This paper presents a new approach for accurate extraction of tissue deformation imaged with tagged MR. Our method, based on banks of Gabor filters, adjusts (1) the aspect and (2) orientation of the filter’s envelope and adjusts (3) the radial frequency and (4) angle of the filter’s sinusoidal grating to extract information about the deformation of tissue. The method accurately extracts tag line spacing, orientation, displacement and effective contrast. Existing, non-adaptive methods often fail to recover useful displacement information in the proximity of tissue boundaries while our method works in the proximity of the boundaries. We also present an interpolation method to recover all tag information at a finer resolution than the filter bank parameters. Results are shown on simulated images of translating and contracting tissue.

XMR systems are a new type of interventional facility in which patients can be rapidly transferred between x-ray and MR systems on a floating table. We have previously developed a technique to register MR and x-ray images obtained from such systems. We are carrying out a program of XMR guided cardiac electrophysiology study (EPS) and radio frequency ablation (RFA). The aim of our work was to apply our registration technology to XMR guided EPS/RFA in order to integrate anatomical, electrophysiological and motion information. This would assist in guidance and allow us to validate and refine electromechanical models. Registration of the imaging modalities was achieved by a combination of system calibration and real-time optical tracking. Patients were initially imaged using MR imaging. An SSFP volume scan of the heart was acquired for anatomical information, followed by tagged scans for motion information. The patients were then transferred to the x-ray system. Tracked biplane x-ray images were acquired while electrical measurements were made from catheters placed in the heart. The relationship between the MR and x-ray images was determined. The MR volume scan of the heart was segmented and the tagged scans were analysed using a non-rigid registration algorithm to compute motion. The position of catheters was reconstructed within the MR cardiac anatomy. The anatomical, electrophysiological, and motion information were displayed in the same coordinate system. Simulations of electrical depolarisation and contraction were performed using electromechanical models of the myocardium. We present results for 2 initial cases. For patient 1, a contact mapping system was used for the EPS and for patient 2, a non-contact mapping system was used. Our XMR registration technique allows the integration of anatomical, electrophysiological, and motion information for patients undergoing EPS/RFA. This integrated approach has assisted in interventional guidance and has been used to validate electromechanical models of the myocardium.

We present current research in which left ventricular deformation is estimated from tagged cardiac magnetic resonance imaging using volumetric deformable models constructed from nonuniform rational B-splines (NURBS). From a set of short and long axis images at end-diastole, the initial NURBS model is constructed by fitting two surfaces with the same parameterization to the set of epicardial and endocardial contours from which a volumetric model is created. Using normal displacements of the three sets of orthogonal tag planes as well as displacements of both tag line and contour/tag line intersection points, one can solve for the optimal homogeneous coordinates, in a least squares sense, of the control points of the NURBS model at a later time point using quadratic programming. After fitting to all time points of data, lofting the NURBS model at each time point creates a comprehensive 4-D NURBS model. From this model, we can extract 3-D myocardial displacement fields and corresponding strain maps, which are local measures of non-rigid deformation.

Respiratory motion compensation for cardiac imaging requires knowledge of the heart's motion and deformation during breathing. We propose a method for measuring the natural tidal respiratory motion of the heart using free breathing coronary angiograms. A 3D deformation field describing the cardiac and respiratory motion of the coronary arteries is recovered from a biplane acquisition. Cardiac and respiratory phase are assigned to the images from an ECG signal synchronized to the image acquisition, and from the diaphragmatic displacement as observed in the images. The resulting motion field is decomposed into cardiac and respiratory components by fitting the field with periodic 2D parametric functions, where one dimension spans one cardiac cycle, and the second dimension spans one respiratory cycle. The method is applied to patient datasets, and an analysis of respiratory motion of the heart is presented.

The state-of-the-art multiple detector-row CT, which usually employs fan beam reconstruction algorithms by approximating a cone beam geometry into a fan beam geometry, has been well recognized as an important modality for cardiac imaging. At present, the multiple detector-row CT is evolving into volumetric CT, in which cone beam reconstruction algorithms are needed to combat cone beam artifacts caused by large cone angle. An ECG-gated cardiac cone beam reconstruction algorithm based upon the so-called semi-CB geometry is implemented in this study. To get the highest temporal resolution, only the projection data corresponding to 180° plus the cone angle are row-wise rebinned into the semi-CB geometry for three-dimensional reconstruction. Data extrapolation is utilized to extend the z-coverage of the ECG-gated cardiac cone beam reconstruction algorithm approaching the edge of a CT detector. A helical body phantom is used to evaluate the ECG-gated cone beam reconstruction algorithm’s z-coverage and capability of suppressing cone beam artifacts. Furthermore, two sets of cardiac data scanned by a multiple detector-row CT scanner at 16 x 1.25 (mm) and normalized pitch 0.275 and 0.3 respectively are used to evaluate the ECG-gated CB reconstruction algorithm’s imaging performance. As a reference, the images reconstructed by a fan beam reconstruction algorithm for multiple detector-row CT are also presented. The qualitative evaluation shows that, the ECG-gated cone beam reconstruction algorithm outperforms its fan beam counterpart from the perspective of cone beam artifact suppression and z-coverage while the temporal resolution is well maintained. Consequently, the scan speed can be increased to reduce the contrast agent amount and injection time, improve the patient comfort and x-ray dose efficiency. Based up on the comparison, it is believed that, with the transition of multiple detector-row CT into volumetric CT, ECG-gated cone beam reconstruction algorithms will provide better image quality for CT cardiac applications.

This work explored the feasibility of using Magnetic Resonance Elastography (MRE) technology to enable in vitro quantification of dynamic mechanical behavior of cartilage through its thickness. A customized system for MRE of cartilage was designed to include components for adequate generation and detection of high frequency mechanical shear waves within small and stiff materials. The system included components for mechanical excitation, motion encoding, and imaging of small samples. Limitations in sensitivity to motion encoding of high frequency propagating mechanical waves using a whole body coil (i.e. G_max = 2.2 G/cm) required the design of a local gradient coil system to achieve a gain in gradient strength of at least 5 times. The performance of the new system was tested using various cartilage-mimicking phantom materials. MRE of a stiff 5% agar gelatin phantom demonstrated gains in sensitivity to motion encoding of high frequency mechanical waves in cartilage like materials. MRE of fetal bovine cartilage samples yielded a distribution of shear stiffness within the thickness of the cartilage similar to values found in the literature, hence, suggesting the feasibility of using MRE to non-invasively and directly assess the dynamic mechanical properties of cartilage.

Strain Encoded Magnetic Resonance Imaging (SENC-MRI) is a new technique that allows real-time quantification of tissue deformation. The technique is based on initially modulating the magnetization of the imaged object with sinusoidal pattern (MR-tagging) in the z-direction (through-plane direction). Compression is then applied to the object resulting in a change of the frequency of the sinusoidal tagging depending, in part, on tissue stiffness; e.g. the softer the material the higher the resulting frequency. By determining the changes in frequency, regional deformations can be determined and quantified. In SENC MRI, this is achieved by acquiring several images (typically 8 images), each with different phase-encoding, which we call tunings, in the z-direction. For each tuning, the intensity of pixels whose tagging frequency coincides with the tuning frequency is higher than other pixels. Since the number of the acquired images is limited, only a limited range of frequencies can be covered and, hence, the accuracy of the estimates may be inefficient if the tuning are not selected carefully. However, in this paper, we show that deformation maps can be obtained with good accuracy from the limited number of tunings. In this regard, we propose three methods and compare between them for maximum achievable accuracy. The methods are 1) center-of-mass, 2) curve fitting, and 3) clustering-based method. The methods are applied to simulated data and MR images obtained from a gel phantom experiment. The results of comparisons shows that good estimates of deformation can be obtained even if the sampled data is distorted by noise or MR artifacts.

Osteoporosis affects an estimated 44 million Americans. This condition results from bone loss, but the measured change in bone mass does not fully account for the marked decrease in whole-bone structural integrity seen in osteoporosis. In order to study structural changes in bone mineral distribution due to normal ageing and osteoporosis, we have developed a method for progressive analysis of whole-bone mechanical integrity from helical CT images. The system provides rapid semi-automated alignment of femur and vertebrae volume images into standard anatomic reference planes, and calculates bone mineral density in any selected 3D sections of bone. Mineral density measures are obtained using both full-width-half-max contours and threshold-derived masks, and are obtained for cortical bone and trabecular bone separately. Biomechanical properties of the bone cross-section are also assessed, including the 2-D bending moment of the cortical bone region and the integrated flexural rigidity of the cortical region or whole-bone region in arbitrary planes. This method facilitates progressive refinement of the analysis protocol by separating the labor-intensive alignment and landmark selection process from the analysis process. As the analysis protocol evolves to include new measures, previously analyzed images can be automatically reanalyzed, using the image regions originally specified. Initial results show inverse correlation of indices of biomechanical bone strength with age, greater loss of bone strength in the lumbar spine than in the femoral neck, and more trabecular than cortical bone loss at both sites.

Regional lung ventilation can be measured via Xenon-enhanced computed tomography (Xe-CT) by determining washin (WI) and washout (WO) rates of stable Xe. It has been assumed that WI = WO, ignoring Xe solubility in blood and tissue and then other geometric isssues. We test this by measuring WO-WI in lung by Xe-CT. Also, we investigate the effect of tidal volume (TV) and end inspiratory (EI) vs end expiratory (EE) scan gating on WO and WI measurements. 3 anesthetized, supine sheep were scanned using multidetector-row computed tomography (MDCT). Imaging was gated to both EE and EI during a WI (33 breaths) and WO (20 breaths) maneuver using 55% Xe for WI and room air for WO. Time constants (TCs) of Xe WI and WO were obtained by exponential fitting. WO and WI TCs were compared: 1) apex and base 2) dependent, middle, and nondependent 3) EE and EI 4) three TVs. The vertical gradient of WO-WI showed WO > WI in dependent vs non-dependent regions. WO-WI in both dependent and nondependent region at the lung base and apex was larger when measured at EE compared to EI. As TV increases, the global WO-WI difference decreased. TV showed greater influence on WO than WI. Xe WO was longer than WI possibly reflecting Xe solubility in blood and tissue. Higher TVs and gating to EE provided greater effects on WO than WI TCs which may relate to the number of partial volumed conducting airways contributing to the regional voxel-based measures. We conclude that WO mode is more susceptible to errors caused by either xenon solubility or tidal volume than WI mode and EE scanning may more accurately reflect alveolar ventilation.

We are developing 4D-CT to provide breathing motion information (trajectories) for radiation therapy treatment planning of lung cancer. Potential applications include optimization of intensity-modulated beams in the presence of breathing motion and intra-fraction target volume margin determination for conformal therapy. The images are acquired using a multi-slice CT scanner while the patient undergoes simultaneous quantitative spirometry. At each couch position, the CT scanner is operated in ciné mode and acquires up to 15 scans of 12 slices each. Each CT scan is associated with the measured tidal volume for retrospective reconstruction of 3D CT scans at arbitrary tidal volumes. The specific tasks of this project involves the development of automated registration of internal organ motion (trajectories) during breathing. A modified least-squares based optical flow algorithm tracks specific features of interest by modifying the eigenvalues of gradient matrix (gradient structural tensor). Good correlations between the measured motion and spirometry-based tidal volume are observed and evidence of internal hysteresis is also detected.

This paper describes an algorithm for automated segmentation of pulmonary vessels from thoracic 3D CT images. The lung region is roughly extracted based on thresholding and labeling in order to reduce computational cost in the following filtering step. Vessels are enhanced by application of a line-filter, which is based on a combination of eigen values of a Hessian matrix to provide higher response to vessels compared with the other structures. Initial segmentation is performed by thresholding of the filter output. Since extracted vessels may contain tiny holes and local discontinuities between segments, especially around branchpoints, tracking algorithm is used to fill these gaps. Though the results may still contain not only vessels but also parts of airway walls and noise, such structures can be eliminated by considering the number of branchpoints associated with each structure since vascular trees are characterized as objects with many branchpoints. Therefore, a thinning algorithm is applied to determine the number of branchpoints and the final segmentation is obtained by thresholding with regard to the number of branchpoints. We applied the algorithm to five healthy human scans and obtained visually promising results. In order to evaluate our segmentation results quantitatively, approximately 2,000 manually identified points inside the vascular tree were selected in each case to check how many were correctly included in the segmentation result. On average, 98% of the manually identified vessel points were properly marked as vessels. This result demonstrates the promising performance of our algorithm and its utility for further analyses.

The accurate segmentation of the human airway tree from volumetric CT images builds an important corner stone in pulmonary image processing. It is the basis for many consecutive processing steps like branch-point labeling and matching, virtual bronchoscopy, and more. Previously reported airway tree segmentation methods often suffer from "leaking" into the surrounding lung tissue, caused by the anatomically thin airway wall combined with the occurrence of partial volume effect and noise. Another common problem with previously proposed airway segmentation algorithms is their difficulties with segmenting low dose scans and scans of heavily diseased lungs. We present a new airway tree segmentation method that works in 3D, avoids leaks, and automatically adapts to different types of scans without the need for the user to iteratively adjust any parameters.

Based on clinical CT-data a patient individual model of the bronchial-tree is constructed, incorporating information on irregular dichotomic branching of the airway bifurcations. Combining this geometric model with analytical models of transport and uptake of gas a patient individual outline of ventilation of the lung and the individual time course of inhalation is given. Conjunctions to multi-breath washout analysis are deduced. Central purpose of the presented model is to assess the significance of patient individual local airway geometry on the global ventilation of the lung and the resulting uptake of inhaled gases in the gas exchange regions of the lung.

In this report we present a brief outline of our technological approaches to developing a comprehensive imaging platform suitable for the investigation of the dynamics of the hemoglobin signal in large tissue structures using NIRS imaging techniques. Our approach includes a combined hardware and software development effort that provides for i) hardware integration, ii) system calibration, iii) data integrity checks, iv) image recovery, v) image enhancement and vi) signal processing. Presented are representative results obtained from human subjects that explore the sensitivity and other capabilities of the measuring system to detect focal hemodynamic responses in the head, breast and limb of volunteers. Results obtained support the contention that time-series NIRS imaging is a powerful and sensitive technique for exploring the hemodynamics of healthy and diseased tissues.

Bronchoscopic biopsy is often used for assisting the assessment of lung cancer. We have found in previous research that live image guidance of bronchoscopy has much potential for improving biopsy outcome. We have devised a system for this purpose. During a guided bronchoscopy procedure, our system simultaneously draws upon both the bronchoscope's video stream and the patient's 3D MDCT volume. The key data-processing step during guided bronchoscopy is the registration of the 3D MDCT data volume to the bronchoscopic video. The registration process is initialized by assuming that the bronchoscope is at a fixed viewpoint, giving a target reference video image, while the virtual-world camera inside the MDCT volume begins at an initial viewpoint that is within a reasonable vicinity of the bronchoscope's viewpoint. During registration, an optimization process searches for the optimal viewpoint to give the virtual image best matching the fixed video target. Overall, we have found that the CT-video registration technique operates robustly over a wide range of conditions, with considerable flexibility in the initial-viewpoint choice. Further, the system appears to be largely insensitive to the differences in lung capacity during the MDCT scan and during bronchoscopy. Finally, the system matches effectively in a wide range of anatomical circumstances.

This paper describes a new method for calculating image similarity between a real bronchoscopic (RB) image and a virtual endoscopic (VE) image for bronchoscope tracking based on image registration. Camera motion tracking is sequentially done by finding viewing parameters (camera position and orientation) that can render the most similar VE image to a currently processing RB frame based on image similarity, since it is difficult to attach a positional sensor at the tip of a bronchoscope. In the previous method, image similarity was calculated between real and virtual endoscopic images by summing gray-level differences up for all pixels of two images. This method could not estimate positions and orientations of a real bronchoscope camera properly, when image similarity changed only a little (but partly changed significantly) due to averaging of gray-level differences for the entire image. The proposed method divides the real and virtual endoscopic images into a set of subregions and selects the subregions that contain characteristic shapes such as the bifurcation and folding patterns of the bronchus. The proposed image similarity measure is implemented in the bronchoscope navigation system that equips the prediction function of the bronchoscope motion based on Kalman filtering. The predicted results are used as initial estimations of image registration. We applied the proposed method to eight pairs of bronchoscopic videos and three-dimensional (3-D) chest CT images. The experimental results showed that the proposed method improved the tracking performance by five orders of magnitude over the previous method. Computation time for one frame decreased to 20% of the previous method's.

A method for performing a virtual pneumoperitoneum based on shape deformation and its application to virtual laparoscopy are described. Laparoscopic surgery is now widely performed as minimum-invasive surgery. Because a laparoscope has a very narrow viewing area, this limits the surgeon's viewable area. A virtual endoscopy system that provides virtual laparoscopic images rendered at arbitrary viewpoints and view directions to a surgeon during surgery would be very helpful for intra-operative surgical navigation or pre-operative surgical planning. Because CT images are taken before a pneumoperitoneum in most cases, the abdominal wall needs to be lifted for generating virtual laparoscopic images. We deform original 3-D abdominal CT images so that the abdominal wall is virtually elevated. The entire process consists of five major steps: (a) extracting the abdominal wall, (b) elastic modeling, (c) elastic deformation of the model, (d) deformation of the original image, and (e) rendering virtual laparoscopic images. Virtual laparoscopic images are then generated from the deformed image. We have applied the method to three cases of 3-D abdominal CT images. From the experimental results, we confirmed that the abdominal wall was appropriately elevated by the proposed method. Laparoscopic views generated were very helpful for intra-operative surgical navigation as additional views of a surgeon or pre-operative surgical planning.

The development of pulmonary airway disease is characterized by mucosal color and topographical changes. Traditionally subjective visual interpretation of a bronchoscope procedure defines the identification of pulmonary airway disease however we have developed an optical imaging system used in conjunction with CT images to potentially quantify and classify these subtle variations. This paper presents a method for the construction of true color 3D images of the pulmonary airways from both optical and CT image data. Shape from Shading methods in the past decade have continually strived to achieve this goal by extracting 3D information from captured 2D images however these attempts have been severely limited in their application to bronchoscope images. Conversely the utilization of CT scans provides a sound tool for determining the gross structural anatomy of the airways however the accuracy of the rendered topographical surface maps is limited due to the resolution of the CT image data. Through integration of both the optical and CT imaging modalities we hope to create high resolution true color 3D images providing the necessary color and texture information to aid in future detection and classification of possible pulmonary airway disease. Preliminary combined color and texture results associated with various pulmonary airway diseases are presented highlighting the usefulness of this analysis technique.

We propose a new partial volume (PV) segmentation scheme to extract bladder wall for computer aided detection (CAD) of bladder lesions using multispectral MR images. Compared with CT images, MR images provide not only a better tissue contrast between bladder wall and bladder lumen, but also the multispectral information. As multispectral images are spatially registered over three-dimensional space, information extracted from them is more valuable than that extracted from each image individually. Furthermore, the intrinsic T₁ and T₂ contrast of the urine against the bladder wall eliminates the invasive air insufflation procedure. Because the earliest stages of bladder lesion growth tend to develop gradually and migrate slowly from the mucosa into the bladder wall, our proposed PV algorithm quantifies images as percentages of tissues inside each voxel. It preserves both morphology and texture information and provides tissue growth tendency in addition to the anatomical structure. Our CAD system utilizes a multi-scan protocol on dual (full and empty of urine) states of the bladder to extract both geometrical and texture information. Moreover, multi-scan of transverse and coronal MR images eliminates motion artifacts. Experimental results indicate that the presented scheme is feasible towards mass screening and lesion detection for virtual cystoscopy (VC).

Functional magnetic resonance imaging (fMRI) has become a powerful tool for investigating the working human brain based on the blood oxygenation level dependent (BOLD) effect on the MR signal. However, despite the widespread use of fMRI techniques for mapping brain activation, the basic physiological mechanisms underlying the observed signal changes are still poorly understood. Arterial spin labeling (ASL) techniques, which measure cerebral blood flow (CBF) and the BOLD effect simultaneously, provide a useful tool for investigating these physiological questions. In this paper, recent results of studies manipulating the baseline CBF both pharmacologically and physiologically will be discussed. These data are consistent with a feed-forward mechanism of neurovascular coupling, and suggest that the CBF change itself may be a more robust reflection of neural activity changes than the BOLD effect. Consistent with these data, a new thermodynamic hypothesis is proposed for the physiological function of CBF regulation: maintenance of the [O₂]/[CO₂] concentration ratio at the mitochondria in order to preserve the free energy available from oxidative metabolism. A kinetic model based on this hypothesis provides a reasonable quantitative description of the CBF changes associated with neural activity and altered blood gases (CO₂ and O₂).

This study investigated the effects of 30% oxygen administration on the visuospatial cognitive performance using fMRI. Eight college students (right-handed, average age 23.5) were selected as subjects for this study. Oxygen supply equipment which gives 21% and 30% oxygen at a constant rate of 8L/min was developed for this study. To measure the performance of visuospatial cognition, two questionnaires with similar difficulty containing 20 questions each were also developed. Experiment was designed as two runs: run for visuospatial cognition test with normal air (21% of oxygen) and run for visuospatial cognition test with highly concentrated air (30% of oxygen). Run consists of 4 blocks and each block has 8 control problems and 5 visuospatial problems. Functional brain images were taken from 3T MRI using single-shot EPI method. Activities of neural network due to performing visuospatial cognition test were identified using subtraction procedure, and activation areas while performing visuospatial cognition test were extracted using double subtraction procedure. Activities were observed at occipital lobe, parietal lobe, and frontal lobe when performing visuospatial cognition test following both 21% and 30% oxygen administration. But in case of only 30% oxygen administration there were more activities at left precuneus, left cuneus, right postcentral gyrus, bilateral middle frontal gyri, right inferior frontal gyrus, left superior frontal gyrus, bilateral uvula, bilateral pyramis, and nodule compared with 21% oxygen administration. From results of visuospatial cognition test, accuracy rate increased in case of 30% oxygen administration. Thus it could be concluded that highly concentrated oxygen administration has positive effects on the visuospatial cognitive performance.

Functional magnetic resonance imaging (fMRI) is a technique that is sensitive to correlates of neuronal activity. The application of fMRI to measure functional connectivity of related brain regions across hemispheres (e.g. left and right motor cortices) has great potential for revealing fundamental physiological brain processes. Primarily, functional connectivity has been characterized by linear correlations in resting-state data, which may not provide a complete description of its temporal properties. In this work, we broaden the measure of functional connectivity to study not only linear correlations, but also those arising from deterministic, non-linear dynamics. Here the delta-epsilon approach is extended and applied to fMRI time series. The method of delays is used to reconstruct the joint system defined by a reference pixel and a candidate pixel. The crux of this technique relies on determining whether the candidate pixel provides additional information concerning the time evolution of the reference. As in many correlation-based connectivity studies, we fix the reference pixel. Every brain location is then used as a candidate pixel to estimate the spatial pattern of deterministic coupling with the reference. Our results indicate that measured connectivity is often emphasized in the motor cortex contra-lateral to the reference pixel, demonstrating the suitability of this approach for functional connectivity studies. In addition, discrepancies with traditional correlation analysis provide initial evidence for non-linear dynamical properties of resting-state fMRI data. Consequently, the non-linear characterization provided from our approach may provide a more complete description of the underlying physiology and brain function measured by this type of data.

We have used MR DTI to identify relevant brain structures involved in visuospatial processing, in an attempt to link perceptual and attentional impairments to WM changes in Alzheimer's disease (AD) patients. Correlation of DTI measured parameters with results of several neuropsychological tests will be reported here. Several issues related to quantitation of DTI parameters in ROI analysis are addressed. In spite of only a small number of subjects were studied so far, we found not only that AD patients showed significant decrease of white matter (WM) integrity in corpus callosum (CC), most prominent at the posterior portion, but also found significant correlations between the DTI parameters and scores from several neuropsychological tests. Our preliminary results suggest that DTI help to improve the overall accuracy rate in distinguishing between early AD onset and age-related functional decline, and potentially may improve efficiency in differentiating between different types of dementia.

Functional Magnetic Resonance Imaging (fMRI) provides the location and regional extent of a task correlated activation in the brain. Recently we have demonstrated, that fMRI of passive sensory tasks (visual, auditory, motor) can be successfully used to map cortical activation in the newborn brain. However the interpretation of the functional response in the immature brain is difficult, as the blood oxygen level dependent (BOLD) physiological signal and location of the activation is quite different compared to adult fMRI responses of similar tasks. We expect, that the major reason for these differences are primarily caused by the immature myelination of the white matter tracts at this age. Diffusion tensor imaging (DTI) can be used to measure the white matter tract development in the newborn brain. The purpose of this paper is to report how to obtain and to combine fMRI and DTI data processing to enhance functional brain mapping in newborns. We obtained simultaneous fMRI and DTI data of 18 newborns, post-conceptional age (gestational age at study) between 34-week and 52-week, which were referred for clinical indicated MRI. 16 out of 18 subjects have been successfully investigated with combined fMRI and DTI and functional activation could be obtained. Fiber tracking was successfully in the visual and auditory cortex, but proofed difficult in the motor-cortex. The additional tract information supported the functional findings and the interpretation in the immature brain. The novel functional imaging in newborn is challenging because of the yet unknown physiological response and location of activation in the newborn brain. Therefore one need additional evidence that the functional findings are valid in the context of structural development. The maturation of myelination is an essential information to compare and to interpret fMRI in newborns. We conclude that the proposed method of combined fMRI and DTI, derived from adult neuroimaging, will be most relevant to understand the physiological response and thus the neurodevelopment of the newborn brain.

The purpose of this study was to perform CT colonography (CTC) without cathartic colon cleansing. Four groups of 3 patients were prepared the day before CTC with a dedicated low residue diet, a hydration control allowing 2 liters of fluid intake and barium as tagging agent. Four different barium regimens were investigated. Groups 1 and 3 ingested barium over 1 day at different concentrations and groups 2 and 4 over 2 days. The barium volume to drink the day before CTC was 750 ml in groups 1 and 2 and 50 ml in groups 3 and 4. The fluid, density measurements of the fecal residue and tagging efficacy were evaluated. All fecal residue with densities ≥ 150 H.U. was electronically labeled. Per segment a visual labeling score (0, 25, 50, 75, 100%) was performed. The fluid was evaluated according to its proportion to the maximum anteroposterior diameter of the colonic segment where it was detected. No significant differences in densities of tagged residue were detected. The visual labeling scores varied between 90 and 100% in all segments. There were 6 fluid levels: 5 covering < 10% and one covering 50% of the colonic lumen. The lowest density of the fluid was 360 H.U. In this preliminary study we could conclude that CTC without cathartic cleansing and with barium produced efficient labeling of fecal residue. The barium intake could be reduced to one day and to 50 ml.

The Geometric Deformable Model is developed for accurate colon lumen segmentation as part of an automatic Virtual Colonoscopy system. The deformable model refines the lumen surface found by an automatic seed location and thresholding procedure. The challenges to applying the deformable model are described, showing the definition of the stopping function as the key to accurate segmentation. The limitations of current stopping criteria are examined and a new definition, tailored to the task of colon segmentation, is given. First, a multiscale edge operator is used to locate high confidence boundaries. These boundaries are then integrated into the stopping function using a distance transform. The hypothesis is that the new stopping function results in a more accurate representation of the lumen surface compared to previous monotonic functions of the gradient magnitude. This hypothesis is tested using observer ratings of colon surface fidelity at 100 hundred randomly selected locations in each of four datasets. The results show that the surfaces determined by the modified deformable model better represent the lumen surface overall.

An improved 3D method for colonic polyp segmentation was developed based on previous 2D method. The method is based on combination of 3D knowledge-guided intensity adjustment, fuzzy clustering, and dynamic deformable surfaces. After intensity adjustment and fuzzy clustering, a deformable surface is employed to locate the polyp boundaries. The surface is also dynamically maintained to preserve the resolution and topology. The deformable surface is operated on a sub-volume of the data set and driven by image forces, balloon forces and internal spline forces. Compared to previous 2D method, the improved method produces much smoother polyp boundaries, and 3D features can be derived from the segmentation, such as 3D aspect ratio, curvatures, and polyp wall thickness etc. The computer segmentations were validated with manual segmentations. Preliminary results showed that the average volume overlap percentage among 25 polyp detections was 80.6%.

The presented method significantly reduces the time necessary to validate a computed tomographic colonography (CTC) computer aided detection (CAD) algorithm of colonic polyps applied to a large patient database. As the algorithm is being developed on Windows PCs and our target, a Beowulf cluster, is running on Linux PCs, we made the application dual platform compatible using a single source code tree. To maintain, share, and deploy source code, we used CVS (concurrent versions system) software. We built the libraries from their sources for each operating system. Next, we made the CTC CAD algorithm dual-platform compatible and validate that both Windows and Linux produced the same results. Eliminating system dependencies was mostly achieved using the Qt programming library, which encapsulates most of the system dependent functionality in order to present the same interface on either platform. Finally, we wrote scripts to execute the CTC CAD algorithm in parallel. Running hundreds of simultaneous copies of the CTC CAD algorithm on a Beowulf cluster computing network enables execution in less than four hours on our entire collection of over 2400 CT scans, as compared to a month a single PC. As a consequence, our complete patient database can be processed daily, boosting research productivity. Large scale validation of a computer aided polyp detection algorithm for CT colonography using cluster computing significantly improves the round trip time of algorithm improvement and revalidation.

Asymmetric stents are promising new devices for endovascular treatment of cerebrovascular aneurysms. For in vitro experiment a patch made of stainless steel mesh is directly attached onto a standard stent and deployed so that the patch is placed over the aneurysm orifice. hus we modify substantially the flow into the aneurysm and decrease the shear stress on the aneurysm walls. We used mesh-patches having different permeabilities and evaluated the flow using Particle Image Velocimetry. PIV provides instantaneous velocity vector measurements in a cross-section of flow containing reflective micro-particles. A pulsed-laser light sheet illuminates the flow in the target area and images are acquired using a CCD camera. By registering the position of the particles in two successive images the fluid velocity vectors components are calculated. From the 2D velocity field a best polynomial fit is made to obtain a smooth function of each velocity with respect to the coordinates. Using the fit, we derived the values of quantities of interest in the plane of acquisition such as: tangent shear stress, vorticity and inflow. We used four meshes of different permeabilities. We found out that by using lower permeability meshes we create better conditions for the embolization of the aneurysm.

We have built new asymmetric stents for minimally invasive endovascular treatment of cerebral aneurysms. Each asymmetric stent consists of a commercial stent with a micro-welded circular mesh patch. The blood flow modification in aneurysm-vessel phantoms due to these stents was evaluated using x-ray angiographic analysis. However, the density difference between the radiographic contrast and the blood gives rise to a gravity effect, which was evaluated using an initial optical dye-dilution experiment. For the radiographic evaluations, curved-vessel phantoms instead of simple straight side-wall aneurysm phantoms were used in the characterization of meshes/stents. Six phantoms (one untreated, one treated with a commercial stent, and four treated with different asymmetric stents) with similar morphologies were used for comparison. We calculated time-density curves of the aneurysm region and then calculated the peak value (Pk) and washout rate (1/τ) after analytical curve fitting. Flow patterns in the angiograms showed reduction of vortex flow and slow washout in the dense mesh patch treated aneurysms. The meshes reduced Pk down to 21% and 1/τ down to 12% of the values for the untreated case. In summary, new asymmetric stents were constructed and their evaluation demonstrates that they may be useful in the endovascular treatment of aneurysms.

Characterization of the blood flow patterns in cerebral aneurysms is important to explore possible correlations between the hemodynamics conditions and the morphology, location, type and risk of rupture of intracranial aneurysms. For this purpose, realistic patient-specific models are constructed from computed tomography angiography and 3D rotational angiography image data. Visualizations of the distribution of hemodynamics forces on the aneurysm walls as well as the intra-aneurysmal flow patterns are presented for a number of cerebral aneurysms of different sizes, types and locations. The numerical models indicate that there are different classes of intra-aneurysmal flow patterns, that may carry different risks of rupture.

Due to histological evidence of the fundamental role of the cerebral vessels in white matter abnormalities, recently there has been an increased interest in analyzing the relationship between brain white matter lesions in multiple sclerosis (MS) and brain vasculature. We developed a method for visualization and measurement of geometrical relationships between MS lesions and the brain vessels imaged with magnetic resonance (MR) imaging techniques. Using MR images we create surface models of lesions and vessels that constitute a base for quantitative analysis. In this work we analyze correlation between basic lesion geometrical characteristics and two features: 1) distances to vessels, and 2) vessel caliber. For the former, we compute a distance map from the vessel structure, such that each voxel stores its distance vector to the closest vessel. This allows the measurements of Euclidean distances to the closest vessels. For the latter, we compute a radius map in which each voxel stores the radius of its closest vessel. It is used to measure distribution of lesions with respect to the vessel caliber. We compute and analyze relations between the basic geometrical characteristics of lesions and the closest vessels locations and calibers. To demonstrate the feasibility of the developed technique we present results from the study of 3 MS cases.

To explore and quantify pulmonary arterial remodeling we used various methods including micro-CT, high-resolution 3-dimensional x-ray imaging, to examine the structure and function of intact pulmonary vessels in isolated rat lungs. The rat is commonly used as an animal model for studies of pulmonary hypertension (PH) and the accompanying vascular remodeling, where vascular remodeling has been defined primarily by changes in the vessel wall composition in response to hypertension inducing stimuli such as chronic hypoxic exposure (CHE) or monocrotaline (MCT) injection. Little information has been provided as to how such changes affect the vessel wall mechanical properties or the lumenal architecture of the pulmonary arterial system that actually account for the hemodynamic consequences of the remodeling. In addition, although the link between primary forms of pulmonary hypertension and inherited genetics is well established, the role that genetic coding plays in hemodynamics and vascular remodeling is not. Therefore, we are utilizing Fawn-Hooded (FH), Sprague-Dawley (SD) and Brown Norway (BN)rat strains along with unique imaging methods to parameterize both vessel distensibility and lumenal morphometry using a principal pulmonary arterial pathway analysis based on self-consistency. We have found for the hypoxia model, in addition to decreased body weight, increased hematocrit, increased right ventricular hypertrophy, the distensibility of the pulmonary arteries is shown to decrease significantly in the presence of remodeling.

On many occasions, it is desirable to image lungs in vivo to perform a pulmonary physiology study. Since the lungs are moving, gating with respect to the ventilatory phase has to be performed in order to minimize motion artifacts. Gating can be done in real time, similar to cardiac imaging in clinical applications, however, there are technical problems that have lead us to investigate different approaches. The problems include breath-to-breath inconsistencies in tidal volume, which makes the precise detection of ventilatory phase difficult, and the relatively high ventilation rates seen in small animals (rats and mice have ventilation rates in the range of a hundred cycles per minute), which challenges the capture rate of many imaging systems (this is particularly true of our system which utilizes cone-beam geometry and a 2 dimensional detector). Instead of pre-capture ventilation gating we implemented a method of post-acquisition gating. We acquire a sequence of projections images at 30 frames per second for each of 360 viewing angles. During each capture sequence the rat undergoes multiple ventilation cycles. Using the sequence of projection images, an automated region of interest algorithm, based on integrated grayscale intensity, tracts the ventilatory phase of the lungs. In the processing of an image sequence, multiple projection images are identified at a particular phase and averaged to improve the signal-to-ratio. The resulting averaged projection images are input to a Feldkamp cone-beam algorithm reconstruction algorithm in order to obtain isotropic image volumes. Minimal motion artifact data sets improve qualitative and quantitative analysis techniques useful in physiologic studies of pulmonary structure and function.

A method for precise and sensitive organ specific quantification of ¹²⁴I (¹²⁴IdUR) activities in small animals has been developed. The method is based upon high precision co-registration between PET and MR imagery, utilizing an innovative mouse fixation system with external point sources visible in both modalities. The methodology is generic, and thus can easily be extended to other types of small animal studies. The methodology comprises PET/MR rigid co-registration utilising external fiducial markers, measurement of activities within small 3D volumes manually drawn in MR data and extracted from the co-aligned PET data, and subsequently corrected for Partial Volume Effect.

We are investigating imaging techniques to study the rapid biochemical and physiological response of tumors to photodynamic therapy (PDT). Positron emission tomography (PET) can provide physiological and functional images of cancers. While MRI can provide high resolution anatomical images and generate serial, noninvasive, in vivo observations of morphological changes. In this study, we investigate image registration methods to combine MRI and micro-PET (μPET) images for improved tumor monitoring. We acquired high resolution MR and PET ¹⁸F-fluorodeoxyglucose (FDG) images from mice with RIF-1 tumors. We used rigid body registration with three translations and three angular variables. We used normalized mutual information as the similarity measure. To assess the quality of registration, we performed slice by slice review of both image volumes, manually segmented feature organs such as the left and right kidneys and the bladder in each slice, and computed the distance between corresponding centroids of the organs. We also used visual inspection techniques such as color overlay displays. Over 40 volume registration experiments were performed with MR and μPET images acquired from three C3H mice. The color overlays showed that the MR images and the PET images matched well. The distance between corresponding centroids of organs was 1.5 ± 0.4 mm which is about 2 pixels of μPET. In conclusion, registration of high resolution MR and μPET images of mice may be useful to combine anatomical and functional information that could be used for the potential application in photodynamic therapy.

The changes in the tumor that occur following photodynamic therapy (PDT) were studied using a small animal MR imager operating at 7Tesla. The animal model used in these studies was mice bearing radiation induced fibrosarcoma (RIF) tumor on the foot dorsum. The mice were injected with 10μM/kg of one of the photosensitizers: (1) Photofrin, (2) Non-fluorinated porphyrin photosensitizer (DOD-1), (3) Fluorinated porphyrin photosensitizer (DOD-2) and, (4) Fluorinated chlorin photosensitizer (DOD-6). Laser light at 630 or 650 nm (150 mW/cm², 270 joules/cm²) was delivered to the tumor at 2-24 hours of photosensitizer administration. The MR spectroscopic and imaging examination of the tumors involved both the ¹H and ³¹P nuclei. The tumor bioenergetics was measured by ³¹P spectroscopy. The water proton relaxivity and diffusion measurements were used to obtain local changes in different regions of the tumor. Changes in ³¹P MR spectra were observed following PDT using Photofrin and fluorinated chlorin sensitizer (DOD-6). However, no significant changes were observed when the fluorinated porphyrin and its nonfluorinated analog were used. The PDT induced changes in tumor volumes showed significant tumor regression with Photofrin, fluorinated porphyrin and chlorin sensitizers. No tumor regression was observed with the non labeled porphyrin sensitizer and the growth profile followed the general pattern of unperturbed tumors. Serial noninvasive measurements of tumor response to PDT are measurable by both MRI and MRS. The MR derived parameters that are characteristic of the tumor status before and after the therapy are discussed here.

The bronchial circulation is thought to be the primary blood supply for pulmonary carcinomas. Thus, we have developed a method for imaging and quantifying changes in perfusion in the rat lung due to development of the bronchial circulation. A dual-modality micro-CT/SPECT system was used to detect change in perfusion in two groups of rats: controls and those with a surgically occluded left pulmonary artery. Both groups were imaged following injections on separate days i) 2mCi of Tc99m labeled macroaggregated albumin (MAA) into the left carotid artery (IA) and ii) a similar injection into the femoral vein (IV). The IA injection resulted in Tc99m accumulation in capillaries of the systemic circulation including the bronchial circulation, whereas the IV resulted in Tc99m accumulation in the pulmonary capillaries. Ordered subset expectation maximization (OSEM) was used to reconstruct the SPECT image volumes and a Feldkamp algorithm was used to reconstruct the micro-CT image volumes. The micro-CT and SPECT volumes were registered, the SPECT image volume was segmented using the right and left lung boundaries defined from the micro-CT volume, and the ratio of IA radioactivity accumulation in the left lung to IV radioactivity accumulation in both lungs was used as a measure of left lung flow via the bronchial circulation. This ratio was ~0.02 for the untreated rats compared to the treated animals that had an increased flow ratio of ~0.21 40 days after left pulmonary artery occlusion. This increase in flow to the occluded left lung via the bronchial circulation suggests this will be a useful model for further investigating antiangiogenic treatments.

High-speed X-ray computed tomography (CT) has the potential to observe the transport of iodinated radio-opaque contrast agent (CA) through tissue enabling the quantification of tissue physiology in organs and tumors. The concentration of Iodine in the tissue and in the left ventricle is extracted as a function of time and is fit to a compartmental model for physiologic parameter estimation. The reproducibility of the physiologic parameters depend on the (1) The image-sampling rate. According to our simulations 5-second sampling is required for CA injection rates of 1.0ml/min (2) the compartmental model should reflect the real tissue function to give meaning results. In order to verify these limits a functional CT study was carried out in a group of 3 mice. Dynamic CT scans were performed on all the mice with 0.5ml/min, 1ml/min and 2ml/min CA injection rates. The physiologic parameters were extracted using 4 parameter and 6 parameter two compartmental models (2CM). Single factor ANOVA did not indicate a significant difference in the perfusion, in the kidneys for the different injection rates. The physiologic parameter obtained using the 6-parameter 2CM model was in line with literature values and the 6-parameter significantly improves chi-square goodness of fits for two cases.

Utilization of computed tomography (CT) for lung cancer screening has attracted significant research interests in recent years. Images reconstructed from CT studies are used for lung nodule characterization and three-dimensional lung lesion sizing. Methodologies have been developed to automatically identify and characterize lung nodules. In this paper, we analyze the impact of acquisition and reconstruction parameters on the accuracy of quantitative lung nodule characterization. The two major data acquisition parameters that impact the accuracy of the lung nodule measurement are acquisition mode and slice aperture. Acquisition mode includes both axial and helical scans. The investigated reconstruction parameters are the reconstruction filters and field-of-view. We first develop theoretical models that predict the system response under various acquisition and reconstruction conditions. These models allow clinicians to compare results under different conditions and make appropriate acquisition and reconstruction decisions. To validate our model, extensive phantom experiments are conducted. Experiments have demonstrated that our analytical models accurately predict the performance parameters under various conditions. Our study indicates that acquisition and reconstruction parameters can significantly impact the accuracy of the nodule volume measurement. Consequently, when conducting quantitative analysis on lung nodules, especially in sequential growth studies, it is important to make appropriate adjustment and correction to maintain the desired accuracy and to ensure effective patient management.

Computer-aided diagnosis (CAD) has been investigated to provide physicians with quantitative information, such as estimates of the malignant likelihood, to aid in the classification of abnormalities detected at screening of lung cancers. The purpose of this study is to develop a method for classifying nodule density patterns that provides information with respect to nodule statuses such as lesion stage. This method consists of three steps, nodule segmentation, histogram analysis of CT density inside nodule, and classifying nodules into five types based on histogram patterns. In this paper, we introduce a two-dimensional (2-D) joint histogram with respect to distance from nodule center and CT density inside nodule and explore numerical features with respect to shape and position of the joint histogram.

In this study we present an automatic algorithm for the detection and functional assessment of lung nodules on three-dimensional slices derived from a hybrid PET/CT scanner. In addition to differentiate malignant from benign lesions, the algorithm was mainly designed for assessing the response of lung cancer to therapy. The automated algorithm involves three major steps. First, the lung region is segmented from low resolution multislice CT images. Once the lung is segmented on CT images, a search of seed pixels with maximum activity of ¹⁸FDG is undertaken into the lung regions of the electronically registered PET images. A 3D growing algorithm identified the lesion pixels around the maximum ¹⁸FDG activity seed pixels. In the third step, the total activity (Bq), concentration (Bq/ml), metabolically active volume (ml) and standard uptake values (SUV) were calculated for lesions on PET images. A threshold and filtering method was applied to high resolution CT scans to determine the CT volume of these lesions identified on PET images. All PET images were corrected for attenuation and partial volume effect and cross calibrated with a standard activity measured in a dose calibrator. Studies were performed using a hybrid PET/CT Discovery LS (GE Medical Systems). The feasibility and robustness of the automatic algorithm was demonstrated in studies with a lung-chest phantom and by retrospective analysis of clinical studies.

Micro CT system is developed for lung function analysis at a high resolution of the micrometer order (up to 5 μm in spatial resolution). This system reveals the lung distal structures such as interlobular septa, terminal bronchiole, respiratory bronchiole, alveolar duct, and alveolus. In order to visualize lung 3-D microstructures using micro CT images and to analyze them, this research presents a computerized approach. In this approach, the following things are performed: (1) extracting lung distal structures from micro CT images, (2) visualizing extracted lung microstructure in three dimensions, and (3) visualizing inside of lung distal area in three dimensions with fly-through. This approach is applied for to micro CT images of human lung tissue specimens that were obtained by surgical excision and were kept in the state of the inflated fixed lung. And this research succeeded in visualization of lung microstructures using micro CT images to reveal the lung distal structures from bronchiole up to alveolus.

We have previously reported a synchrotron radiation (SR) microtomography system constructed at the bending magnet beamline at the SPring-8. This system has been applied to the lungs obtained at autopsy and inflated and fixed by Heitzman’s method. Normal lung and lung specimens with two different types of pathologic processes (fibrosis and emphysema) were included. Serial SR microtomographic images were stacked to yield the isotropic volumetric data with high-resolution (12 μm³ in voxel size). Within the air spaces of a subdivision of the acinus, each voxel is segmented three-dimensionally using a region growing algorithm (“rolling ball algorithm”). For each voxel within the segmented air spaces, two types of voxel coding have been performed: single-seeded (SS) coding and boundary-seeded (BS) coding, in which the minimum distance from an initial point as the only seed point and all object boundary voxels as a seed set were calculated and assigned as the code values to each voxel, respectively. With these two codes, combinations of surface rendering and volume rendering techniques were applied to visualize three-dimensional morphology of a subdivision of the acinus. Furthermore, sequentially filling process of air into a subdivision of the acinus was simulated under several conditions to visualize the ventilation procedure (air flow and diffusion). A subdivision of the acinus was reconstructed three-dimensionally, demonstrating the normal architecture of the human lung. Significant differences in appearance of ventilation procedure were observed between normal and two pathologic processes due to the alteration of the lung architecture. Three-dimensional reconstruction of the microstructure of a subdivision of the acinus and visualization of the ventilation procedure (air flow and diffusion) with SR microtomography would offer a new approach to study the morphology, physiology, and pathophysiology of the human respiratory system.

One of the promising applications of Electrical Impedance Tomography (EIT) is to monitor lung ventilation. However, during respiratory activity, chest moves as much as 10% in the anterior-posterior direction. No EIT reconstruction algorithm has considered this change. For EIT system, which is sensitive to electrode positions, whether chest expansion needs to be incorporated becomes an important problem to study. The results using a 2D thorax model showed that chest expansion accounted for up to 20% of the reconstructed image amplitude and introduced an artifact in the center of the image. To better characterize this issue, a 3D model with detailed anatomical structures, which was developed based on a series of MR thorax images covering from neck to abdomen, was used to simulate the influence of chest expansion. At one selected layer of the model, sixteen electrodes were placed evenly to simulate real measurement setup. Chest expansions were simulated by changing the size of 3D human thorax model. Then EIT dynamic images were reconstructed based on stimulated measurements before and after chest expansion. By analyzing these images, the relationship between chest expansion and the amplitude in the lung regions was established. The results show that chest expansion accounts for up to 31% of the changes in the reconstructed resistivity images. And they further confirm that chest expansion introduces an artifact in the center of the images. Based on this study, chest expansion should be considered in EIT image reconstruction.

Pulmonary CT images can provide detailed information about the regional structure and function of the respiratory system. Prior to any of these analyses, however, the lungs must be identified in the CT data sets. A popular animal model for understanding lung physiology and pathophysiology is the sheep. In this paper we describe a lung segmentation algorithm for CT images of sheep. The algorithm has two main steps. The first step is lung extraction, which identifies the lung region using a technique based on optimal thresholding and connected components analysis. The second step is lung separation, which separates the left lung from the right lung by identifying the central fissure using an anatomy-based method incorporating dynamic programming and a line filter algorithm. The lung segmentation algorithm has been validated by comparing our automatic method to manual analysis for five pulmonary CT datasets. The RMS error between the computer-defined and manually-traced boundary is 0.96 mm. The segmentation requires approximately 10 minutes for a 512x512x400 dataset on a PC workstation (2.40 GHZ CPU, 2.0 GB RAM), while it takes human observer approximately two hours to accomplish the same task.

Most of the three-dimensional (3-D) reconstructions of human organs rely on medical volume data. In this paper, we propose to use the endoscopic image sequences as a new image modality for surface reconstruction of human organs or tissues. Using the pinhole camera model, the original 3-D points are projected to two-dimensional (2-D) images by multiplying transformation matrices. We assume the intrinsic camera parameters, such as the focal length and principal points, are known and simplify the transformation matrices to only include the camera motion, i.e. camera rotation and translation. Using the factorization method for recovering the shape of the object and the motion of the camera from an image sequence, the 3-D structures are computed. 3-D reconstruction from endoscopic image sequences is a new exploration. It provides additional information that facilitates the understanding for the lesion areas three-dimensionally. And the reconstructed structures directly correspond to the original images and can be rendered with precise texture-mapping easily. It has potential to be used to guide surgery and serve as an alternative data source for constructing new stereo endoscopy systems using one camera.

Segmentation and measurement of vascular structures is an important topic in medical image analysis. The aim of this paper is to present a hybrid approach to accurate segmentation of vascular structures from MRA images using level set methods and deformable geometric model, constructed with 3D Delaunay triangulation. Based on the analysis of local intensity structure, multiple scale filtering is derived from the Hessian matrix and then is used to effectively enhance vessel structures with various diameters. We apply the level set method to automatically segment vessels enhanced by the filtering. The segmentation of vessels from 3D vessel enhanced images can be regarded as an evolution of a propagating implicit surface in a 3D space that separates vessel volumes from another and moves in a normal direction to the vessel boundaries with a given speed function over time. The speed function used is derived from the results of filtering. In subsequent step, in order to make the segmented vessel surface fit the actual vessel surface more accurately, we triangulate the segmented vessel surface using 3D Delaunay triangulation and use the triangulated surface as a deformable model in order to minimize an energy functional in which the internal force is defined as linear equations with the local surface patch given by 3D Delaunay triangulation and the external force is derived from the gradient information of original images. Using the proposed method, vessels can be effectively and accurately segmented from MRA images.

Very large 3D digital images of arterial trees can be produced by many imaging scanners. While many automatic approaches have been proposed that can begin the process of defining the 3D arterial tree captured in such an image, none guarantee complete, accurate definition. This paper proposes semi-automatic techniques for coming closer to the ultimate goal of defining a complete and accurate 3D arterial tree. As pointed out previously, automated techniques are essential for beginning the process of extracting a complex 3D arterial tree from a large 3D micro-CT image. Yet, many problems arise in this definition of the tree. Our system, initially expounded upon in an earlier effort, uses a series of interactive and semi-automatic tools to examine and correct the identified problems. The system has 3D graphical tools for viewing global and local renderings of the extracted tree, revealing sliding thin-slab views, maximum-intensity projections, multi-planar reformatted slices, and a global/local 2D graphical tree map. The user can invoke several semi-automatic tools for changing the tree as well. The presented results demonstrate the potential of the system.

In this article we demonstrate the use of an automated technique to visualize lesions in the abdominal aorta, gathered from MRI imagery, and displayed as a stained pathology specimen. We have developed this technique in response to the suggestion from clinical colleagues that such a representation of the data is more understandable to the pathologist than presentation of axial MRI slices, even if the atheroma is of high contrast and very thick. This virtual dissection relies on an initial manual segmentation of the inner and outer walls of the vessel, which I achieved using a commercial cardiac analysis package. The algorithm consists of (1) decoding the file that describes the contours; (2) generating a center for each slice; (3) determining the appropriate posterior position on all slices; (4) interpolating along the (approximately circular) lumen and through all slices; (5) false color presentation of the wall thickness as a pathological stain.

The dynamics of curvature and torsion are important for the geometric description of arteries and for the distribution of accumulating plaque. In this research, two methods for estimating curvature and torsion are analyzed with respect to their accuracy. The first method is based on estimating the curvature and torsion of the artery centerline using the Fourier transform. Since the centerline always represents an open curve, extensions ensuring a minimal spectral energy are added on both ends to obtain a closed curve suitable for Fourier analysis. The second method has been previously used for analyzing the motion of coronary arteries and is based on the least squares fitting of a cubic polynomial to the centerline of the artery. Validation is performed using two mathematical, time-varying phantoms as well as 4-D (3-D plus time) in-vivo data of coronary arteries reconstructed by fusion of biplane angiograms and intravascular ultrasound images. Results show that both methods are accurate for estimating curvature and torsion, and that both methods have average errors below 2.15%.

Carotid imaging is a Gold Standard test that provides useful information about the structure and functions of carotid arteries. Spectral imaging helps to evaluate the vessel and hemodynamic changes. High resolution B-mode imaging has emerged as one of the methods of choice for determining the anatomic extent of atherosclerosis and its progression and for assessing cardiovascular risks. The measurements made with Doppler correlate well with pathologic measurements. Recent prospective studies have clearly demonstrated that these measurements of carotid intimal thickness are potent predictors of Myocardial Infarction and Stroke. This method appears very attractive as it is non-invasive, extremely safe, well accepted by the patient and relatively inexpensive. It can be performed serially and has the advantage of visualizing the arterial wall in contrast to angiographic techniques which provide only an outline of the arterial lumen. Recently, there has been an interest in the clinical use of this technique in making difficult clinical decisions like deciding on preventive therapies. 30 subjects aged 21-60 years and 30 subjects aged 61-85 years of both sexes are selected after doing a baseline study to exclude Hypertension, Diabetes, Obesity and Hyperlipidemia. The carotid arteries were examined for intimal thickening, blood flow velocities and luminal diameter. With aging there is a narrowing of the carotid vessels and significant increase in intimal thickening with a consequent increase in the blood flow velocities. Inter-observer, intra-observer and instrument variations are seen and there is no significant change in the values when the distal flow pattern is considered for measurements. Aging produces major cardiovascular changes including decreased elasticity and compliance of great arteries leading to structural and functional alterations in heart and vessels. With aging there is increased intimal thickness and increased pulse wave velocity which is clearly understood by using high resolution B-Mode carotid imaging techniques.

Hemodynamic conditions in the carotid artery may be associated with progression of carotid artery (CA) disease and with the risk of stroke. A methodology for objective measurement of distension (D_CA) and blood-flow (Q_CA) waveform from phase-contrast (PC) magnetic resonance (MR) imaging is presented. Measurement of D_CA is obtained using a modified Hough Transform (mHT) applied to the magnitude-component of the PC MR. The mHT is based on the optimization of an objective function which is the sum of the gradient magnitude of the image sampled at 10° increments around the circle using bilinear interpolation. The mHT detects the boundary of the vessel in the cross-sectional image at 0.05-pixel resolution. Measurement of Q_CA is obtained by integration of the image intensity in the phase-component of the PC MR within the circular region detected by the mHT.

Determination of the hemodynamics conditions in atherosclerotic carotid arteries is important for quantitative assessment of the severity of the disease. Since there are no reliable experimental methods to determine the in vivo wall shear stress distribution in the region of the stenosis, realistic patient-specific image-based finite element models are constructed. The purpose of this paper is to present validation studies based on multi-modality image data of patients with carotid artery disease. The velocity profiles and peak velocity at stenoses computed by the computational fluid dynamics models are in very good agreement with phase-contrast magnetic resonance and carotid Doppler ultrasound measurements, respectively.

Accurate 3-D rectangular meshes construction of pipes is important and valuable for Engineering and Medicine to analyze the fluid mechanics. Hypertrophied prostate is suffered common by aged patient. Medicine trusts that the pathogenic reason can be interpreted using the statistics of analysis of fluid mechanics. We use serials of methods to construct the 3-D rectangular meshes of a patient urethra CT volumetric data given from medicine, and for mechanic engineering to gather statistics.

The aim of this study was to assess the validation of the local density random walk (LDRW) function to correct the delayed and dispersed arterial input function (AIF) data derived from dynamic susceptibility contrast magnetic resonance imaging (DSC-MRI). Instead of using the gamma-variate function to smooth and extrapolate the AIF curves, we suggested a method which was based on diffusion with drift approach. Forty-seven AIF curves from ten patients were segmented to test the effectiveness of the proposed method. The results of the comparisons with the gamma-variate function showed that the LDRW distribution function may provide a new means for more accurate correction of AIF curves.

In order to extract quantitative information from dynamic contrast-enhanced MR images (DCE-MRI) it is usually necessary to identify an arterial input function. This is not a trivial problem if there are no major vessels present in the field of view. Most existing techniques rely on operator intervention or use various curve parameters to identify suitable pixels but these are often specific to the anatomical region or the acquisition method used. They also require the signal from several pixels to be averaged in order to improve the signal to noise ratio, however this introduces errors due to partial volume effects. We have described previously how factor analysis can be used to automatically separate arterial and venous components from DCE-MRI studies of the brain but although that method works well for single slice images through the brain when the blood brain barrier technique is intact, it runs into problems for multi-slice images with more complex dynamics. This paper will describe a factor analysis method that is more robust in such situations and is relatively insensitive to the number of physiological components present in the data set. The technique is very similar to that used to identify spectral end-members from multispectral remote sensing images.

The human cerebral ventricular system is a complex structure that is essential for the well being and changes in which reflect disease. It is clinically imperative that the ventricular system be studied in details. For this reason computer assisted algorithms are essential to be developed. We have developed a novel (patent pending) and robust anatomical knowledge-driven algorithm for automatic extraction of the cerebral ventricular system from MRI. The algorithm is not only unique in its image processing aspect but also incorporates knowledge of neuroanatomy, radiological properties, and variability of the ventricular system. The ventricular system is divided into six 3D regions based on the anatomy and its variability. Within each ventricular region a 2D region of interest (ROI) is defined and is then further subdivided into sub-regions. Various strict conditions that detect and prevent leakage into the extra-ventricular space are specified for each sub-region based on anatomical knowledge. Each ROI is processed to calculate its local statistics, local intensity ranges of cerebrospinal fluid and grey and white matters, set a seed point within the ROI, grow region directionally in 3D, check anti-leakage conditions and correct growing if leakage occurs and connects all unconnected regions grown by relaxing growing conditions. The algorithm was tested qualitatively and quantitatively on normal and pathological MRI cases and worked well. In this paper we discuss in more detail inclusion of anatomical knowledge in the algorithm and usefulness of our approach from clinical perspective.

Cerebrospinal fluid filled ventricular system is an essential part of brain. The volume, shape and size of this ventricular system remain more or less constant and various pathologies directly or indirectly affect them. Morphometric analysis of cerebral ventricular system is important for evaluating changes due to growth, aging, intrinsic and extrinsic pathologies. Previous quantification efforts using ex vivo techniques suffered considerable error due to deformation of slices during sectioning, and numerous other factors. In vivo studies using air or contrast media also introduce volumetric changes in the ventricles thus giving erroneous quantitative information. Imaging of ventricular anatomy avoids these problems and allows repetitive studies following progression of ventricular system changes due to disease or natural processes. We have developed a methodology for automated extraction of ventricular system from MR neuroimages. Once extracted, landmarks are located on the surface of ventricular system automatically. These landmarks are then used for calculation of the ventricular shape, volume and size. A total of 20 brain ventricular systems were analyzed. The morphometric dimensions of the ventricles are presented in this paper. This study forms an initial basis for more advanced work on ventricular segmentation and morphometry.

We have developed a new technique to analyze correlations between brain anatomy and its neurological functions. The technique is based on the anatomic MRI of patients with brain lesions who are administered neuropsychological tests. Brain lesions of the MRI scans are first manually segmented. The MRI volumes are then normalized to a reference map, using the segmented area as a mask. After normalization, the brain lesions of the MRI are segmented again in order to redefine the border of the lesions in the context of the normalized brain. Once the MRI is segmented, the patient's score on the neuropsychological test is assigned to each voxel in the lesioned area, while the rest of the voxels of the image are set to 0. Subsequently, the individual patient's MRI images are superimposed, and each voxel is reassigned the average score of the patients who have a lesion at that voxel. A threshold is applied to remove regions having less than three overlaps. This process leads to an anatomo-functional map that links brain areas to functional loss. Other maps can be created to aid in analyzing the functional maps, such as one that indicates the 95% confidence interval of the averaged scores for each area. This anatomo-clinical overlapping map (AnaCOM) method was used to obtain functional maps from patients with lesions in the superior frontal gyrus. By finding particular subregions more responsible for a particular deficit, this method can generate new hypotheses to be tested by conventional group methods.

Intensity and volume features of the hippocampus from MR images of the brain are known to be useful in detecting the abnormality and consequently candidacy of the hippocampus for temporal lobe epilepsy surgery. However, currently, intracranial EEG exams are required to determine the abnormal hippocampus. These exams are lengthy, painful and costly. The aim of this study is to evaluate texture characteristics of the hippocampi from MR images to help physicians determine the candidate hippocampus for surgery. We studied the MR images of 20 epileptic patients. Intracranial EEG results as well as surgery outcome were used as gold standard. The hippocampi were manually segmented by an expert from T1-weighted MR images. Then the segmented regions were mapped on the corresponding FLAIR images for texture analysis. We calculate the average energy features from 2D wavelet transform of each slice of hippocampus as well as the energy features produced by 3D wavelet transform of the whole hippocampus volume. The 2D wavelet transform is calculated both from the original slices as well as from the slices perpendicular to the principal axis of the hippocampus. In order to calculate the 3D wavelet transform we first rotate each hippocampus to fit it in a rectangular prism and then fill the empty area by extrapolating the intensity values. We combine the resulting features with volume feature and compare their ability to distinguish between normal and abnormal hippocampi using linear classifier and fuzzy c-means clustering algorithm. Experimental results show that the texture features can correctly classify the hippocampi.

An algorithm to automatically detect brain tumors in MR images is presented. The key concern is speed in order to process efficiently large brain image databases and provide quick outcomes in clinical setting. The method is based on study of asymmetry of the brain. Tumors cause asymmetry of the brain, so we detect brain tumors in 3D MR images using symmetry analysis of image grey levels with respect to the midsagittal plane (MSP). The MSP, separating the brain into two hemispheres, is extracted using our previously developed algorithm. By removing the background pixels, the normalized grey level histograms are calculated for both hemispheres. The similarity between these two histograms manifests the symmetry of the brain, and it is quantified by using four symmetry measures: correlation coefficient, root mean square error, integral of absolute difference (IAD), and integral of normalized absolute difference (INAD). A quantitative analysis of brain normality based on 42 patients with tumors and 55 normals is presented. The sensitivity and specificity of IAD and INAD were 83.3% and 89.1%, and 85.7% and 83.6%, respectively. The running time for each symmetry measure for a 3D 8bit MR data was between 0.1 - 0.3 seconds on a 2.4GHz CPU PC.

Diffusion weighted imaging (DWI) is the gold standard for imaging of acute stroke. Today, high-field systems operating at 3T become increasingly available in clinical settings. But, with b-value increasing, lesion SNR of DWI image descends, and anisotropy increases significantly. Aim of the study is to develop an automatic volumetric measure method of ischemic lesions on diffusion weighted imaging (DWI) images at high magnetic field, without the disturbance of anisotropy. Using a home-built interactive platform, we rated SNR and anisotropy. The extent of anisotropy was evaluated by the intensity ratio of white matter versus gray matter. Based on this knowledge, we developed an automatic segmentation method, involving firstly non-linear anisotropic diffusion filtering, secondly expert pieces of information applied to determine the scopes of parameters according to different b-value, and finally multi-scale adaptive statistical classification with intensity inhomogeneity correction. Results of the automatic segmentation are compared with lesion delineations by experts, showing the rapid identification of ischemic lesion with accuracy and reproducibility, even in the presence of radio frequency (RF) inhomogeneity. There has been considerable interest in using DWI at 3T to detect ischemic lesion in stroke patients. The proposed method is promising for rapid, accurate, and quantitatively diagnosis of ischemic stroke.

We have investigated the feasibility of obtaining high-quality Diffusion Tensor Magnetic Resonance Imaging (DTI) data in newborn humans. We show that the use of an MR-compatible incubator with customized RF headcoils can provide diffusion tensor maps of sufficient quality for quantitative DTI measurements and 3D fiber tracking. We have also investigated the effect of performing affine co-registration on the diffusion-weighted images, as is conventionally believed to be necessary to correct for eddy current distortion effects. We have found that co-registration indeed successfully eliminates the well-known bright band of high anisotropy that forms in the peripheral brain regions, and that such co-registration also reduces smaller interior regions of artifactually high diffusion anisotropy. In addition, we have investigated whether non-affine distortions exist in the diffusion-weighted images, as might be expected due to the existence of large susceptibility gradients. The results of performing 2nd order mutual information polynomial registration of the diffusion-weighted images to the non-diffusion-weighted (b=0) image in each slice show that subtle differences between affine and 2nd order co-registration do exist, which suggests that care must be taken when interpreting FA values in cortical brain regions. Finally, we present results of 3D white matter fiber tracking in the newborn brain. To preserve the full information content of the DTI data, we used simple Euler integration without noise filtering or fiber crossing detection. Our results show that the directionality of the major white matter pathways can be visualized in newborns.

The radiotherapic treatment of cerebral arterious malformations (AVM) requires an accurate estimation of the AVM shape. This estimation is classically obtained from the delineation of the AVM in several 2D angiographic views. In this paper, a clinical study of the inter-observer variability in the AVM detection is first performed. It proves tha the estimated volume varie a lot between observers. For thee reasons, we propose a framewok for AVM delineation which makes use of 2D and 3D angiographic images: the initial estimate obtained with 2D angiographic images is then refined within the 3D volume using deformable models. Results are presented demonstrating shape delineation on various AVMs.

Fuzzy c-means (FCM) suffers from some limitations such as the need for a priori knowledge of the number of clusters, and unknown statistical significance and instability of the results, when it is applied to the raw fMRI time series. Based on randomization, we developed a method to control the false positive detection rate in FCM and estimate the statistical significance of the results. Using this novel approach, we proposed an fMRI activation detection method which uses FCM with controlled false positive rate. The ability of the method in controlling the false positive rate is shown by an analysis of false positives in activation maps of resting-state fMRI data. Controlling the false positive rate allows comparison of different feature spaces and fuzzy clustering methods. A new feature space, in multi and scalar wavelet domain, is proposed for activation detection in fMRI to address the stability problem. Finally, using the proposed method for controlling the false positive rate, the proposed feature space is compared to the cross-correlation feature space.

Functional magnetic resonance imaging (fMRI) intends to detect significant neural activity by means of statistical data processing. Commonly used statistical tests include the Student-t test, analysis of variance, and the generalized linear model test. A key assumption underlying these methods is that the data are Gaussian distributed. Moreover, although MR data are intrinsically complex valued, fMRI data analysis is usually performed on single valued magnitude data. Whereas complex MRI data are Gaussian distributed, magnitude data are Rician distributed. In this paper, we describe five Generalized Likelihood Ratio Tests (GLRTs) that fully exploit the knowledge of the distribution of the data: one is based on Rician distributed magnitude data and two are based on Gaussian distributed complex valued data. By means of Monte Carlo simulations, the performance of the GLRTs is compared with the classical statistical tests.

This study presents a fast orthogonal search (FOS) method for modeling fMRI time series. Based on the system identification theory, an orthogonalization procedure to model the fMRI time series is described. FOS method does not require equally space data, and can resolve sinusoidal frequencies much more closely than Fourier transform method. After the time series are modeled by means of the FOS, F-test is employed to detect the activation regions. Eight volunteers' data were collected to validate the proposed method. The results demonstrate the feasibility of the proposed method.

Non-negative Matrix Factorization (NMF) has previously been shown to be a useful decomposition for multivariate data. In this paper, we introduce this new technique to the field of fMRI data analysis. In order to make the representation suitable for task-related brain activation detection, we imposed some additional constraints, and defined an improved contrast function. We deduced the update rules and proved the convergence of the algorithm. In the procedure, the number of factors was determined by visual assessment. We studied 8 healthy right-handed adult volunteers by a 3.0T GE Signa scanner. A block design motor paradigm (bilateral finger tapping) stimulated the blood oxygenation level-dependent (BOLD) response. Gradient Echo EPI sequence was utilized to acquire BOLD contrast functional images. With this constrained NMF (cNMF) we could obtain major activation components and the corresponding time courses, which showed high correlation with the reference function (r>0.7). The results showed that our method would be feasible for detection brain activations from task-related fMRI series.

Synchronized oscillations in resting state timecourses have been detected in recent fMRI studies. These oscillations are low frequency in nature (<0.08 Hz), and seem to be a property of symmetric cortices. These fluctuations are important as a pontential signal of interest, which could indicate connectivity between functionally related areas of the brain. It has also been shown that the synchronized oscillations decrease in some spontaneous pathological states (such as cocaine injection). Thus, detection of these functional connectivity patterns may help to serve as a guage of normal brain activity. Currently, functional connectivity detection is applied only in offline post-processing analysis. Online detection methods have been applied to detect task activation in functional MRI. This allows real-time analysis of fMRI results, and could be important in detecting short-term changes in functional states. In this work, we develop an outline algorithm to detect low frequency resting state functional connectivity in real time. This will extend connectivity analysis to allow online detection of changes in "resting state" brain networks.

Resting state oscillations have been detected in functional MRI studies, and appear to be synchronized between functionally related areas. It has also been shown that these synchronized oscillations decrease in some pathological states. Thus, these fluctuations are important as a potential signal of interest, which could indicate connectivity between functionally related areas of the brain. A current challenge is to detect these patterns without using an external reference. ICA analysis is a promising model-free technique that finds the independent components in a data set. A drawback to using ICA is the possibility of convergence problems in the presence of noise, and signal mixing across components. This work utilizes a recently developed denoising method as a preprocessing step to condition task and resting state functional MRI data for ICA analysis. The advantages of this approach include increased reliability of ICA results and allowing region specific signal patterns to be separated using a model-free analysis.

Using cardiac magnetic resonance (MR) imaging, a combination of late contrast enhanced MR (ceMR) and cinematographic (CINE) images, a myocardial viability score can be derived. At present this score is produced by visual evaluation of wall motion abnormalities in combination with presence or absence of late hyper enhancement (LE) on ceMR. We set out to develop and validate image processing techniques derived from stereo vision capable of reducing the observer dependence and improving accuracy in the diagnosis of viable myocardium.

Measuring changes in cardiac motion patterns can assist in diagnosing the onset of arrhythmia and ischaemia and in the follow-up of treatment. This work presents a methodology for measuring such motion changes from MR images. Non-rigid registration is used to track cardiac motion in a sequence of 3D tagged MR images. We use a cylindrical coordinate system to subdivide the myocardium into smaller anatomically meaningful regions and to express motion derived measurements such as displacement and strain for each myocardial region during the cardiac cycle. In the first experiment we have evaluated the proposed methods using synthetic image sequences where the ground truth was available. These images were generated using a cardiac motion simulator for tagged MRI. Normal and abnormal motion fields were produced by modifying parameters in a small region of the myocardium. In the second experiment we have acquired two separate tagged MR image sequences from five healthy volunteers. Both acquisitions have been carried out without moving the volunteer inside the scanner, thus avoiding potential misregistration errors due to subject motion between scans. In addition, one of volunteers was subjected to stress during one of the scans. In the final experiment we acquired tagged MR images from a patient with super-ventricular tachyarrhythmia, before and after radio frequency ablation. The image acquisition and catheter intervention were performed with a combined X-ray and MRI system. Detection results were correct on synthetic data and no region was incorrectly classified as having significant changes in the repetition studies. Significant changes in motion pattern were measured in the stress and ablation studies. Furthermore, results seem to corroborate that the ablation regularised cardiac contraction.

This project is a result of a marriage between two independent activities that existed for quite some time within two collaborative groups: (1) the development of a mechanical linkage device and its utilization to test externally the flexibility characteristics of the ankle complex under load; (2) the development of an MR imaging and image analysis methodology to characterize the internal 3D movements of bones of the ankle complex. In the resulting methodology, which we term stress MRI (sMRI for short), the ankle is MR imaged in various foot configurations while held in place by the linkage device with controlled load proven to detect hindfoot instability. Subsequently the acquired images are subjected to a series of image processing and analysis steps to yield a set of parameters to describe the morphology, architecture, and kinematics of the bones of the ankle complex. These parameters are computed from images acquired for 14 normal ankles (of 7 subjects, including the left and the right ankle) and for 8 cadaveric ankles, the latter in five different situations consisting of intact ankle, two ligaments - the CFL and ATFL - sectioned serially, and then after the two ligaments are surgically reconstructed by using two procedures. The results indicate that (i) there is considerable left-to-right symmetry in the ankles; (ii) ligament damage causes a few parameters to change significantly; (iii) both ankle and subtalar motions occur in inversion and anterior drawer; (iv) in vitro motion is generally greater than in vivo motion; (v) the surgical procedures are effective in achieving normalcy, yet there are differences in their performance.

To more precisely measure and monitor bone health, The Johns Hopkins University Applied Physics Lab and School of Medicine have developed the Advanced Multiple Projection Dual-Energy X-Ray Absorptiometry (AMPDXA) scanner. This system provides improvements over conventional DXA scanners in image resolution and multiple projection capability. These improvements allow us to determine structural information about the bone in addition to the standard bone mineral density (BMD) measurements. Algorithms and software were developed to process data acquired from the AMPDXA scanner and to determine important structural parameters, such as the center of mass axis in three dimensions and cross-sectional moments of inertia. The analysis operates on three projections about 15 degrees apart, calculates BMD for each projection, and then combines the data into a three dimensional coordinate system. By knowing the patient position in three dimensions, bone structural parameters are calculated more precisely. Using repeated testing of cadaver bones, the precision of determining these structural parameters is approximately the detector pixel size of 0.127 mm. Data on artificial bone cylinders indicate accuracy of about 3%. Comparisons between bone structural parameters derived from AMPDXA and CT scans show very similar results.

Methods for quantifying hip prosthesis induced changes in the adjacent bone are of great interest to orthopedics. In this work, we present a semi-automated technique for measuring the differences in bone density between the prosthetic and contra lateral hips within a CT volumetric data set. In order to reliably compare the bone-density measurements between the prosthetic and the contra laterals hips, a standardized zoning was developed. Using a spherical model of the outer surface of the prosthetic cup, the superior volume of the acetabular region was subdivided into four distance zones: 0-1mm, 1-6mm, 6-11mm, and 11-16mm, respectively. Furthermore, these regions were divided into four positional zones: medial, lateral, anterior, and posterior. At the same time, the positional zones were divided into four angular regions 15, 30, 45, and 60 degrees from the apex of the acetabular cup. The bone density is computed as the average density in Hounsfield unit (HU) measured from the CT scan using all the voxels within each of the 64 zones. Preliminary analysis has been completed on 3 subjects with total hip replacement. The zonal densities on the prosthetic hip and the contra lateral hip were computed and compared. Contrary to initial expectations, a paired t-test showed no statistical significance between the prosthetic and contra lateral bone-densities at any of the four distance zones. Further analysis with a larger sample subject is needed to detect differences in bone-density between the hips in the stress/weight bearing areas in the 15 to 30 degree regions. A method for reliably and consistently measuring the bone-density within standardized zones has been developed and applied on prosthetic and contra lateral hips. The average bone-density for each of the zones takes into account the entire volumetric data set within that region, which is a considerable improvement over the subjective, user driven region of interest estimate selected within one slice practiced in previous methods. Therefore, our method will offer a more reliable metric in prosthetic changes within the bone. Furthermore, this method may be applicable to analysis of density measurements of different prosthetis and surgical procedures.

Trabecular bone (TB) is a network of interconnected bony struts and plates mostly occurring near the joints of long bones and in the axial skeleton. Several bone diseases including osteoporosis are characterized by fragile bone and increased fracture risk and most fractures occur at locations rich in TB. The mechanical competence of this type of bone can only be partly explained by variations in the bone’s mass density (BMD), and there is now compelling evidence for the role of TB architecture in conferring skeletal strength. Our previous studies have demonstrated that a reduction in BMD is accompanied by a greatly magnified topological process that involves the conversion of trabeculae from plates to struts. Current in vivo technologies yield voxel sizes comparable to TB thickness resulting in inherently fuzzy representations and thus making in vivo assessment of TB architecture challenging. Most existing methods require binarization of an image into bone and non-bone regions and thus are associated with significant errors. Here, a new approach is presented for assessing TB architecture (e.g. classification of plates versus struts) using 3D tensor scale - a local morphometric index - that (1) obviates the need for binary segmentation and is applicable to grayscale bone volume fraction images, and (2) provides precise topological classification over the continuum between a perfect rod and a perfect plate. At any TB location p, tensor scale is the parametric representation of the largest ellipsoid that is centered at p and contained inside the bone region. Accuracy and reproducibility of the method under varying voxel size, and image rotation is presented and its applicability on TB images at in vivo MR resolution is demonstrated.

For practical clinical researches and applications, image segmentation is capable of extracting desired region information plays an important and essential role on a number of medical image preprocessing, including image visualization, malignant tissue recognition, multi-modalities image registration, and so forth. To classify the tissues by physiological characteristics is not satisfactory in high noise functional medical images. In this study, we incorporated both tissue time-activity curves (TACs) and derived "kinetic parametric curves (KPCs)" information to develop a novel image segmentation method for liver tissues classification in dynamic FDG-PET studies. Validation of proposed method, four commmon clustering techniques, include K-mean (KM), Fuzzy C-mean (FCM), Isodata (ISO), Markov Random Fields (MRF), and proposed method were compared to evaluate its precision of segmentation performance. As results, 35.6% and 6.7% less mean errors in mean difference for KPCs and TACs are performed, respectively, than other methods. With combined KPCs and TACs based clustering method can provide the ability to diagnose ill liver tissues exactly.

The purpose of this study is to demonstrate that dynamic CT provides the necessary sensitivity to quantify tumor physiology and differences in chemotherapeutic response. A compartmental mouse model utilizing measured contrast-enhanced dynamic CT scans is used to simulate systematic and statistical errors associated with tumor physiology: perfusion, permeability (PS), fractional plasma volume (f_p), and fractional interstitial volume. The solute utilized is a small molecular weight radio-opaque contrast agent (isovue). For such an intravascular-interstitial medium, the kinematics simplifies to a two compartmental diffusive dominated set of coupled differential equations. Each combination of physiological parameters is repeatedly simulated fifteen times from which statistical errors calculated. The fractional change relative to the true value (systematic error) and standard deviation (statistical error) are plotted as a function of PS, f_p, scanner temporal resolution and noise, and contrast media injection rates. By extrapolating from experimental data found in literature, a relative change in PS and f_p of approximately 40% is required. Thus, the longitudinal response of two chemotherapeutic drugs under investigation - proteasome and IMPDH inhibitors - are hypothesized to induce different physiological responses. The first set of simulations varies PS from 0.05 to 0.40 mL/min/mL and f_p from 0.01 to 0.07 mL/mL while holding all other physiological parameters constant. Errors in PS remain below 3% while statistical errors for f_p increase significantly as the volume decreases toward 1-2%: errors remain less than 6% for f_p>0.03 while increasing to above 15% for f_p<0.02. The second set of simulations are performed quantifying the relationship between scanner temporal resolution and contrast media injection rate for various tumor permeabilities. For the majority of cases, the errors remain below 5%. As PS approaches perfusion, a total error less than 6% can be maintained for a temporal resolution less than or equal to 3 seconds, and an error less than 9% up to 5-7 seconds. As the injection rate decreases from 2 mL/min down to 0.25 mL/min, inadequate sampling of the contrast dynamics necessary to decouple the physiological parameters is lost increasing both systematic and statistical errors from 10% when sampling at 5 seconds in excess of 20-25% at a 9 second sampling rate. In each case, dynamic CT provides the necessary sensitivity to distinguish between the differing therapeutic reponses of proteasome and IMPDH inhibitors.

This paper investigates methods of processing mobility related static images to enhance the effectiveness of a visual prosthesis system. Eight images were processed into 50x50 pixel binary, greyscale, Sobel and Canny edge detected images. 10 subjects were asked 5 mobility related identification tasks for each (randomly ordered) image. Results indicate that edge detection may be useful at this resolution. However, there was not a significant difference found between the results achieved using the Canny and Sobel algorithms. These results support the development of an adaptive device. A mobility display framework has been proposed to assist in this development. Future work will focus on processing image sequences and the development of a visual prosthesis simulation device.