Accurate measurements of human body fat distribution are desirable because excessive body fat is associated with impaired insulin sensitivity, type 2 diabetes mellitus (T2DM) and cardiovascular disease. In this study, we hypothesized that the performance of water suppressed (WS) MRI is superior to non-water suppressed (NWS) MRI for volumetric assessment of abdominal subcutaneous (SAT), intramuscular (IMAT), visceral (VAT), and total (TAT) adipose tissues. We acquired T₁-weighted images on a 3T MRI system (TIM Trio, Siemens), which was analyzed using semi–automated segmentation software that employs a fuzzy c-means (FCM) clustering algorithm. Sixteen contiguous axial slices, centered at the L4–L5 level of the abdomen, were acquired in eight T2DM subjects with water suppression (WS) and without (NWS). Histograms from WS images show improved separation of non-fatty tissue pixels from fatty tissue pixels, compared to NWS images. Paired t-tests of WS versus NWS showed a statistically significant lower volume of lipid in the WS images for VAT (145.3 cc less, p=0.006) and IMAT (305 cc less, p<0.001), but not SAT (14.1 cc more, NS). WS measurements of TAT also resulted in lower fat volumes (436.1 cc less, p=0.002). There is strong correlation between WS and NWS quantification methods for SAT measurements (r=0.999), but poorer correlation for VAT studies (r=0.845). These results suggest that NWS pulse sequences may overestimate adipose tissue volumes and that WS pulse sequences are more desirable due to the higher contrast generated between fatty and non-fatty tissues.

We introduce a new protocol to obtain radial Ultra-Short TE (UTE) MRI Cine of the beating mouse heart within reasonable measurement time. The method is based on a self-gated UTE with golden angle radial acquisition and compressed sensing reconstruction. The stochastic nature of the retrospective triggering acquisition scheme produces an under-sampled and random kt-space filling that allows for compressed sensing reconstruction, hence reducing scan time. As a standard, an intragate multislice FLASH sequence with an acquisition time of 4.5 min per slice was used to produce standard Cine movies of 4 mice hearts with 15 frames per cardiac cycle. The proposed self-gated sequence is used to produce Cine movies with short echo time. The total scan time was 11 min per slice. 6 slices were planned to cover the heart from the base to the apex. 2X, 4X and 6X under-sampled k-spaces cine movies were produced from 2, 1 and 0.7 min data acquisitions for each slice. The accelerated cine movies of the mouse hearts were successfully reconstructed with a compressed sensing algorithm. Compared to the FLASH cine images, the UTE images showed much less flow artifacts due to the short echo time. Besides, the accelerated movies had high image quality and the undersampling artifacts were effectively removed. Left ventricular functional parameters derived from the standard and the accelerated cine movies were nearly identical.

Polycystic kidney disease (PKD) is a major cause of renal failure. Despite recent advances in understanding the biochemistry and genetics of PKD, the functional mechanisms underpinning the declines in renal function observed in the disorder are not well established. No studies investigating the distribution of cysts within polycystic kidneys exist. This work introduces regional cyst concentration as a new biomarker for evaluation of patients suffering from PKD. We derive a method to define central and peripheral regions of the kidney, approximating the anatomical division between cortex and medulla, and apply it to two cohorts of ten patients with early/mild or late/severe disease. Our results from the late/severe cohort show peripheral cyst concentration correlates with the current standard PKD biomarker, total kidney volume (TKV), signi cantly better than central cyst concentration (p < 0.05). We also find that cyst concentration was globally increased in the late/severe cohort (p << 0.01) compared to the early/mild cohort, for both central and peripheral regions. These findings show cysts in PKD are not distributed homogeneously throughout the renal tissues.

While the plethora of information from multiple imaging and non-imaging data streams presents an opportunity for discovery of fused multimodal, multiscale biomarkers, they also introduce multiple independent sources of noise that hinder their collective utility. The goal of this work is to create fused predictors of disease diagnosis and prognosis by combining multiple data streams, which we hypothesize will provide improved performance as compared to predictors from individual data streams. To achieve this goal, we introduce supervised multiview canonical correlation analysis (sMVCCA), a novel data fusion method that attempts to find a common representation for multiscale, multimodal data where class separation is maximized while noise is minimized. In doing so, sMVCCA assumes that the different sources of information are complementary and thereby act synergistically when combined. Although this method can be applied to any number of modalities and to any disease domain, we demonstrate its utility using three datasets. We fuse (i) 1.5 Tesla (T) magnetic resonance imaging (MRI) features with cerbrospinal fluid (CSF) proteomic measurements for early diagnosis of Alzheimer’s disease (n = 30), (ii) 3T Dynamic Contrast Enhanced (DCE) MRI and T2w MRI for in vivo prediction of prostate cancer grade on a per slice basis (n = 33) and (iii) quantitative histomorphometric features of glands and proteomic measurements from mass spectrometry for prediction of 5 year biochemical recurrence postradical prostatectomy (n = 40). Random Forest classifier applied to the sMVCCA fused subspace, as compared to that of MVCCA, PCA and LDA, yielded the highest classification AUC of 0.82 +/- 0.05, 0.76 +/- 0.01, 0.70 +/- 0.07, respectively for the aforementioned datasets. In addition, sMVCCA fused subspace provided 13.6%, 7.6% and 15.3% increase in AUC as compared with that of the best performing individual view in each of the three datasets, respectively. For the biochemical recurrence dataset, Kaplan-Meier curves generated from classifier prediction in the fused subspace reached the significance threshold (p = 0.05) for distinguishing between patients with and without 5 year biochemical recurrence, unlike those generated from classifier predictions of the individual modalities.

Recent neural signal decoding studies based on functional magnetic resonance imaging (fMRI) have identified the specific image presenting to the subject from a set of potential images, and some studies extend neural decoding into image reconstruction, i.e. image contents that the subject perceived were decoded from the fMRI signals recorded during the subject looking at images. In this paper, we decoded the target images using fMRI signals and described a target image searching method based on the relationship between target image stimuli and fMRI activity. We recorded fMRI data during a serial visual stimuli image presentation task, some of the stimuli images were target images and the rest images were non-target ones. Our fMRI data analysis results showed that in the serial visual presentation task, target images elicited a stereotypical response in the fMRI, which can be detected by multi-voxel pattern analysis (MVPA). Classifiers designed with support vector machine (SVM) used this response to decipher target images from non-target images. The leave-one-run-out cross-validation showed that we can pick out the target images with a possibility far above the chance level, which indicate that there’s a neural signatures correlated with the target image recognition process in the human systems.

Previous behavioral research has proved that individuals process own- and other-race faces differently. One well-known effect is the other-race effect (ORE), which indicates that individuals categorize other-race faces more accurately and faster than own-race faces. The existed functional magnetic resonance imaging (fMRI) studies of the other-race effect mainly focused on the racial prejudice and the socio-affective differences towards own- and other-race face. In the present fMRI study, we adopted a race-categorization task to determine the activation level differences between categorizing own- and other-race faces. Thirty one Chinese participants who live in China with Chinese as the majority and who had no direct contact with Caucasian individual were recruited in the present study. We used the group independent component analysis (ICA), which is a method of blind source signal separation that has proven to be promising for analysis of fMRI data. We separated the entail data into 56 components which is estimated based on one subject using the Minimal Description Length (MDL) criteria. The components sorted based on the multiple linear regression temporal sorting criteria, and the fit regression parameters were used in performing statistical test to evaluate the task-relatedness of the components. The one way anova was performed to test the significance of the component time course in different conditions. Our result showed that the areas, which coordinates is similar to the right FFA coordinates that previous studies reported, were greater activated for own-race faces than other-race faces, while the precuneus showed greater activation for other-race faces than own-race faces.

A functional connectivity (FC) analysis from resting-state functional MRI (rsfMRI) is gaining its popularity toward the clinical application such as diagnosis of neuropsychiatric disease. To delineate the brain networks from rsfMRI data, non-neuronal components including head motions and physiological artifacts mainly observed in cerebrospinal fluid (CSF), white matter (WM) along with a global brain signal have been regarded as nuisance variables in calculating the FC level. However, it is still unclear how the non-neuronal components can affect the performance toward diagnosis of neuropsychiatric disease. In this study, a systematic comparison of classification performance of schizophrenia patients was provided employing the partial correlation coefficients (CCs) as feature elements. Pair-wise partial CCs were calculated between brain regions, in which six combinatorial sets of nuisance variables were considered. The partial CCs were used as candidate feature elements followed by feature selection based on the statistical significance test between two groups in the training set. Once a linear support vector machine was trained using the selected features from the training set, the classification performance was evaluated using the features from the test set (i.e. leaveone- out cross validation scheme). From the results, the error rate using all non-neuronal components as nuisance variables (12.4%) was significantly lower than those using remaining combination of non-neuronal components as nuisance variables (13.8 ~ 20.0%). In conclusion, the non-neuronal components substantially degraded the automated diagnosis performance, which supports our hypothesis that the non-neuronal components are crucial in controlling the automated diagnosis performance of the neuropsychiatric disease using an fMRI modality.

Amblyopia is a common yet hard-to-cure disease in children and results in poor or blurred vision. Some efforts such as voxel-based analysis, cortical thickness analysis have been tried to reveal the pathogenesis of amblyopia. However, few studies focused on alterations of the functional connectivity (FC) in amblyopia. In this study, we analyzed the abnormalities of amblyopia patients by both the seed-based FC with the left/right primary visual cortex and the network constructed throughout the whole brain. Experiments showed the following results: (1)As for the seed-based FC analysis, FC between superior occipital gyrus and the primary visual cortex was found to significantly decrease in both sides. The abnormalities were also found in lingual gyrus. The results may reflect functional deficits both in dorsal stream and ventral stream. (2)Two increased functional connectivities and 64 decreased functional connectivities were found in the whole brain network analysis. The decreased functional connectivities most concentrate in the temporal cortex. The results suggest that amblyopia may be caused by the deficits in the visual information transmission.

Cortical thinning and metabolic reduction can be possible imaging biomarkers for Alzheimer’s disease (AD) diagnosis and monitoring. Many techniques have been developed for the cortical measurement and widely used for the clinical statistical studies. However, the measurement consistency of individuals, an essential requirement for a clinically useful technique, requires proper further investigation. Here we leverage our previously developed BSIM technique ¹ to measure cortical thickness and thinning and use it with longitudinal MRI from ADNI to investigate measurement consistency and spatial resolution. 10 normal, 10 MCI, and 10 AD subjects in their 70s were selected for the study. Consistent cortical thinning patterns were observed in all baseline and follow up images. Rapid cortical thinning was shown in some MCI and AD cases. To evaluate the correctness of the cortical measurement, we compared longitudinal cortical thinning with clinical diagnosis and longitudinal PET metabolic reduction measured using 3D-SSP technique² for the same person. Longitudinal MR cortical thinning and corresponding PET metabolic reduction showed high level pattern similarity revealing certain correlations worthy of further studies. Severe cortical thinning that might link to disease conversion from MCI to AD was observed in two cases. In summary, our results suggest that consistent cortical measurements using our technique may provide means for clinical diagnosis and monitoring at individual patient’s level and MR cortical thinning measurement can complement PET metabolic reduction measurement.

Alzheimer’s disease (AD) is a progressive neurodegenerative disease with the clinical symptom of the continuous deterioration of cognitive and memory functions. Multiple diffusion tensor imaging (DTI) indices such as fractional anisotropy (FA) and mean diffusivity (MD) can successfully explain the white matter damages in AD patients. However, most studies focused on the univariate measures (voxel-based analysis) to examine the differences between AD patients and normal controls (NCs). In this investigation, we applied a multivariate independent component analysis (ICA) to investigate the white matter covariances based on FA measurement from DTI data in 35 AD patients and 45 NCs from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database. We found that six independent components (ICs) showed significant FA reductions in white matter covariances in AD compared with NC, including the genu and splenium of corpus callosum (IC-1 and IC-2), middle temporal gyral of temporal lobe (IC-3), sub-gyral of frontal lobe (IC-4 and IC-5) and sub-gyral of parietal lobe (IC-6). Our findings revealed covariant white matter loss in AD patients and suggest that the unsupervised data-driven ICA method is effective to explore the changes of FA in AD. This study assists us in understanding the mechanism of white matter covariant reductions in the development of AD.

The common squirrel monkey, Saimiri sciureus, is a New World monkey with functional and microstructural organization of central nervous system similar to that of humans. It is one of the most commonly used South American primates in biomedical research. Unlike its Old World macaque cousins, no digital atlases have described the organization of the squirrel monkey brain. Here, we present a multi-modal magnetic resonance imaging (MRI) atlas constructed from the brain of an adult female squirrel monkey. In vivo MRI acquisitions include high resolution T2 structural imaging and low resolution diffusion tensor imaging. Ex vivo MRI acquisitions include high resolution T2 structural imaging and high resolution diffusion tensor imaging. Cortical regions were manually annotated on the co-registered volumes based on published histological sections.

Spectral-domain optical coherence tomography (SD-OCT) finds widespread use clinically for the detection and management of ocular diseases. This non-invasive imaging modality has also begun to find frequent use in research studies involving animals such as mice. Numerous approaches have been proposed for the segmentation of retinal surfaces in SD-OCT images obtained from human subjects; however, the segmentation of retinal surfaces in mice scans is not as well-studied. In this work, we describe a graph-theoretic segmentation approach for the simultaneous segmentation of 10 retinal surfaces in SD-OCT scans of mice that incorporates learned shape priors. We compared the method to a baseline approach that did not incorporate learned shape priors and observed that the overall unsigned border position errors reduced from 3.58 +/- 1.33 μm to 3.20 +/- 0.56 μm.

Anterior segment optical coherence tomography (AS-OCT) is a non-invasive imaging modality that allows for the quantitative assessment of corneal thicknesses. Automated approaches for these measurements are not readily available and therefore measurements are often obtained manually. While graph-based approaches that enable the optimal simultaneous segmentation of multiple 3D surfaces have been proposed and applied to 3D optical coherence tomography volumes of the back of the eye, such approaches have not been applied for the segmentation of the corneal surfaces. In this work we propose adapting this graph-based method for the automated 3D segmentation of three corneal surfaces in AS-OCT images and to measure total corneal thickness. The approach is evaluated based on 34 AS-OCT volumes obtained from both eyes of 17 mice with varying corneal thicknesses. The segmentation accuracy was assessed using unsigned border positioning errors and was found to be 1.82 +/- 0.81 μm. We also assessed an average relative error in total layer thickness measurements which was found to be 2.27%.

Optical coherence tomography (OCT) is a powerful imaging tool that is particularly useful for exploring retinal abnormalities in ophthalmological diseases. Recently, it has been used to track changes in the eye associated with neurological diseases such as multiple sclerosis (MS) where certain tissue layer thicknesses have been associated with disease progression. A small percentage of MS patients also exhibit what has been called microcystic macular edema (MME), where uid collections that are thought to be pseudocysts appear in the inner nuclear layer. Very little is known about the cause of this condition so it is important to be able to identify precisely where these pseudocysts occur within the retina. This identi cation would be an important rst step towards furthering our understanding. In this work, we present a detection algorithm to nd these pseudocysts and to report on their spatial distribution. Our approach uses a random forest classi er trained on manual segmentation data to classify each voxel as pseudocyst or not. Despite having a small sample size of ve subjects, the algorithm correctly identi es 84.6% of pseudocysts as compared to manual delineation. Finally, using our method, we show that the spatial distribution of pseudocysts within the macula are generally contained within an annulus around the fovea.

Despite the current availability in resource-rich regions of advanced technologies in scanning and 3-D imaging in current ophthalmology practice, world-wide screening tests for early detection and progression of glaucoma still consist of a variety of simple tools, including fundus image-based parameters such as CDR (cup to disc diameter ratio) and CAR (cup to disc area ratio), especially in resource -poor regions. Reliable automated computation of the relevant parameters from fundus image sequences requires robust non-rigid registration and segmentation techniques. Recent research work demonstrated that proper non-rigid registration of multi-view monocular fundus image sequences could result in acceptable segmentation of cup boundaries for automated computation of CAR and CDR. This research work introduces a composite diffeomorphic demons registration algorithm for segmentation of cup boundaries from a sequence of monocular images and compares the resulting CAR and CDR values with those computed manually by experts and from 3-D visualization of stereo pairs. Our preliminary results show that the automated computation of CDR and CAR from composite diffeomorphic segmentation of monocular image sequences yield values comparable with those from the other two techniques and thus may provide global healthcare with a cost-effective yet accurate tool for management of glaucoma in its early stage.

4D or dynamic imaging of the thorax has many potential applications [1, 2]. CT and MRI offer sufficient speed to acquire motion information via 4D imaging. However they have different constraints and requirements. For both modalities both prospective and retrospective respiratory gating and tracking techniques have been developed [3, 4]. For pediatric imaging, x-ray radiation becomes a primary concern and MRI remains as the de facto choice. The pediatric subjects we deal with often suffer from extreme malformations of their chest wall, diaphragm, and/or spine, as such patient cooperation needed by some of the gating and tracking techniques are difficult to realize without causing patient discomfort. Moreover, we are interested in the mechanical function of their thorax in its natural form in tidal breathing. Therefore free-breathing MRI acquisition is the ideal modality of imaging for these patients. In our set up, for each coronal (or sagittal) slice position, slice images are acquired at a rate of about 200-300 ms/slice over several natural breathing cycles. This produces typically several thousands of slices which contain both the anatomic and dynamic information. However, it is not trivial to form a consistent and well defined 4D volume from these data. In this paper, we present a novel graph-based combinatorial optimization solution for constructing the best possible 4D scene from such data entirely in the digital domain. Our proposed method is purely image-based and does not need breath holding or any external surrogates or instruments to record respiratory motion or tidal volume. Both adult and children patients’ data are used to illustrate the performance of the proposed method. Experimental results show that the reconstructed 4D scenes are smooth and consistent spatially and temporally, agreeing with known shape and motion of the lungs.

The optimal management of unruptured aneurysms is controversial, and current decision making is mainly based on aneurysm size and location. Incidentally detected unruptured aneurysms less than 5mm in diameter should be treated conservatively. However, small unruptured aneurysms also bleed. Risk factors based on the hemodynamic forces exerted over the arterial wall have been investigated using image-based computational fluid dynamic (CFD) methodologies during the last decade. Accurate estimation of wall shear stress (WSS) is required to properly study associations between flow features and aneurysm processes. Previous works showed that Newtonian and non-Newtonian (Casson) models produce similar WSS distributions and characterization, with no significant differences. Other authors showed that the WSS distribution computed from time-averaged velocity fields is significantly higher for the Newtonian model where WSS is low. In this work we reconstructed ten patient-specific CFD models from angiography images to investigate the time evolution of WSS at selected locations such as aneurysm blebs (low WSS), and the parent artery close to the aneurysm neck (high WSS). When averaging all cases it is seen that the estimation of the time-averaged WSS, the peak WSS and the minimum WSS value before the systolic peak were all higher when the Casson rheology was considered. However, none of them showed statistically significant differences. At the afferent artery Casson rheology systematically predicted higher WSS values. On the other hand, at the selected blebs either Newtonian or Casson WSS estimations are higher in some phases of the cardiac cycle. Those observations differ among individual cases.

Biological cells are composed primarily of water; and as such are challenging to image without staining since the induced intensity modulation of transmitted or reflected light is typically insufficient to permit acceptable contrast for optical imaging. This issue may be resolved with the aid of exogenous contrast agents, but this often has a deleterious effect on the cell and precludes in vivo imaging. A unique approach to this problem is afforded by the phase contrast microscope in which optical-path differences in transmitted light is exploited as a contrast mechanism for qualitative imaging. In recent years however, several quantitative phase imaging techniques have been developed which allow for diffraction limited endogenous-contrast imaging with excellent temporal resolution. We hereby present a laser scanning technique for quantitative phase imaging which achieves sub-diffraction limited resolution at the expense of temporal resolution. This instrument is based on a stabilized fiber interfometer which is incorporated into a near-field scanning optical microscope (NSOM) for tri-modal imaging. Our latest results will focus on modifications made to this system to facilitate imaging in a liquid environment. A simple approach for achieving stable shear-force feedback operation in a liquid will be presented. Acquired high resolution images of white blood cells revealed detailed sub-cellular features. Images of fibroblast cells in air and in a liquid environment confirm the efficacy of the feedback operation in a liquid. Moreover, we demonstrate cell refractometry capability without the need for ad hoc modifications. These results clearly highlight the unique potential of this instrument for the study of living cells.

Parametric imaging maps (PIM’s) derived from digital subtraction angiography (DSA) for the cerebral arterial flow assessment in clinical settings have been proposed, but experiments have yet to determine the reliability of such studies. For this study, we have observed the effects of different injection techniques on PIM’s. A flow circuit set to physiologic conditions was created using an internal carotid artery phantom. PIM’s were derived for two catheter positions, two different contrast bolus injection volumes (5ml and 10 ml), and four injection rates (5, 10, 15 and 20 ml/s). Using a gamma variate fitting approach, we derived PIM’s for mean-transit-time (MTT), time-to-peak (TTP) and bolus-arrivaltime (BAT). For the same injection rates, a larger bolus resulted in an increased MTT and TTP, while a faster injection rate resulted in a shorter MTT, TTP, and BAT. In addition, the position of the catheter tip within the vasculature directly affected the PIM. The experiment showed that the PIM is strongly correlated with the injection conditions, and, therefore, they have to be interpreted with caution. PIM images must be taken from the same patient to be able to be meaningfully compared. These comparisons can include pre- and post-treatment images taken immediately before and after an interventional procedure or simultaneous arterial flow comparisons through the left and right cerebral hemispheres. Due to the strong correlation between PIM and injection conditions, this study indicates that this assessment method should be used only to compare flow changes before and after treatment within the same patient using the same injection conditions.

Additive manufacturing (3D printing) technology offers a great opportunity towards development of patient-specific vascular anatomic models, for medical device testing and physiological condition evaluation. However, the development process is not yet well established and there are various limitations depending on the printing materials, the technology and the printer resolution. Patient-specific neuro-vascular anatomy was acquired from computed tomography angiography and rotational digital subtraction angiography (DSA). The volumes were imported into a Vitrea 3D workstation (Vital Images Inc.) and the vascular lumen of various vessels and pathologies were segmented using a “marching cubes” algorithm. The results were exported as Stereo Lithographic (STL) files and were further processed by smoothing, trimming, and wall extrusion (to add a custom wall to the model). The models were printed using a Polyjet printer, Eden 260V (Objet-Stratasys). To verify the phantom geometry accuracy, the phantom was reimaged using rotational DSA, and the new data was compared with the initial patient data. The most challenging part of the phantom manufacturing was removal of support material. This aspect could be a serious hurdle in building very tortuous phantoms or small vessels. The accuracy of the printed models was very good: distance analysis showed average differences of 120 μm between the patient and the phantom reconstructed volume dimensions. Most errors were due to residual support material left in the lumen of the phantom. Despite the post-printing challenges experienced during the support cleaning, this technology could be a tremendous benefit to medical research such as in device development and testing.

Patient-specific blood pressure values in the human aorta are an important parameter in the management of cardiovascular diseases. A direct measurement of these values is only possible by invasive catheterization at a limited number of measurement sites. To overcome these drawbacks, two non-invasive approaches of computing patient-specific relative aortic blood pressure maps throughout the entire aortic vessel volume are investigated by our group. The first approach uses computations from complete time-resolved, three-dimensional flow velocity fields acquired by phasecontrast magnetic resonance imaging (PC-MRI), whereas the second approach relies on computational fluid dynamics (CFD) simulations with ultrasound-based boundary conditions. A detailed evaluation of these computational methods under realistic conditions is necessary in order to investigate their overall robustness and accuracy as well as their sensitivity to certain algorithmic parameters. We present a comparative study of the two blood pressure computation methods in an experimental phantom setup, which mimics a simplified thoracic aorta. The comparative analysis includes the investigation of the impact of algorithmic parameters on the MRI-based blood pressure computation and the impact of extracting pressure maps in a voxel grid from the CFD simulations. Overall, a very good agreement between the results of the two computational approaches can be observed despite the fact that both methods used completely separate measurements as input data. Therefore, the comparative study of the presented work indicates that both non-invasive pressure computation methods show an excellent robustness and accuracy and can therefore be used for research purposes in the management of cardiovascular diseases.

Dynamic cardiac CT perfusion (CTP) is a high resolution, non-invasive technique for assessing myocardial blood ow (MBF), which in concert with coronary CT angiography enable CT to provide a unique, comprehensive, fast analysis of both coronary anatomy and functional ow. We assessed perfusion in a porcine model with and without coronary occlusion. To induce occlusion, each animal underwent left anterior descending (LAD) stent implantation and angioplasty balloon insertion. Normal ow condition was obtained with balloon completely de ated. Partial occlusion was induced by balloon in ation against the stent with FFR used to assess the extent of occlusion. Prospective ECG-triggered partial scan images were acquired at end systole (45% R-R) using a multi-detector CT (MDCT) scanner. Images were reconstructed using FBP and a hybrid iterative reconstruction (iDose4, Philips Healthcare). Processing included: beam hardening (BH) correction, registration of image volumes using 3D cubic B-spline normalized mutual-information, and spatio-temporal bilateral ltering to reduce partial scan artifacts and noise variation. Absolute blood ow was calculated with a deconvolutionbased approach using singular value decomposition (SVD). Arterial input function was estimated from the left ventricle (LV) cavity. Regions of interest (ROIs) were identi ed in healthy and ischemic myocardium and compared in normal and occluded conditions. Under-perfusion was detected in the correct LAD territory and ow reduction agreed well with FFR measurements. Flow was reduced, on average, in LAD territories by 54%.

We propose a novel framework of robust kinetic parameter estimation applied to absolute ow quanti cation in dynamic PET imaging. Kinetic parameter estimation is formulated as a nonlinear least squares with spatial constraints problem (NLLS-SC) where the spatial constraints are computed from a physiologically driven clustering of dynamic images, and used to reduce noise contamination. An ideal clustering of dynamic images depends on the underlying physiology of functional regions, and in turn, physiological processes are quanti ed by kinetic parameter estimation. Physiologically driven clustering of dynamic images is performed using a clustering algorithm (e.g. K-means, Spectral Clustering etc) with Kinetic modeling in an iterative handshaking fashion. This gives a map of labels where each functionally homogenous cluster is represented by mean kinetics (cluster centroid). Parametric images are acquired by solving the NLLS-SC problem for each voxel which penalizes spatial variations from its mean kinetics. This substantially reduces noise in the estimation process for each voxel by utilizing kinetic information from physiologically similar voxels (cluster members). Resolution degradation is also substantially minimized as no spatial smoothing between heterogeneous functional regions is performed. The proposed framework is shown to improve the quantitative accuracy of Myocardial Perfusion (MP) PET imaging, and in turn, has the long-term potential to enhance capabilities of MP PET in the detection, staging and management of coronary artery disease.

A feasibility study of improved PET-CT tumor imaging approach is presented. A single PET-CT routine includes three different techniques: 18F-FDG early dynamic blood flow intended for perfusion assessment; standard late 18F-FDG uptake; and high-resolution contrast-enhanced CT enabling tissue texture analysis. Both PET protocols utilize the same single standard radiotracer dose administration. Quantitative volumetric arterial perfusion maps are derived from the reconstructed dynamic PET images corresponding to successive acquisition time intervals of 3 seconds only. For achieving high accuracy, the analysis algorithm differentiates the first-pass arterial flow from other interfering dynamic effects, and a noise reduction scheme based on adaptive total-variation minimization aims to provide appreciable quantitative map in physical conditions of high noise and low spatial resolution. The CT texture analysis comprises a practical and robust method for generating volumetric tissue irregularity maps. A local map value is represented by the entropy function which is derived from a weighted co-occurrence matrix histogram of the corresponding image voxel three-dimensional vicinity. Unique entropy scaling scheme and parameter optimization process, as well as appropriate scaling for varying image noise levels and contrast agent concentrations, improve the results toward quantitative absolute measure with respect to diverse scanning conditions and key analysis parameters. Representative imaging results are demonstrated on several clinical cases involving different organs and cancer types. In these cases, significant tumor characterization relative to the normal surrounding tissues is seen on the quantitative maps of all three imaging techniques. This proof of concept can lead the way to a new practical diagnostic imaging application.

Benign radiation-induced lung injury is a common finding following stereotactic ablative radiotherapy (SABR) for lung cancer, and is often difficult to differentiate from a recurring tumour due to the ablative doses and highly conformal treatment with SABR. Current approaches to treatment response assessment have shown limited ability to predict recurrence within 6 months of treatment. The purpose of our study was to evaluate the accuracy of second order texture statistics for prediction of eventual recurrence based on computed tomography (CT) images acquired within 6 months of treatment, and compare with the performance of first order appearance and lesion size measures. Consolidative and ground-glass opacity (GGO) regions were manually delineated on post-SABR CT images. Automatic consolidation expansion was also investigated to act as a surrogate for GGO position. The top features for prediction of recurrence were all texture features within the GGO and included energy, entropy, correlation, inertia, and first order texture (standard deviation of density). These predicted recurrence with 2-fold cross validation (CV) accuracies of 70–77% at 2– 5 months post-SABR, with energy, entropy, and first order texture having leave-one-out CV accuracies greater than 80%. Our results also suggest that automatic expansion of the consolidation region could eliminate the need for manual delineation, and produced reproducible results when compared to manually delineated GGO. If validated on a larger data set, this could lead to a clinically useful computer-aided diagnosis system for prediction of recurrence within 6 months of SABR and allow for early salvage therapy for patients with recurrence.

To match recent clinical dual energy (DE) CT studies focusing on the lung, similar developments for DE micro-CT of the rodent lung are required. Our group has been actively engaged in designing pulmonary gating techniques for micro- CT, and has also introduced the first DE micro-CT imaging method of the rodent lung. The aim of this study was to assess the feasibility of DE micro-CT imaging for the evaluation of airway obstruction in mice, and to compare the method with micro single photon emission computed tomography (micro-SPECT) using technetium-99m labeled macroaggregated albumin (^99mTc-MAA). The results suggest that the induced pulmonary airway obstruction causes either atelectasis, or air-trapping similar to asthma or chronic bronchitis. Atelectasis could only be detected at early time points in DE micro-CT images, and is associated with a large increase in blood fraction and decrease in air fraction. Air trapping had an opposite effect with larger air fraction and decreased blood fraction shown by DE micro-CT. The decrease in perfusion to the hypoventilated lung (hypoxic vasoconstriction) is also seen in micro-SPECT. The proposed DE micro-CT technique for imaging localized airway obstruction performed well in our evaluation, and provides a higher resolution compared to micro-SPECT. Both DE micro-CT and micro-SPECT provide critical, quantitative lung biomarkers for image-based anatomical and functional information in the small animal. The methods are readily linked to clinical methods allowing direct comparison of preclinical and clinical results.

Regional changes in lung microstructure are an important component of several common lung disorders and even in healthy lungs alveolar mechanics are poorly understood. Existing techniques capable of studying the lung microstructure have various limitations including poor temporal resolution. We present a technique, which can measure the distribution of alveolar diameters from a single, phase contrast chest X-ray. We present the results of analysis of synchrotron images of a rabbit pup’s lungs, which we compare with high-resolution computed tomography images. We demonstrate that measurements can be made with an exposure time of 40 ms, highlighting the unique potential for performing dynamic in vivo measurements. Applications include disease detection, assessment of therapeutics and physiological studies.

Alterations in the structure and function of the pulmonary system that occur in disease or injury often give rise to measurable spectral, spatial and/or temporal changes in lung sound production and transmission. These changes, if properly quantified, might provide additional information about the etiology, severity and location of trauma, injury, or pathology. With this in mind, the authors are developing a comprehensive computer simulation model of pulmonary acoustics, known as The Audible Human Project™. Its purpose is to improve our understanding of pulmonary acoustics and to aid in interpreting measurements of sound and vibration in the lungs generated by airway insonification, natural breath sounds, and external stimuli on the chest surface, such as that used in elastography. As a part of this development process, finite element (FE) models were constructed of an excised pig lung that also underwent experimental studies. Within these models, the complex airway structure was created via two methods: x-ray CT image segmentation and through an algorithmic means called Constrained Constructive Optimization (CCO). CCO was implemented to expedite the segmentation process, as airway segments can be grown digitally. These two approaches were used in FE simulations of the surface motion on the lung as a result of sound input into the trachea. Simulation results were compared to experimental measurements. By testing how close these models are to experimental measurements, we are evaluating whether CCO can be used as a means to efficiently construct physiologically relevant airway trees.

Dual modality micro-CT and SPECT imaging can play an important role in preclinical studies designed to investigate mechanisms, progression, and therapies for acute lung injury in rats. SPECT imaging involves examining the uptake of radiopharmaceuticals within the lung, with the hypothesis that uptake is sensitive to the health or disease status of the lung tissue. Methods of quantifying lung uptake and comparison of right and left lung uptake generally begin with identifying and segmenting the lung region within the 3D reconstructed SPECT volume. However, identification of the lung boundaries and the fissure between the left and right lung is not always possible from the SPECT images directly since the radiopharmaceutical may be taken up by other surrounding tissues. Thus, our SPECT protocol begins with a fast CT scan, the lung boundaries are identified from the CT volume, and the CT region is coregistered with the SPECT volume to obtain the SPECT lung region. Segmenting rat lungs within the CT volume is particularly challenging due to the relatively low resolution of the images and the rat’s unique anatomy. Thus, we have developed an automated segmentation algorithm for low resolution micro-CT scans that utilizes depth maps to detect fissures on the surface of the lung volume. The fissure’s surface location is in turn used to interpolate the fissure throughout the lung volume. Results indicate that the segmentation method results in left and right lung regions consistent with rat lung anatomy.

Objectives: We aimed to develop a way to rapidly generate multi-modality (MRI-CT) pulmonary imaging structurefunction maps using novel non-rigid image registration methods. This objective is part of our overarching goal to provide an image processing pipeline to generate pulmonary structure-function maps and guide airway-targeted therapies.

Methods: Anatomical ¹H and functional ³He MRI were acquired in 5 healthy asymptomatic ex-smokers and 7 ex-smokers with chronic obstructive pulmonary disease (COPD) at inspiration breath-hold. Thoracic CT was performed within ten minutes of MRI using the same breath-hold volume. Landmark-based affine registration methods previously validated for imaging of COPD, was based on corresponding fiducial markers located in both CT and ¹H MRI coronal slices and compared with shape-based CT-MRI non-rigid registration. Shape-based CT-MRI registration was developed by first identifying the shapes of the lung cavities manually, and then registering the two shapes using affine and thin-plate spline algorithms. We compared registration accuracy using the fiducial localization error (FLE) and target registration error (TRE).

Results: For landmark-based registration, the TRE was 8.4±5.3 mm for whole lung and 7.8±4.6 mm for the R and L lungs registered independently (p=0.4). For shape-based registration, the TRE was 8.0±4.6 mm for whole lung as compared to 6.9±4.4 mm for the R and L lung registered independently and this difference was significant (p=0.01). The difference for shape-based (6.9±4.4 mm) and landmark-based R and L lung registration (7.8±4.6 mm) was also significant (p=.04)

Conclusion: Shape-based registration TRE was significantly improved compared to landmark-based registration when considering L and R lungs independently.

There are several disease conditions that lead to upper airway restrictive disorders. In the study of these conditions, it is important to take into account the dynamic nature of the upper airway. Currently, dynamic MRI is the modality of choice for studying these diseases. Unfortunately, the contrast resolution obtainable in the images poses many challenges for an effective segmentation of the upper airway structures. No viable methods have been developed to date to solve this problem. In this paper, we demonstrate the adaptation of the iterative relative fuzzy connectedness (IRFC) algorithm for this application as a potential practical tool. After preprocessing to correct for background image non-uniformities and the non-standardness of MRI intensities, seeds are specified for the airway and its crucial background tissue components in only the 3D image corresponding to the first time instance of the 4D volume. Subsequently the process runs without human interaction and completes segmenting the whole 4D volume in 10 sec. Our evaluations indicate that the segmentations are of very good quality achieving true positive and false positive volume fractions and boundary distance with respect to reference manual segmentations of about 93%, 0.1%, and 0.5 mm, respectively.

Regional trabecular bone quality estimation for purposes of femoral bone strength prediction is important for improving the clinical assessment of osteoporotic fracture risk. In this study, we explore the ability of 3D Minkowski Functionals derived from multi-detector computed tomography (MDCT) images of proximal femur specimens in predicting their corresponding biomechanical strength. MDCT scans were acquired for 50 proximal femur specimens harvested from human cadavers. An automated volume of interest (VOI)-fitting algorithm was used to define a consistent volume in the femoral head of each specimen. In these VOIs, the trabecular bone micro-architecture was characterized by statistical moments of its BMD distribution and by topological features derived from Minkowski Functionals. A linear multiregression analysis and a support vector regression (SVR) algorithm with a linear kernel were used to predict the failure load (FL) from the feature sets; the predicted FL was compared to the true FL determined through biomechanical testing. The prediction performance was measured by the root mean square error (RMSE) for each feature set. The best prediction result was obtained from the Minkowski Functional surface used in combination with SVR, which had the lowest prediction error (RMSE = 0.939 ± 0.345) and which was significantly lower than mean BMD (RMSE = 1.075 ± 0.279, p<0.005). Our results indicate that the biomechanical strength prediction can be significantly improved in proximal femur specimens with Minkowski Functionals extracted from on MDCT images used in conjunction with support vector regression.

Current assessment of cartilage is primarily based on identification of indirect markers such as joint space narrowing and increased subchondral bone density on x-ray images. In this context, phase contrast CT imaging (PCI-CT) has recently emerged as a novel imaging technique that allows a direct examination of chondrocyte patterns and their correlation to osteoarthritis through visualization of cartilage soft tissue. This study investigates the use of topological and geometrical approaches for characterizing chondrocyte patterns in the radial zone of the knee cartilage matrix in the presence and absence of osteoarthritic damage. For this purpose, topological features derived from Minkowski Functionals and geometric features derived from the Scaling Index Method (SIM) were extracted from 842 regions of interest (ROI) annotated on PCI-CT images of healthy and osteoarthritic specimens of human patellar cartilage. The extracted features were then used in a machine learning task involving support vector regression to classify ROIs as healthy or osteoarthritic. Classification performance was evaluated using the area under the receiver operating characteristic (ROC) curve (AUC). The best classification performance was observed with high-dimensional geometrical feature vectors derived from SIM (0.95 ± 0.06) which outperformed all Minkowski Functionals (p < 0.001). These results suggest that such quantitative analysis of chondrocyte patterns in human patellar cartilage matrix involving SIM-derived geometrical features can distinguish between healthy and osteoarthritic tissue with high accuracy.

Trabecular bone and its microarchitecture are of prime importance for health. Studying vascularization helps to better know the relationship between bone and vascular microarchitecture. This research is an animal study (nine Lewis rats), based on the perfusion of vascularization by a contrast agent (a mixture of 50% barium sulfate with 1.5% of gelatin) before euthanasia. The samples were studied by micro CT at a resolution of 9μm. Softwares were used to show 3D volumes of bone and vessels, to calculate bone and vessels microarchitecture parameters. This study aims to understand simultaneously the bone microarchitecture and its vascular microarchitecture.

High-throughput micro-CT imaging has been used in our laboratory to evaluate fetal skeletal morphology in developmental toxicology studies. Currently, the volume-rendered skeletal images are visually inspected and observed abnormalities are reported for compounds in development. To improve the efficiency and reduce human error of the evaluation, we implemented a framework to automate the evaluation process. The framework starts by dividing the skull into regions of interest and then measuring various geometrical characteristics. Normal/abnormal classification on the bone segments is performed based on identifying statistical outliers. In pilot experiments using rabbit fetal skulls, the majority of the skeletal abnormalities can be detected successfully in this manner. However, there are shape-based abnormalities that are relatively subtle and thereby difficult to identify using the geometrical features. To address this problem, we introduced a model-based approach and applied this strategy on the squamosal bone. We will provide details on this active shape model (ASM) strategy for the identification of squamosal abnormalities and show that this method improved the sensitivity of detecting squamosal-related abnormalities from 0.48 to 0.92.

Bone mineral density (BMD) obtained from a CT image is an imaging biomarker used pre-clinically for characterizing the Rheumatoid arthritis (RA) phenotype. We use this biomarker in animal studies for evaluating disease progression and for testing various compounds. In the current setting, BMD measurements are obtained manually by selecting the regions of interest from three-dimensional (3-D) CT images of rat legs, which results in a laborious and low-throughput process. Combining image processing techniques, such as intensity thresholding and skeletonization, with mathematical techniques in curve fitting and curvature calculations, we developed an algorithm for quick, consistent, and automatic detection of joints in large CT data sets. The implemented algorithm has reduced analysis time for a study with 200 CT images from 10 days to 3 days and has improved the robust detection of the obtained regions of interest compared with manual segmentation. This algorithm has been used successfully in over 40 studies.

The segmentation of microscopic images is an important issue in biomedical image processing. Many works can be found in the literature; however, there is not a gold standard method that is able to provide good results for all kinds of microscopic images. Then, authors propose methods for a given kind of microscopic images. This paper deals with new segmentation framework based on evidence theory, called ESA (Evidential Segmentation Algorithm) to segment blood cell images. The proposed algorithm allows solving the segmentation problem of blood cell images. Herein, our goal is to extract the components of a given cell image by using evidence theory, that allows more flexibility to classify the pixels. The obtained results showed the efficiency of the proposed algorithm compared to other competing methods.

Magnetic Particle Imaging is capable of three-dimensional real-time imaging. Due to high spatial and temporal resolution, the method offers a great potential to be used in interventional scenarios. In this contribution, a design study integrating a single-sided coil assembly into a patient table is presented. An elliptical and an approximated elliptical coil topology are compared and proposed as alternatives to the commonly used circular shaped coils. Through this, the size of the field of view can be extended while not exceeding the lateral width of the patient table.

Nanoscale bubbles (nanobubbles) have been reported to improve contrast in tumor-targeted ultrasound imaging due to the enhanced permeation and retention effects at tumor vascular leaks. In this work, a self-made nanobubble ultrasound contrast agent was preliminarily characterized and evaluated in-vitro and in-vivo. Fundamental properties such as morphology appearance, size distribution, zeta potential, bubble concentration (bubble numbers per milliliter contrast agent suspension) and the stability of nanobubbles were assessed by light microscope and particle sizing analysis. Then the concentration intensity curve and time intensity curves (TICs) were acquired by ultrasound imaging experiment in-vitro. Finally, the contrast-enhanced ultrasonography was performed on rat to investigate the procedure of liver perfusion. The results showed that the nanobubbles had good shape and uniform distribution with the average diameter of 507.9 nm, polydispersity index (PDI) of 0.527, and zeta potential of -19.17 mV. Significant contrast enhancement was observed in in-vitro ultrasound imaging, demonstrating that the self-made nanobubbles can enhance the contrast effect of ultrasound imaging efficiently in-vitro. Slightly contrast enhancement was observed in in-vivo ultrasound imaging, indicating that the nanobubbles are not stable enough in-vivo. Future work will be focused on improving the ultrasonic imaging performance, stability, and antibody binding of the nanoscale ultrasound contrast agent.

Fluorescence techniques have been extensively employed to develop non-invasive methodologies for tracking and understanding complex biological processes both in vitro and in vivo, which is of high importance in modern life science research. Among a variety of fluorescent probes, inorganic semiconductor quantum dots (QDs) have shown advantages in terms of better photostability, larger Stokes shift and more feasible surface functionalization. However, their intrinsic toxic heavy metal components and unstable fluorescence at low pH greatly impede the applications of QDs in in vivo studies. In this work, we developed novel fluorescent probes that can outperform currently available QD based probes in practice. Using conjugated oligomer with aggregation-induced emission characteristics as the fluorescent domain and biocompatible lipid-PEG derivatives as the encapsulation matrix, the obtained organic dots have shown higher brightness, better stability in biological medium and comparable size and photostability as compared to their counterparts of inorganic QDs. More importantly, unlike QD-based probes, the organic fluorescent dots do not blink, and also do not contain heavy metal ions that could be potentially toxic when applied for living biosubstrates. Upon surface functionalization with a cell-penetrating peptide, the organic dots greatly outperform inorganic quantum dots in both in vitro and in vivo long-term cell tracing studies, which will be beneficial to answer crucial questions in stem cell/immune cell therapies. Considering the customized fluorescent properties and surface functionalities of the organic dots, a series of biocompatible organic dots will be developed to serve as a promising platform for multifarious bioimaging tasks in future.

There is ongoing interest in extending CT from anatomical to functional imaging. Recent successes with dual energy CT, the introduction of energy discriminating x-ray detectors, and novel, target-specific, nanoparticle contrast agents enable functional imaging capabilities via spectral CT. However, many challenges related to radiation dose, photon flux, and sensitivity still must be overcome. Here, we introduce a post-reconstruction algorithm called spectral diffusion that performs a robust material decomposition of spectral CT data in the presence of photon noise to address these challenges. Specifically, we use spectrally joint, piece-wise constant kernel regression and the split Bregman method to iteratively solve for a material decomposition which is gradient sparse, quantitatively accurate, and minimally biased relative to the source data. Spectral diffusion integrates structural information from multiple spectral channels and their corresponding material decompositions within the framework of diffusion-like denoising algorithms. Using a 3D, digital bar phantom and a material sensitivity matrix calibrated for use with a polychromatic x-ray source, we quantify the limits of detectability (CNR = 5) afforded by spectral diffusion in the triple-energy material decomposition of iodine (3.1 mg/mL), gold (0.9 mg/mL), and gadolinium (2.9 mg/mL) concentrations.

In magnetic resonance elastography (MRE), displacement fields from shear waves are inverted to estimate underlying material properties. Modulus differences detected by MRE may be used to distinguish tumors or other localized pathology in tissue. The accuracy of modulus estimates depends on the choice of the assumed constitutive model, as well as on the inversion algorithm, image resolution, and signal-to-noise ratio. In particular, in simpler inversion methods such as direct inversion and three-dimensional local frequency estimation (3D-LFE) the constitutive model is minimal (linear, elastic or viscoelastic, and isotropic) and the simplifying assumption of local homogeneity is usually made. The assumption of local homogeneity is often inaccurate [1], since the shear wavelength is typically comparable to the size of the structures of interest. Notably, the residual error (in direct inversion) between the model and the experimental data increases sharply at the boundaries of inclusions, while the “certainty” of the 3D-LFE estimate decreases. These error metrics may be used to detect local stiffness heterogeneity, as well as indicate variations in appropriate constitutive models. The utility of model uncertainty is demonstrated in simulations and with MRE data from a heterogeneous gel phantom.

There is currently no reliable method for early characterization of breast cancer response to neoadjuvant chemotherapy (NAC) [1,2]. Given that disruption of normal structural architecture occurs in cancer-bearing tissue, we hypothesize that further structural changes occur in response to NAC. Consequently, we are investigating the use of modalityindependent elastography (MIE) [3-8] as a method for monitoring mechanical integrity to predict long term outcomes in NAC. Recently, we have utilized a Demons non-rigid image registration method that allows 3D elasticity reconstruction in abnormal tissue geometries, making it particularly amenable to the evaluation of breast cancer mechanical properties. While past work has reflected relative elasticity contrast ratios [3], this study improves upon that work by utilizing a known stiffness reference material within the reconstruction framework such that a stiffness map becomes an absolute measure. To test, a polyvinyl alcohol (PVA) cryogel phantom and a silicone rubber mock mouse tumor phantom were constructed with varying mechanical stiffness. Results showed that an absolute measure of stiffness could be obtained based on a reference value. This reference technique demonstrates the ability to generate accurate measurements of absolute stiffness to characterize response to NAC. These results support that ‘referenced MIE' has the potential to reliably differentiate absolute tumor stiffness with significant contrast from that of surrounding tissue. The use of referenced MIE to obtain absolute quantification of biomarkers is also translatable across length scales such that the characterization method is mechanics-consistent at the small animal and human application.

Breast conserving therapy (BCT) of breast cancer is now widely accepted due to improved cosmetic outcome and improved patients’ quality of life. One of the critical issues in performing breast-conserving surgery is trying to achieve microscopically clear surgical margins while maintaining excellent cosmesis. Unfortunately, unacceptably close or positive surgical margins occur in at least 20-25% of all patients undergoing BCT requiring repeat surgical excision days or weeks later, as permanent histopathology routinely takes days to complete. Our aim is to develop a better method for intraoperative imaging of non-palpable breast malignancies excised by wire or needle localization. Providing non-deformed three dimensional imaging of the excised breast tissue should allow more accurate assessment of tumor margins and consequently allow further excision at the time of initial surgery thus limiting the enormous financial and emotional burden of additional surgery. We have designed and constructed a device that allows preservation of the excised breast tissue in its natural anatomic position relative to the breast as it is imaged to assess adequate excision. We performed initial tests with needle-guided lumpectomy specimens using micro-CT and digital breast tomosynthesis (DBT). Our device consists of a plastic sphere inside a cylindrical holder. The surgeon inserts a freshly excised piece of breast tissue into the sphere and matches its anatomic orientation with the fiducial markers on the sphere. A custom-shaped foam is placed inside the sphere to prevent specimen deformation due to gravity. DBT followed by micro-CT images of the specimen were obtained. We confirmed that our device preserved spatial orientation of the excised breast tissue and that the location error was lower than 10mm and 10 degrees. The initial obtained results indicate that breast lesions containing microcalcifications allow a good 3D imaging of margins providing immediate intraoperative feedback for further excision as needed at the initial operation.

Mammography is one of the most important tools for the early detection of breast cancer typically through detection of characteristic masses and/or micro calcifications. Digital mammography has become commonplace in recent years. High quality mammogram images are large in size, providing high-resolution data. Estimates of the false negative rate for cancers in mammography are approximately 10%–30%. This may be due to observation error, but more frequently it is because the cancer is hidden by other dense tissue in the breast and even after retrospective review of the mammogram, cannot be seen. In this study, we report on the results of novel image processing algorithms that will enhance the images providing decision support to reading physicians. Techniques such as Butterworth high pass filtering and Gabor filters will be applied to enhance images; followed by segmentation of the region of interest (ROI). Subsequently, the textural features will be extracted from the ROI, which will be used to classify the ROIs as either masses or non-masses. Among the statistical methods most used for the characterization of textures, the co-occurrence matrix makes it possible to determine the frequency of appearance of two pixels separated by a distance, at an angle from the horizontal. This matrix contains a very large amount of information that is complex. Therefore, it is not used directly but through measurements known as indices of texture such as average, variance, energy, contrast, correlation, normalized correlation and entropy.

Clinical observations have long suggested that cancer progression is accompanied by extracellular matrix remodeling and concomitant increases in mechanical stiffness. Due to the strong association of mechanics and tumor progression, there has been considerable interest in incorporating methodologies to diagnose cancer through the use of mechanical stiffness imaging biomarkers, resulting in commercially available US and MR elastography products. Extension of this approach towards monitoring longitudinal changes in mechanical properties along a course of cancer therapy may provide means for assessing early response to therapy; therefore a systematic study of the elasticity biomarker in characterizing cancer for therapeutic monitoring is needed. The elastography method we employ, modality independent elastography (MIE), can be described as a model-based inverse image-analysis method that reconstructs elasticity images using two acquired image volumes in a pre/post state of compression. In this work, we present preliminary data towards validation and reproducibility assessment of our elasticity biomarker in a pre-clinical model of breast cancer. The goal of this study is to determine the accuracy and reproducibility of MIE and therefore the magnitude of changes required to determine statistical differences during therapy. Our preliminary results suggest that the MIE method can accurately and robustly assess mechanical properties in a pre-clinical system and provide considerable enthusiasm for the extension of this technique towards monitoring therapy-induced changes to breast cancer tissue architecture.

We present results from a recent study involving the Imagio^TM breast imaging system, which produces fused real-time two-dimensional color-coded opto-acoustic (OA) images that are co-registered and temporally inter- leaved with real-time gray scale ultrasound using a specialized duplex handheld probe. The use of dual optical wavelengths provides functional blood map images of breast tissue and tumors displayed with high contrast based on total hemoglobin and oxygen saturation of the blood. This provides functional diagnostic information pertaining to tumor metabolism. OA also shows morphologic information about tumor neo-vascularity that is complementary to the morphological information obtained with conventional gray scale ultrasound. This fusion technology conveniently enables real-time analysis of the functional opto-acoustic features of lesions detected by readers familiar with anatomical gray scale ultrasound. We demonstrate co-registered opto-acoustic and ultrasonic images of malignant and benign tumors from a recent clinical study that provide new insight into the function of tumors in-vivo. Results from the Feasibility Study show preliminary evidence that the technology may have the capability to improve characterization of benign and malignant breast masses over conventional diagnostic breast ultrasound alone and to improve overall accuracy of breast mass diagnosis. In particular, OA improved speci city over that of conventional diagnostic ultrasound, which could potentially reduce the number of negative biopsies performed without missing cancers.

The choroid is a densely layer under the retinal pigment epithelium (RPE). Its deeper boundary is formed by the sclera, the outer fibrous shell of the eye. However, the inhomogeneity within the layers of choroidal Optical Coherence Tomography (OCT)-tomograms presents a significant challenge to existing segmentation algorithms. In this paper, we performed a statistical study of retinal OCT data to extract the choroid. This model fits a Gaussian mixture model (GMM) to image intensities with Expectation Maximization (EM) algorithm. The goodness of fit for proposed GMM model is computed using Chi-square measure and is obtained lower than 0.04 for our dataset. After fitting GMM model on OCT data, Bayesian classification method is employed for segmentation of the upper and lower border of boundary of retinal choroid. Our simulations show the signed and unsigned error of -1.44 +/- 0.5 and 1.6 +/- 0.53 for upper border, and -5.7 +/- 13.76 and 6.3 +/- 13.4 for lower border, respectively.

In clinical settings, application of the most recent modelling techniques is usually unfeasible due to the limited acquisition time. Localised acquisitions enclosing only the object of interest by reducing the field-of-view (FOV) counteract the time limitation but pose new challenges to the subsequent processing steps like motion correction. We use datasets from the Human Connectome Project (HCP) to simulate head motion distorted reduced FOV acquisitions and present an evaluation of head motion correction approaches: the commonly used affine regis- tration onto an unweighted reference image guided by the mutual information (MI) metric and a model-based approach, which uses reference images computed from approximated tensor data to improve the performance of the MI metric. While the standard approach using the MI metric yields up to 15% outliers (error>5 mm) and a mean spatial error above 1.5 mm, the model-based approach reduces the number of outliers (1%) and the spatial error significantly (p<;0.01). The behavior is also reflected by the visual analysis of the MI metric. The evaluation shows that the MI metric is of very limited use for reduced FOV data post-processing. The model-based approach has proven more suitable in this context.

This paper presents a novel fully automatic unsupervised framework for the segmentation of brain tissues in magnetic resonance (MR) images. The framework is a combination of our proposed Bayesian-based adaptive mean shift (BAMS), a priori spatial tissue probability maps and fuzzy c-means. BAMS is applied to cluster the tissues in the joint spatialintensity feature space and then a fuzzy c-means algorithm is employed with initialization by a priori spatial tissue probability maps to assign the clusters into three tissue types; white matter (WM), gray matter (GM) and cerebrospinal fluid (CSF). The proposed framework is validated on multimodal synthetic as well as on real T1-weighted MR data with varying noise characteristics and spatial intensity inhomogeneity. The performance of the proposed framework is evaluated relative to our previous method BAMS and other existing adaptive mean shift framework. Both of these are based on the mode pruning and voxel weighted k-means algorithm for classifying the clusters into WM, GM and CSF tissue. The experimental results demonstrate the robustness of the proposed framework to noise and spatial intensity inhomogeneity, and that it exhibits a higher degree of segmentation accuracy in segmenting both synthetic and real MR data compared to competing methods.

The acute ischemic stroke is a leading cause for death and disability in the industry nations. In case of a present acute ischemic stroke, the prediction of the future tissue outcome is of high interest for the clinicians as it can be used to support therapy decision making. Within this context, it has already been shown that the voxel-wise multi-parametric tissue outcome prediction leads to more promising results compared to single channel perfusion map thresholding. Most previously published multi-parametric predictions employ information from perfusion maps derived from perfusion-weighted MRI together with other image sequences such as diffusion-weighted MRI. However, it remains unclear if the typically calculated perfusion maps used for this purpose really include all valuable information from the PWI dataset for an optimal tissue outcome prediction. To investigate this problem in more detail, two different methods to predict tissue outcome using a k-nearest-neighbor approach were developed in this work and evaluated based on 18 datasets of acute stroke patients with known tissue outcome. The first method integrates apparent diffusion coefficient and perfusion parameter (Tmax, MTT, CBV, CBF) information for the voxel-wise prediction, while the second method employs also apparent diffusion coefficient information but the complete perfusion information in terms of the voxel-wise residue functions instead of the perfusion parameter maps for the voxel-wise prediction. Overall, the comparison of the results of the two prediction methods for the 18 patients using a leave-one-out cross validation revealed no considerable differences. Quantitatively, the parameter-based prediction of tissue outcome led to a mean Dice coefficient of 0.474, while the prediction using the residue functions led to a mean Dice coefficient of 0.461. Thus, it may be concluded from the results of this study that the perfusion parameter maps typically derived from PWI datasets include all valuable perfusion information required for a voxel-based tissue outcome prediction, while the complete analysis of the residue functions does not add further benefits for the voxel-wise tissue outcome prediction and is also computationally more expensive.

Age-related cortical thinning has been studied by many researchers using quantitative MR images for the past three decades and vastly differing results have been reported. Although results have shown age-related cortical thickening in elderly cohort statistically in some brain regions under certain conditions, cortical thinning in elderly cohort requires further systematic investigation. This paper leverages our previously reported brain surface intensity model (BSIM)¹ based technique to measure cortical thickness to study cortical changes due to normal aging. We measured cortical thickness of cognitively normal persons from 60 to 89 years old using Australian Imaging Biomarkers and Lifestyle Study (AIBL) data. MRI brains of 56 healthy people including 29 women and 27 men were selected. We measured average cortical thickness of each individual in eight brain regions: parietal, frontal, temporal, occipital, visual, sensory motor, medial frontal and medial parietal. Unlike the previous published studies, our results showed consistent age-related thinning of cerebral cortex in all brain regions. The parietal, medial frontal and medial parietal showed fastest thinning rates of 0.14, 0.12 and 0.10 mm/decade respectively while the visual region showed the slowest thinning rate of 0.05 mm/decade. In sensorimotor and parietal areas, women showed higher thinning (0.09 and 0.16 mm/decade) than men while in all other regions men showed higher thinning than women. We also created high resolution cortical thinning rate maps of the cohort and compared them to typical patterns of PET metabolic reduction of moderate AD and frontotemporal dementia (FTD). The results seemed to indicate vulnerable areas of cortical deterioration that may lead to brain dementia. These results validate our cortical thickness measurement technique by demonstrating the consistency of the cortical thinning and prediction of cortical deterioration trend with AIBL database.

This study examines the spatial distribution of microhemorrhages defined using susceptibility weighted images (SWI) in 46 patients with Traumatic Brain Injury (TBI) and applying region of interest (ROI) analysis using a brain atlas. SWI and 3D T1-weighted images were acquired on a 3T clinical Siemens scanner. A neuroradiologist reviewed all SWI images and manually labeled all identified microhemorrhages. To characterize the spatial distribution of microhemorrhages in standard Montreal Neurological Institute (MNI) space, the T1-weighted images were nonlinearly registered to the MNI template. This transformation was then applied to the co-registered SWI images and to the microhemorrhage coordinates. The frequencies of microhemorrhages were determined in major structures from ROIs defined in the digital Talairach brain atlas and in white matter tracts defined using a diffusion tensor imaging atlas. A total of 629 microhemorrhages were found with an average of 22±42 (range=1-179) in the 24 positive TBI patients. Microhemorrhages mostly congregated around the periphery of the brain and were fairly symmetrically distributed, although a number were found in the corpus callosum. From Talairach ROI analysis, microhemorrhages were most prevalent in the frontal lobes (65.1%). Restricting the analysis to WM tracts, microhemorrhages were primarily found in the corpus callosum (56.9%).

The analysis of large ensembles of time series is a fundamental challenge in different domains of biomedical image processing applications, specifically in the area of functional MRI data processing. An important aspect of such analysis is the ability to reconstruct community network structures based on interactive behavior between different nodes of the network which are captured in such time series. In this study, we start with a previously proposed novel approach that applies the linear Granger Causality concept to very high-dimensional time series. This approach is based on integrating dimensionality reduction into a multivariate time series model. If residuals of dimensionality reduced models can be transformed back into the original space, prediction errors in the high–dimensional space may be computed, and a large scale Granger Causality Index (lsGCI) is properly defined. The primary goal of this study was then to present an approach for recovering network structure from such lsGCI interactions through the application of pair-wise clustering. We specifically focus on a clustering approach based on topographic mapping of proximity data (TMP) for this purpose. We demonstrate our approach with a simulated network composed of five pair-wise different internal networks with varying strengths of community structure (based on the number of inter-network vertices). Our results suggest that such pair-wise clustering with TMP is capable of reconstructing the structure of the original network from lsGCI matrices that record the interactions between different nodes of the network when there is sufficient disparity between the intra- and inter-network vertices.

In this paper, we propose a method to automatically segment the corticospinal tract (CST) in diffusion tensor images (DTIs) by incorporating the anatomical features from multi-modality images generated in DTI using multiple DTI atlases. The to-be-segmented test subject, and each atlas, is comprised of images with different modalities – the mean diffusivity, the fractional anisotropy, and the images representing the three elements of the primary eigenvector. Each atlas had a paired image containing the manually delineated segmentations of the three regions of interest - the left and right CST and the background surrounding the CST. We solve the problem via maximum a posteriori estimation using generative models. Each modality image is modeled as a conditional Gaussian mixture random field, conditioned on the atlas-label pair and the local change of coordinates for each label. The expectation-maximization algorithm is used to alternatively estimate the local optimal diffeomorphisms for each label and the maximizing segmentations. The algorithm is evaluated on six subjects with a wide range of pathology. We compare the proposed method with two state-of-the-art multi-atlas based label fusion methods, against which the method displayed a high level of accuracy.

The sequential application of whole brain radiotherapy (WBRT) and more targeted stereotactic radiosurgery (SRS) is frequently used to treat metastatic brain tumors. However, SRS has side effects related to necrosis and edema, and requires separate and relatively invasive localization procedures. Helical tomotherapy (HT) allows for a SRS-type simultaneous infield boost (SIB) of multiple brain metastases, synchronously with WBRT and without separate stereotactic procedures. However, some patients’ tumors may not respond to HT+SIB, and would be more appropriately treated with radiosurgery or conventional surgery despite the additional risks and side effects. As a first step toward a broader objective of developing a means for response prediction to HT+SIB, the goal of this study was to investigate whether quantitative measurements of tumor size and appearance (including first- and second-order texture features) on a magnetic resonance imaging (MRI) scan acquired prior to treatment could be used to differentiate responder and nonresponder patient groups after HT+SIB treatment of metastatic disease of the brain. Our results demonstrated that smaller lesions may respond better to this form of therapy; measures of appearance provided limited added value over measures of size for response prediction. With further validation on a larger data set, this approach may lead to a means for prediction of individual patient response based on pre-treatment MRI, supporting appropriate therapy selection for patients with metastatic brain cancer.

New imaging technologies enable one to assess T-cell proliferation, an important feature of the immunological response. However, none of the traditional imaging modalities allow one to examine quantiatively T-cell function with microscopic resolution and single cell sensitivity over an entire mouse. To address this need, we established T-cells proliferation assays using 3D microscopic cryo-imaging. Assays include: (1) biodistribution of T-cells, (2) secondary lymphoid organ (SLO) volume measurement, (3) carboxyfluorescein succinimidyl ester (CFSE) dilution per cell as cells divide. To demonstrate the application, a graft-versus-host-disease (GVHD) model was used. 3D visualization show that T-cells specifically homed to the SLOs (spleen and lymph nodes) as well as GVHD target organs (such as GI-tract, liver, skin and thymus).The spleen was chosen as representative of the SLOs. For spleen size analysis, volumes of red and white pulp were measured. Spleen volumes of the allogeneic mice (with GVHD) were significantly larger than those of the syngeneic mice (without GVHD) at 72 to 120 hours post-transplant. For CFSE dilution approach, we employed color-coded volume rendering and probability density function (PDF) of single cell intensity to assess T-cell proliferation in the spleen. As compared to syngeneic T-cells, the allogeneic T-cells quickly aggregated in the spleen as indicated by increasing of CFSE signal over the first 48 hours. Then they rapidly proliferated as evidenced by reduced CFSE intensity (at 48-96 hours). Results suggest that assays can be used to study GVHD treatments using T-cell proliferation and biodistibution as assays. In summary, this is the first time that we are able to track and visualize T-cells in whole mouse with single cell sensitivity. We believe that our technique can be an alternative choice to traditional in vitro immunological proliferation assays by providing assessment of proliferation in an in vivo model.

Retinal 3D Optical coherence tomography (OCT) is a non-invasive imaging modality in ocular diseases. Due to large volumes of OCT data, it is better to utilize automatic extraction of information from OCT images, such as total retinal thickness and retinal nerve fiber layer thickness (RNFLT). These two thickness values have become useful indices to indicate the progress of diseases like glaucoma, according to the asymmetry between two eyes of an individual. Furthermore, the loss of ganglion cells may not be diagnosable by other tests and even not be evaluated when we only consider the thickness of one eye (due to dramatic different thickness among individuals). This can justify our need to have a comparison between thicknesses of two eyes in symmetricity. Therefore, we have proposed an asymmetry analysis of the retinal nerve layer thickness and total retinal thickness around the macula in the normal Iranian population. In the first step retinal borders are segmented by diffusion map method and thickness profiles were made. Then we found the middle point of the macula by pattern matching scheme. RNFLT and retinal thickness are analyzed in 9 sectors and the mean and standard deviation of each sector in the right and left eye are obtained. The maximums of the average RNFL thickness in right and left eyes are seen in the perifoveal nasal, and the minimums are seen in the fovea. Tolerance limits in RNFL thickness is shown to be between 0.78 to 2.4 μm for 19 volunteers used in this study.

Cardiac ischemic injuries can be classified into two main categories: reversible and irreversible. Treatment of reversible damages is possible through revascularization therapies. Clinically, it is quite vital to determine the reversibility of ischemic injuries and local efficiency using accurate diagnostics techniques. For this purpose, a number of imaging techniques have been developed. To our knowledge, while some of these techniques are capable of assessing tissue viability which is believed to be correlated with ischemic injuries reversibility, none of them are capable of providing information about local myocardial tissue efficiency. Note that this efficiency indicates the local tissue contribution to the overall (global) heart mechanical function which is characterized by parameters such as ejection fraction. While contraction force generation of the myocardium is a reliable and straightforward mechanical measure for the local myocardium functionality, it is also hypothesized that the level of damage reversibility expected from therapy is proportional to the intensity and distribution of these forces. As such this research involves developing a new imaging technique for cardiac contraction force quantification. This work is also geared towards another application, namely Cardiac Resynchronization Therapy (CRT), specifically for electrode leads configuration optimization. The latter has not been tackled through a systematic technique thus far. In the proposed method, contraction force reconstruction is accomplished by an inverse problem algorithm solved through an optimization framework which uses forward mechanical modelling of the myocardium iteratively to obtain the contraction forces field. As a result, the method requires a forward mechanical model of the myocardium which is computationally efficient and robust against divergence. Therefore, we developed such a model which considers all aspects of the myocardial mechanics including hyperelasticity, anisotropy, and active contraction forces of the fibers. This model assumes two major parts for the myocardium consisting background tissue and reinforcement bars simulating myocardial fibers. The finite element simulations of this model demonstrated reasonably good performance in mimicking left ventricle (LV) contractile function.

Since mechanical ventilation can cause harm to lung tissue it should be as protective as possible. Whereas numerous options exist to set ventilator parameters, an adequate monitoring is lacking up to date. The Electrical Impedance Tomography (EIT) provides a non-invasive visualization of ventilation which is relatively easy to apply and commercially available. Although there are a number of published measures and parameters derived from EIT, it is not clear how to use EIT to improve clinical outcome of e.g. patients suffering from acute respiratory distress syndrome (ARDS), a severe disease with a high mortality rate. On the one hand, parameters should be easy to obtain, on the other hand clinical algorithms should consider them to optimize ventilator settings. The so called Global inhomogeneity (GI) index bases on the fact that ARDS is characterized by an inhomogeneous injury pattern. By applying positive endexpiratory pressures (PEEP), homogeneity should be attained. In this study, ARDS was induced by a double hit procedure in six pigs. They were randomly assigned to either the EIT or the control group. Whereas in the control group the ARDS network table was used to set the PEEP according to the current inspiratory oxygen fraction, in the EIT group the GI index was calculated during a decremental PEEP trial. PEEP was kept when GI index was lowest. Interestingly, PEEP was significantly higher in the EIT group. Additionally, two of these animals died ahead of the schedule. Obviously, not only homogeneity of ventilation distribution matters but also limitation of over-distension.

Accurate measurement and analysis of 3D kinematics of temporomandibular joint (TMJ) is very important for assisting clinical diagnosis and treatment of prosthodontics and orthodontics, and oral surgery. This study presents a new 3D kinematic measurement technique of the TMJ using X-ray fluoroscopic images, which can easily obtain the TMJ kinematic data in natural motion. In vivo kinematics of the TMJ (maxilla and mandibular bone) is determined using a feature-based 2D/3D registration, which uses beads silhouette on fluoroscopic images and 3D surface bone models with beads. The 3D surface models of maxilla and mandibular bone with beads were created from CT scans data of the subject using the mouthpiece with the seven strategically placed beads. In order to validate the accuracy of pose estimation for the maxilla and mandibular bone, computer simulation test was performed using five patterns of synthetic tantalum beads silhouette images. In the clinical applications, dynamic movement during jaw opening and closing was conducted, and the relative pose of the mandibular bone with respect to the maxilla bone was determined. The results of computer simulation test showed that the root mean square errors were sufficiently smaller than 1.0 mm and 1.0 degree. In the results of clinical application, during jaw opening from 0.0 to 36.8 degree of rotation, mandibular condyle exhibited 19.8 mm of anterior sliding relative to maxillary articular fossa, and these measurement values were clinically similar to the previous reports. Consequently, present technique was thought to be suitable for the 3D TMJ kinematic analysis.

The ability of Anisotropic Minkowski Functionals (AMFs) to capture local anisotropy while evaluating topological properties of the underlying gray-level structures has been previously demonstrated. We evaluate the ability of this approach to characterize local structure properties of trabecular bone micro-architecture in ex vivo proximal femur specimens, as visualized on multi-detector CT, for purposes of biomechanical bone strength prediction. To this end, volumetric AMFs were computed locally for each voxel of volumes of interest (VOI) extracted from the femoral head of 146 specimens. The local anisotropy captured by such AMFs was quantified using a fractional anisotropy measure; the magnitude and direction of anisotropy at every pixel was stored in histograms that served as a feature vectors that characterized the VOIs. A linear multi-regression analysis algorithm was used to predict the failure load (FL) from the feature sets; the predicted FL was compared to the true FL determined through biomechanical testing. The prediction performance was measured by the root mean square error (RMSE) for each feature set. The best prediction performance was obtained from the fractional anisotropy histogram of AMF Euler Characteristic (RMSE = 1.01 ± 0.13), which was significantly better than MDCT-derived mean BMD (RMSE = 1.12 ± 0.16, p<0.05). We conclude that such anisotropic Minkowski Functionals can capture valuable information regarding regional trabecular bone quality and contribute to improved bone strength prediction, which is important for improving the clinical assessment of osteoporotic fracture risk.

Cone beam computed tomography (CBCT) has found use in mammography for imaging the entire breast with sufficient spatial resolution at a radiation dose within the range of that of conventional mammography. Recently, enhancement of lesion tissue through the use of contrast agents has been proposed for cone beam CT. This study investigates whether the use of such contrast agents improves the ability of texture features to differentiate lesion texture from healthy tissue on CBCT in an automated manner. For this purpose, 9 lesions were annotated by an experienced radiologist on both regular and contrast-enhanced CBCT images using two-dimensional (2D) square ROIs. These lesions were then segmented, and each pixel within the lesion ROI was assigned a label – lesion or non-lesion, based on the segmentation mask. On both sets of CBCT images, four three-dimensional (3D) Minkowski Functionals were used to characterize the local topology at each pixel. The resulting feature vectors were then used in a machine learning task involving support vector regression with a linear kernel (SVRlin) to classify each pixel as belonging to the lesion or non-lesion region of the ROI. Classification performance was assessed using the area under the receiver-operating characteristic (ROC) curve (AUC). Minkowski Functionals derived from contrastenhanced CBCT images were found to exhibit significantly better performance at distinguishing between lesion and non-lesion areas within the ROI when compared to those extracted from CBCT images without contrast enhancement (p < 0.05). Thus, contrast enhancement in CBCT can improve the ability of texture features to distinguish lesions from surrounding healthy tissue.

Obesity is a global health problem, particularly in the U.S. where one third of adults are obese. A reliable and accurate method of quantifying obesity is necessary. Visceral adipose tissue (VAT) and subcutaneous adipose tissue (SAT) are two measures of obesity that reflect different associated health risks, but accurate measurements in humans or rodent models are difficult. In this paper we present an automatic, registration-based segmentation method for mouse adiposity studies using microCT images. We co-register the subject CT image and a mouse CT atlas. Our method is based on surface matching of the microCT image and an atlas. Surface-based elastic volume warping is used to match the internal anatomy. We acquired a whole body scan of a C57BL6/J mouse injected with contrast agent using microCT and created a whole body mouse atlas by manually delineate the boundaries of the mouse and major organs. For method verification we scanned a C57BL6/J mouse from the base of the skull to the distal tibia. We registered the obtained mouse CT image to our atlas. Preliminary results show that we can warp the atlas image to match the posture and shape of the subject CT image, which has significant differences from the atlas. We plan to use this software tool in longitudinal obesity studies using mouse models.

In this paper, the relationship between image intensity and ultrasound contrast agent (UCA) concentration is investigated. Experiments are conducted in water bath using a silicon tube filled with UCA (SonoVue) at different concentrations (100μl/l to 6000μl/l) at around 37 °C to simulate the temperature in human body. The mean gray-scale intensity within the region of interest (ROI) is calculated to obtain the plot of signal intensity to UCA concentration. The results show that the intensity firstly exhibits a linear increase to the peak at approximately 1500μl/l then appears a downward trend due to the multiple scattering (MS) effects.

Intravascular ultrasound (IVUS) is an important imaging technique that is used to study vascular wall architecture for diagnosis and assessment of the vascular diseases. Segmentation of lumen and media-adventitia boundaries from IVUS images is a basic and necessary step for quantitative assessment of the vascular walls. Due to ultrasound speckles, artifacts and individual differences, automated segmentation of IVUS images represents a challenging task. In this paper, a random walk based method is proposed for fully automated segmentation of IVUS images. Robust and accurate determination of the seed points for different regions is the key to successful use of the random walk algorithm in segmentation of IVUS images and is the focus of our work. The presented method mainly comprises five steps: firstly, the seed points inside the lumen and outside the adventitia are roughly estimated with intensity information, respectively; secondly, the seed points outside the adventitia are refined, and those of the media are determined through the results of applying random walk to the IVUS image with the roughly estimated seed points; thirdly, the media-adventitia boundary is detected by using random walk with the seed points obtained in the second step and the image gradient; fourthly, the seed points for media and lumen are refined; finally, the lumen boundary is extracted by using random walk again with the seed points obtained in the fourth step and the image gradient. The tests of the proposed algorithm on the in vivo dataset demonstrate the effectiveness of the presented IVUS image segmentation approach.