectors exhibited the origin of cortical activity, the development of stimulation as expansion or contraction of such flow vectors, and the flow of activation patterns may suggest prediction in cortical activity. The present study demonstrates that optical flow provides a power tool for the analysis of NIRS signals. Finally, we suggested a novel idea to identify pain status in nonverbal patients by using optical flow motion vectors; however, this idea will be study further in our future research.