The spinal canal contributes to the overall compliance of the craniospinal compartment. Thus it plays an important role in the regulation of craniospinal hydrodynamics and intracranial pressure. Limited information is available concerning the spinal canal compliance and its distribution along the spinal canal. Current methods of compliance measurement require injection of fluid into the spinal canal cerebrospinal fluid (CSF) spaces and thus are associated with morbidity. A noninvasive method of deriving the spinal canal compliance and its distribution is being developed. A motion-sensitive Magnetic Resonance Imaging technique is employed to quantify the oscillating CSF flow at several locations along the spinal canal. The differential equations governing CSF flow are derived using Bond Graph methodology. Flow dynamics satisfying the differential equations is then compared iteratively with actual flow measurements to yield spinal canal compliance, and CSF resistance and inertia. The model was validated using CSF flow measurements obtained from 4 healthy volunteers. The model predicted CSF flow was compared with measured CSF flow waveforms at intermediate locations. Compliance values ranged from 1.7 mL/mmHg to 45.2 mL/mmHg. The model further provides new information about the relative contribution sub segments of the canal to the overall spinal canal compliance.

This paper presents a methodology to construct realistic patient-specific computational fluid dynamics models of the circle of Willis (CoW) using magnetic resonance angiography (MRA) data. Anatomical models are reconstructed from MRA images using tubular deformable models along each arterial segment and a surface-merging algorithm. The resulting models are smoothed and used to generate finite element (FE) grids. The incompressible Navier-Stokes equations are solved using a stabilized FE formulation. Physiologic flow conditions are derived from phase-contrast MR velocity measurements. The methodology was tested on image data of a normal volunteer. A pulsatile flow solution was obtained. Measured flow rates were prescribed in the internal carotid arteries, vertebral arteries, middle cerebral arteries and anterior cerebral arteries. Pressure boundary conditions were imposed in the posterior cerebral arteries. Visualizations of the complex flow patterns and wall shear stress distributions were produced. Potential applications of these FE models include: study the role of the communicating arteries during arterial occlusions and after endovsascular interventions, calculate transport of drugs, evaluate accuracy of 1D flow models, and evaluate vascular bed models used to impose boundary conditions when flow data is unavailable or incomplete.

Atherosclerotic disease of the renal artery can reduce the blood flow leading to renovascular hypertension and ischemic nephopathy. The kidney responds to a decrease in blood flow by activation of the renin-angiotensin system that increases blood pressure and can result in severe hypertension. Percutaneous translumenal angioplasty (PTA) may be indicated for treatment of renovascular hypertension (RVH). However, direct measurement of renal artery caliber and degree of stenosis has only moderate specificity for detection of RVH. A confounding factor in assessment of the proximal renal artery is that diffuse atherosclerotic disease of the distal branches of the renal artery can produce the same effect on blood-flow as atherosclerotic disease of the proximal renal artery. A methodology is proposed for estimation of pressure gradients at renal artery stenoses from magnetic resonance imaging that could improve the evaluation of renal artery disease. In the proposed methodology, pressure gradients are estimated using computational fluid dynamics (CFD) modeling. Realistic CFD models are constructed from images of vessel shape and measurements of blood-flow rates which are available from magnetic resonance angiography (MRA) and phase-contrast magnetic resonance (MR) imaging respectively. CFD measurement of renal artery pressure gradients has been validated in a physical flow-through model.

In this paper a very fast algorithm for non-invasive intra-vascular pressure measurement using Phase-Contrast MRI data is introduced. A noise robust finite difference numerical method has been developed to quickly calculate the relative static pressure in steady flow by using the Navier-Stokes equation in axisymmetric geometries. We show that in such geometries, pressure calculation may be performed in less than two minutes and by using a single PC MR slice containing the axis of symmetry, in turn resulting in dramatic reductions in the acquisition time. The method has been validated under a variety of additive noise conditions with computational fluid dynamics (CFD) flow simulations within phantom geometries. Effects of noise, resolution, and velocity (flow) are discussed.

Most cardiac diseases, like coronary artery disease or valve defects, are related to wall motion malfunctions. Much research has been done in the estimation of cardiac movement using magnetic resonance (MR) techniques, like MR tagging or phase contrast MR. But echocardiography (cardiac ultrasound) is still the method of choice in clinical routine. Besides visual, subjective wall motion scoring using gray-scale data, myocardial velocities can be obtained using tissue Doppler techniques. A major limitation of Doppler techniques is the one-dimensional estimation of velocities: only the component towards transducer position is measured. Assessing the true velocity field will lead to a more objective measurement. In this paper we describe a method for reconstructing velocity fields using a-priori information about the solution. The feasibility is quantitatively evaluated using simulated data of typical velocity patterns. Translation, contraction and shear movement can be reconstructed well. The proposed method can also be used for the reconstruction of blood flow velocities using color Doppler.

The past few decades have seen phenomenal growth in medical imaging technologies and their impact in biomedicine. This trend will continue, and indeed expand with the development of new techniques. However, another development likely to have major importance is the development of systems offering combined modality imaging in the same platform. Such hybrids include PET/CT, PET/MRI, SPECT/CT, and x-ray/MR systems, and offer not only the convenience of access to multiple modalities in the same session, but more importantly, synergistic uses in which the combined modality is more than the simple sum of previous methods. A simple example is the use of CT data for attenuation correction in PET and SPECT imaging, which is greatly facilitated in a hybrid system. In our own work, we have developed a hybrid x-ray/MR system for image guided interventions. It brings together the high spatial and temporal resolution of x-ray fluoroscopy with the 3D imaging capabilities, soft tissue contrast, and sensitivity to physiological processes of MRI. System development required exploration of a number of interesting problems, including the effect of high magnetic fields on x-ray tubes. The system has been used for several diagnostic and minimally invasive applications, including biopsies, arthrograms, and TIPS procedures.

Using magnetic resonance imaging (MRI), real-time guidance is feasible for radiofrequency (RF) current ablation of pathologic tissue. Lesions have a characteristic two-zone appearance: an inner core (Zone I) surrounded by a hyper-intense rim (Zone II). A better understanding of both the immediate (hyper-acute) and delayed (sub-acute) physiological response of the target tissue will aid development of minimally invasive tumor treatment strategies. We performed in vivo RF ablations in a rabbit thigh model and characterized the tissue response to treatment through contrast enhanced (CE) T1 and T2 weighted MR images at two time points. We measured zonal grayscale changes as well as zone volume changes using a 3D computationally fitted globally deformable parametric model. Comparison over time demonstrated an increase in the volume of both the inner necrotic core (mean 56.5% increase) and outer rim (mean 16.8% increase) of the lesion. Additionally, T2 images of the lesion exhibited contrast greater than or equal to CE T1 (mean 35% improvement). This work establishes a foundation for the clinical use of T2 MR images coupled with a geometric model of the ablation for noninvasive lesion monitoring and characterization.

A fluoroscopy based method for determining heart surface geometry has been developed and validated in phantom and human studies. Biplane fluoroscopic projections were calibrated independently. The heart contour was segmented in each projection and corresponding contour points were matched using epipolar geometry. Points in 3D were reconstructed from the corresponding contour points using point reconstruction. B-splines were approximated from the reconstructed points and meshed to form the heart surface. The fluoroscopy-reconstructed heart was validated in a phantom and human study by comparison to CT imaging. Mean, minimum, maximum and standard deviation of the absolute distance errors were computed for the fluoroscopy-reconstructed heart relative to the CT heart. The mean absolute distance error for the phantom was 4mm. The mean absolute distance error for the human subject was 10 mm. In addition to validating the geometry, we also evaluated in the human subject the feasibility of noninvasive imaging of normal cardiac electrical activity on the fluoroscopy-reconstructed heart by comparing the results to those obtained on the CT heart. Noninvasive images on the fluoroscopy-reconstructed heart by showed close correlation with those obtained on the CT heart (CC=0.70).

This paper quantitatively evaluates observation methods based on undisplayed regions in virtual endoscopy (VE). When we use a VE system as a diagnostic tool, especially as a tool for detecting colorectal polyps, a medical doctor observes the state of the colonic wall. Automated fly-through is often used to reduce system operation load. The system generates automated fly-through paths using the medial axes extracted from colon regions. Because there are many folds in the colon, these folds may cause some blind areas in the views. The aim of this study is to evaluate automated fly-through method in VE based on undisplayed regions. There is no other research on evaluation of undisplayed regions. We evaluate various kinds of automated fly-through paths from the viewpoint of the rate of undisplayed regions. Flattened views are also used for comparison. The experimental results show that about 30% of colon regions are not displayed at all during automated fly-through. The flattened views show smaller rate of undisplayed regions than one of automated fly-through.

Lung cancer assessment involves an initial evaluation of 3D CT image data followed by interventional bronchoscopy. The physician, with only a mental image inferred from the 3D CT data, must guide the bronchoscope through the bronchial tree to sites of interest. Unfortunately, this procedure depends heavily on the physician's ability to mentally reconstruct the 3D position of the bronchoscope within the airways. In order to assist physicians in performing biopsies of interest, we have developed a method that integrates live bronchoscopic video tracking and 3D CT registration. The proposed method is integrated into a system we have been devising for virtual-bronchoscopic analysis and guidance for lung-cancer assessment. Previously, the system relied on a method that only used registration of the live bronchoscopic video to corresponding virtual endoluminal views derived from the 3D CT data. This procedure only performs the registration at manually selected sites; it does not draw upon the motion information inherent in the bronchoscopic video. Further, the registration procedure is slow. The proposed method has the following advantages: (1) it tracks the 3D motion of the bronchoscope using the bronchoscopic video; (2) it uses the tracked 3D trajectory of the bronchoscope to assist in locating sites in the 3D CT "virtual world" to perform the registration. In addition, the method incorporates techniques to: (1) detect and exclude corrupted video frames (to help make the video tracking more robust); (2) accelerate the computation of the many 3D virtual endoluminal renderings (thus, speeding up the registration process). We have tested the integrated tracking-registration method on a human airway-tree phantom and on real human data.

This paper describes a method for generating an unfolded view of organs based on elastic deformation. When we observe the inside of an organ that has a large cavity with a virtual endoscopy system, the viewpoint and the view direction need to be changed many times. Unfolded views can visualize the entire organ wall at a glance and are very useful for diagnosis. The unfolding process requires creating a model that approximates the shape of the target organ. The approximated shape is generated from the outer wall of the organ by extracting the organ wall. The user interactively specifies a cutting line on the approximated shape. The stretching is performed by adding forces and calculating the elastic deformation. The volumetric image where the target organ is unfolded is reconstructed from the original image by using the relation between the approximated and the stretched shapes. We applied the proposed method to six 3-D abdominal CT images. The experimental results showed that the method generates adequate unfolded views of the target organs.

To improve computer aided diagnosis (CAD) for CT colonography we designed a hybrid classification scheme that uses a committee of support vector machines (SVMs) combined with a genetic algorithm (GA) for variable selection. The genetic algorithm selects subsets of four features, which are later combined to form a committee, with majority vote for classification across the base classifiers. Cross validation was used to predict the accuracy (sensitivity, specificity, and combined accuracy) of each base classifier SVM. As a comparison for GA, we analyzed a popular approach to feature selection called forward stepwise search (FSS). We conclude that genetic algorithms are effective in comparison to the forward search procedure when used in conjunction with a committee of support vector machine classifiers for the purpose of colonic polyp identification.

This paper describes a software-based fast volume rendering (VolR) method on a PC platform by using multimedia instructions, such as SIMD instructions, which are currently available in PCs' CPUs. This method achieves fast rendering speed through highly optimizing software rather than an improved rendering algorithm. In volume rendering using a ray casting method, the system requires fast execution of the following processes: (a) interpolation of voxel or color values at sample points, (b) computation of normal vectors (gray-level gradient vectors), (c) calculation of shaded values obtained by dot-products of normal vectors and light source direction vectors, (d) memory access to a huge area, and (e) efficient ray skipping at translucent regions. The proposed software implements these fundamental processes in volume rending by using special instruction sets for multimedia processing. The proposed software can generate virtual endoscopic images of a 3-D volume of 512x512x489 voxel size by volume rendering with perspective projection, specular reflection, and on-the-fly normal vector computation on a conventional PC without any special hardware at thirteen frames per second. Semi-translucent display is also possible.

The effects of CT reconstruction algorithm on density mask score (percentage of voxels < -910 HU) and total lung volume were investigated for emphysema patients with low density mask scores (approximately 20% or below) and patients with higher scores. Based on MTF curves, reconstruction algorithms were classed as standard (i.e. non-enhancing) or over-enhancing. Each image data set was reconstructed with both the standard reconstruction algorithm and the over-enhancing algorithm. All other factors, such as slice collimation and reconstruction interval, were constant. Twenty-nine patients were divided into a high density mask score group (n=10) and low density mask score group (n=19). For the low density mask subgroup, the over-enhancing category yielded an average increase in density mask score of 12.6% compared to standard (i.e. a shift in average score from 14.8% to 27.4%). The maximum shift in score for the low density mask group was 15.9% while the minimum shift was 9.2%. The high density mask group yielded an average shift of 8.7%, with a minimum shift of 3.8% and a maximum shift of 15.3%. The low density group displayed a 1.2% decrease in volume for the over-enhancing category and a 0.8% decrease for the standard category. These volume changes are likely clinically insignificant. Reconstruction algorithm does, however, have a significant effect on the density mask quantitative measure of emphysema. This effect may be significantly larger for density mask scores in patients with smaller amounts of emphysema.

A recently introduced dynamic T1 mapping technique was used to investigate changes of tumor microcirculatory parameters in 16 patients with clinically staged T3) primary rectal carcinoma during a course of preoperative combined chemoradiation. For dynamic T1 mapping an ultra-fast snapshot FLASH T1 mapping sequence was implemented on a 1.5T whole body MR scanner. Acquiring a series of T1 maps contrast media (CM) uptake and washout over an examination time of 40 min was monitored. From the obtained series of T1-maps perfusion-indices (PI) were calculated as the ratio of maximum slope of the tumor CM curve and the maximum of the arterial CM curve. Using pathologic classification of the resected tumors after therapy the patient group could be divided into patients with and without response to therapy. It was found that mean pre-therapy PI values of tumors showing therapy-response were significantly lower than for tumors without no therapy-response. In addition a different behavior of PI distributions within tumors for both groups was observed. The presented study indicates that PI values and their distributions within a tumor seem to be of predictive value for therapy outcome of preoperative therapy in patients with primary rectal carcinoma.

Micro-computed tomography (microCT) is capable of obtaining high-resolution images of skeletal tissues. However its image contrast among soft tissues remains inadequate for tumor detection. High speed functional computed tomography will be needed to image tumors by employing x-ray contrast medium. The functional microCT development will not only facilitate the image contrast enhancement among different tissues but also provide information of tumor physiology. To demonstrate the feasibility of functional CT in mouse imaging, sequential computed tomography is performed in mice after contrast material administration using a high-speed clinical CT scanner. Although the resolution of the clinical scanner is not sufficient to dissolve the anatomic details of rodents, bulky physiological parameters in major organs such as liver, kidney, pancreas, and ovaries (testicular) can be examined. For data analysis, a two-compartmental model is employed and implemented to characterize the tissue physiological parameters (regional blood flow, capillary permeability, and relative compartment volumes.) The measured contrast dynamics in kidneys are fitted with the compartmental model to derive the kidney tissue physiology. The study result suggests that it is feasible to extract mouse tissue physiology using functional CT imaging technology.

Oxidative stress is associated with a variety of pathological processes of clinical relevance. Some of the intermediates generated during the chain reactions associated with oxidation of lipids and proteins are electronically excited and decay emitting photons, which may be detected with the help of sensitive photomultipliers. This technique has been used to monitor oxidative stress in a variety of scenarios including intact organs in vivo or in vitro, and simple models such as proteins and lipids exposed to oxidants. The main drawback of this technique is that the emission of light is extremely weak and it is subjected to substantial interference from spurious sources. In addition, the quantum efficiency of photomultipliers varies with wavelength making it even more difficult to collect reliable data using photomultipliers sensitive to relatively broad spectral ranges. In order to identify the peak emission wavelengths in the visible region, we exposed model systems (proteins, lipids and amino acids) to peroxynitrite and sources of hydroxyl and alcoxyl radicals, analyzing the emission of light with interference filters. The results indicate that the peak emission for most biological models occurs between 450 and 700 nm. The emission at higher wavelengths (lower energy levels) was observed mostly in the presence of less powerful oxidants such as tert-butyl hydroperoxide.

The project was to assess by gamma and MR angiography the bulk variations of chest blood volume related to deep and slow breathing movements. The acquisitions were performed at constant intervals on the widely moving system, without cardiac gating. Two fast enough modalities were used: a gamma-stethoscope working at 30 msec intervals for bulk volumic detection (of ⁹⁹Tc labelled red cells), and MR imaging at 0.5 sec intervals well depicting displacements but not yet performing true angiography. The third modality yielding quantitative imaging was the scintillation gamma camera, but which required 30 sec signal acquisitions for each image. Frames were acquired at 1 sec intervals for up to 30 breathing cycles, and later sorted with double (inspiration and expiration) synchronization for the reconstruction of an average breathing cycle. Convergent results were obtained from the three angiographic modalities, confirming that the deep breathing movements produced inspiratory increases in bulk blood volume and caudal-median displacement of heart and great vessels, and expiratory decreases in blood volume and cranial-left displacement of heart and great vessels. Deep and slow breathing contributed effectively to thoracic blood pumping. The design of a 64x64 channels collimator has been undertaken to speed up the scintillation camera imaging acquisitions.

Multislice CT angiography (MSCTA) is an emerging modality for assessing the coronary arteries. The use of MSCTA for coronary artery disease (CAD) quantification requires an assessment procedure of the coronary arteries that is automated as much as possible. We present an algorithm for the segmentation of the coronary tree with simultaneous extraction of the centerline and the tree-structure. Our approach limits the required user interaction to the placement of one landmark in the left and right main coronary artery respectively. The whole segmentation process takes about 15 s on a mid-sized PC (1GHz) including a real-time visualization of the segmentation in progress. The presented method combines a fast region expansion method (fast marching/front propagation) with heuristic reasoning. The spreading front is monitored for front-splitting enabling branch detection and simultaneous tree reconstruction of the segmented object. This approach allows for the individual treatment of tree-branches with respect to, e.g., threshold settings and reasoning on tree and sub-tree level. This approach can be applied quite generally to the segmentation of tree-like structures. The segmentation results support efficient reporting by enabling automatic generation of overview visualizations, guidance for virtual endoscopy, generation of curved MPRs along the vessels, or cross-sectional area graphs.

Modern micro-CT and multidetector helical CT scanners can produce high-resolution 3D digital images of various anatomical tree structures, such as the coronary or hepatic vasculature and the airway tree. The sheer size and complexity of these trees make it essentially impossible to define them interactively. Automatic approaches, using techniques such as image segmentation, thinning, and centerline definition, have been proposed for a few specific problems. None of these approaches, however, can guarantee extracting geometrically accurate multigenerational tree structures. This limits their utility for detailed quantitative analysis of a tree. This paper proposes an approach for accurately defining 3D trees depicted in large 3D CT images. Our approach utilizes a three-stage analysis paradigm: (1) Apply an automated technique to make a "first cut" at defining the tree. (2) Analyze the automatically defined tree to identify possible errors. (3) Use a series of interactive tools to examine and correct each of the identified errors. At the end of this analysis, in principle, a more useful tree will be defined. Our paper will present a preliminary description of this paradigm and give some early results with 3D micro-CT images.

The functional understanding of the pulmonary anatomy as well as the tracking of the natural course of respiratory diseases are critically dependent on our ability to repeatedly evaluate the same region of the lungs time after time and perform accurate and reliable positionally corresponding measurements. We present a method for accurate labeling of airway branchpoints with their anatomical names as well as an approach for accurate matching of airway tree branchpoints beyond those with anatomical names. An intra-subject tree-matching as well as matching across subjects is achieved. The labeling process is based on matching against a population average. This population average incorporates the anatomical variability that is typically observed across the population. The matching algorithm is based on an association graph method. The computing time is drastically reduced by introducing a hierarchical splitting and only matching two sub-trees at a time. Both steps well tolerate possible false branches. Validation against an independent standard provided by human experts shows a high degree of accuracy (> 90%) for both labeling and matching. The average error compares well to the inter-observer variability among human experts.

To study the relationship between transpulomnary pressure (Ptp), intravascular pressure (Pv), and the pulmonary arterial tree structure, morphometric measurements of pulmonary arterial trees were made in intact lungs from Sprague-Dawley rats. Using cone beam micro-CT and techniques we developed for imaging small animal lungs, volumetric CT data were acquired for Ptp from 0 - 12 mmHg and Pv from 5 - 30 mmHg. The diameter, D (measured range approximately 0.08-2.0 mm), vs. pressure, P, relation can be described by D(P) = D(0)(1+ α P), where α is a distensibility coefficient. Unlike studies performed in larger animals, where changes in either Ptp or Pv had nearly identical effect on vessel distensibility, we found that there is only a small dependence of arterial diameter on Ptp in the rat. For example, using the above relation where P=Ptp and Pv is held constant at 12mmHg, alpha = 0.55±0.42(SE) %/mmHg, compared with when P=Pv and Ptp is held at 12mmHg, alpha = 2.59±0.17(SE) %/mmHg.

Plaque in native coronary arteries is hypothesized to accumulate more likely along the inner curvature of a vessel segment as compared to its outer curvature. This behavior is likely associated with differences in local shear stress, which tends to be lower on the inner bend of a curved vessel than on the outer bend. The reported in-vivo study evaluated how the circumferential plaque distribution depends on local vessel curvature in coronaries from a limited set of 12 patients. Geometrically correct models of the vessel segments were generated utilizing fusion between biplane angiography and intravascular ultrasound. The plaque thickness was derived from the 3-D borders of the lumen/plaque and media/adventitia interfaces. Within each frame, plaque thickness was classified into "below-average" and "above-average" regions. A local curvature index was defined for each point: A positive value indicates the "inner" curvature, a negative value the "outer" curvature, with the magnitude determined from differential geometry. In the majority of the examined vessels, regions of "below-average/outer-curvature" and "above-average/inner-curvature" combined outweighed the "below-average/inner-curvature" and "above-average/outer-curvature" regions. The ratio increased with the threshold to exclude lower-curvature regions, confirming the hypothesis that plaque is more likely to accumulate on the luminal surface along the inner curvature of the coronary segment.

In living organs, microcirculation in the capillaries and high order branches can be seen as a macroscopically random process. The Intra Voxel Incoherent Motion (IVIM) method uses a diffusion-weighted magnetic resonance imaging sequence to register this pseudo-random motion. It is able to observe perfusion in addition to the brownian diffusion by its relatively large distance of movement. The dependence of the MR signal (S) on the diffusion weighting b can be approximated as a bi-exponential relation: (S/S₀)=(1-f).exp(-bD)+f.exp[-b(D+D*)], where S₀ is the signal intensity for b=0, f the vascular volume fraction, D the molecular diffusion coefficient and D* a flow index. This effect, largely investigated in the brain, has never been applied in the heart, where the diffusion-weighted sequence is highly sensitive to bulk motion. We have studied microcirculation in the canine heart in vivo, with a well-controlled cardiac and respiratory gating protocol that overcomes the bulk motion effects. We demonstrated that the IVIM effect could be applied in the myocardium. The IVIM parameters were found equal to D=1.26*10^-3 mm²/s, f=11.98%, D*=12.87*10^-3 mm²/s. Moreover, the microcirculation is directionally anisotropic. The preferred direction of capillaries/small vessels is aligned with the myofibers in mid-myocardium in the left ventricle.

Purpose: To evaluate whether functional multi-slice computed tomography (MSCT) can identify regional areas of normally perfused and ischemic myocardium in a porcine model. Material and Methods: Three out bred pigs, two of which had ameroids surgically implanted to constrict flow within the LAD and LCx coronary arteries, were injected with 25 mL of iopromide (Isovue) at a rate of 5 mL/second via the femoral or jugular vein. Sixty axial scans along the short axis of the heart was acquired on a 16-slice CT scanner (Philips MX8000-IDT) triggered at end-diastole of the cardiac cycle and acquiring an image within 270 msec. A second series of scans were taken after an intravenous injection of a vasodilator, 150 μg/kg/min of adenosine. ROIs were drawn around the myocardial tissue and the resulting time-density curves were used to extract perfusion values. Results: Determination of the myocardial perfusion and fractional blood volume implementing three different perfusion models. A 5-point averaging or 'smoothing' algorithm was employed to effectively filter the data due to its noisy nature. The (preliminary) average perfusion and fractional blood volume values over selected axial slices for the pig without an artificially induced stenosis were measured to be 84 ± 22 mL/min/100g-tissue and 0.17 ± 0.04 mL/g-tissue, the former is consistent with PET scan and EBCT results. The pig with a stenosis in the left LAD coronary artery showed a reduced global perfusion value -- 45 mL/min/100g-tissue. Correlations in regional perfusion values relative to the stenosis were weak. During the infusion of adenosine, averaged perfusion values for the three subjects increased by 46 (±45) percent, comparable to increases measured with PET. Conclusion: Quantifying global perfusion values using MDCT appear encouraging. Future work will focus resolving the systematic effects from noise due to signal fluctuation from the porcine tachyardia (80-93 BPM) and provide a more robust measurement of regional myocardial perfusion throughout the heart.

The measurement of time-density relationships of the myocardium in studies of Magnetic Resonance perfusion data sets is a clinical technique used in assessing myocardial perfusion. Traditionally, to measure the time-density relationship a physician draws a region on the same 2-D image of the myocardium in sequential cardiac cycles. Throughout multiple cardiac cycles the density changes in this region are measured. A major limitation of this technique is change in anatomy relative to the selected region on the myocardium during consecutive cardiac cycles. This causes measurement errors, which are amplified if the traced region does not encompass the entire myocardial thickness, or includes a boundary exterior to the epicardial or endocardial surface. The technique described in this paper uses approximately the same myocardial region throughout the entire perfusion study, which insures inclusion of the entire endocardial to epicardial region and exclusion of exterior regions. Moreover, this region can be subdivided into smaller regions of interest. This can be accomplished by careful segmentation and reformatting of the data into polar coordinates. This allows sectioning both axially and transaxially through the myocardium permitting regional assessment of perfusion specific values such as maximum and/or the time to reach maximum density. These values can then be illustrated using density-mapped colors or time-density curves. This measurement and display technique may provide enhanced detection and evaluation of regional deficits in myocardial contractility and perfusion.

Post myocardial infarction, the identification and assessment of non-viable (necrotic) tissues is necessary for effective development of intervention strategies and treatment plans. Delayed Enhancement Magnetic Resonance (DEMR) imaging is a technique whereby non-viable cardiac tissue appears with increased signal intensity. Radiologists typically acquire these images in conjunction with other functional modalities (e.g., MR Cine), and use domain knowledge and experience to isolate the non-viable tissues. In this paper, we present a technique for automatically segmenting these tissues given the delineation of myocardial borders in the DEMR and in the End-systolic and End-diastolic MR Cine images. Briefly, we obtain a set of segmentations furnished by an expert and employ an artificial intelligence technique, Support Vector Machines (SVMs), to "learn" the segmentations based on features culled from the images. Using those features we then allow the SVM to predict the segmentations the expert would provide on previously unseen images.

Small animal imaging is experiencing rapid development due to its importance in providing high-throughput phenotypic data for functional genomics studies. We have developed a single photon emission computed tomography (SPECT) system to image the pulmonary perfusion distribution in the rat. A standard gamma camera, equipped with a pinhole collimator, was used to acquire SPECT projection images at 40 sec/view of the rat thorax following injection of Tc99m labeled albumin that accumulated in the rat's lungs. A voxel-driven, ordered-subset expectation maximization reconstruction was implemented. Following SPECT imaging, the rat was imaged using micro-CT with Feldkamp conebeam reconstruction. The two reconstructed image volumes were fused to provide a structure/function image of the rat thorax. Reconstruction accuracy and performance were evaluated using numerical simulations and actual imaging of an experimental phantom consisting of Tc99m filled chambers with known diameters and count rates. Full-width half-maximum diameter measurement errors decreased with increasing chamber diameter, ranging from < 6% down to 0.1%. Errors in the ratio of count rate estimates between tubes were also diameter dependent but still relatively small. This preliminary study suggests that SPECT will be useful for imaging and quantifying the pulmonary blood flow distribution and the distribution of Tc99m labeled ligands in the lungs of small laboratory animals.

Conventional CT imaging methods lack the accuracy that is necessary for detailed assessment of liver metastasis. However, quantitative measurement of hepatic perfusion has the potential to provide important information necessary for both the evaluation and treatment of liver metastasis. VX2 carcinoma cells were injected into the liver of healthy male rabbits. A two-phase scan protocol was used to acquire CT images for the determination of liver perfusion prior to and post implantation. Contrast enhancement curves from the aorta, portal vein and liver parenchyma were obtained from the reconstructed images. The weighted summation of the aortic and portal venous curve were de-convolved against the liver parenchymal curve to derive functional parameters such as total hepatic blood flow (HBF) and hepatic arterial fraction (HAF). Results show that the HAF of the implanted tumor increased throughout the period of the study. Twelve days after tumors were implanted the HAF in the liver significantly increased (P < 0.05) from the normal tissue value of 35 ± 1 to 57 ± 9 %. Functional maps of the HAF have the potential to improve treatment outcome of patients owing to the earlier diagnosis of liver cancer. We are able to detect VX2 tumors in the liver as early as 8 days after they were implanted.

Alzheimer's is a progressive brain disease and is clinically characterized by cognitive symptoms that, in combination with behavioral disturbances, significantly interfere with activities of daily living. The purpose of this study is to investigate the possibility of developing volumetric measures of the structural damage and atrophy of brain derived from multiprotocol MR imaging. Our approach first applies intensity inhomogeneity correction and intensity standardization to PD and T2 weighted MR images to create base images for quantitative image analysis. Then, vectorial scale-based fuzzy connectedness segmentation (VSFCS) and morphological operations are applied to the base images to extract masks of cerebrospinal fluid (CSF), grey matter (GM), and white matter (WM), and further to create a clean and accurate intracranial (IC) mask. After separating CSF from brain parenchyma (BP), VSFCS is applied to BP (PD and T2) images to generate pure GM and WM masks, and then subtracting these pure from the BP mask to detect AD lesions. This method was applied to a set of conventional PD and T2 weighted MR images that were obtained from 5 patients with probable AD and 5 healthy normal control subjects. The segmented images of individual brain tissue regions (CSF, GM, WM, and AD lesion) are consistent with a Neuroradiologist's examination. The quantitative analysis shows that patients with AD have more atrophy. The mean value of the volume of brain parenchyma of patients with AD is about 10% less than that of healthy controls.

This paper presents our recent study to evaluate how effectively the image texture information within the hippocampus structure can help the physicians to determine the candidates for epilepsy surgery. First we segment the hippocampus from T1-weighted images using our newly developed knowledge-based segmentation method. To extract the texture features we use multiwavelet, wavelet, and wavelet packet transforms. We calculate the energy and entropy features on each sub-band obtained by the wavelet decomposition. These texture features can be used by themselves or along with other features such as shape and average intensity to classify the hippocampi. The features are calculated on the T1-weighted and FLAIR MR images. Using these features, a clustering algorithm is applied to classify each hippocampus. To find the optimal basis, we use several different bases for wavelet and multiwavelet transforms, and compare the final classification performances, which is evaluated by correct classification rate (CCR). We use MRI of 14 epileptic patients along with their EEG results in our study. We use the pre-operative MR images of the patients who have already been determined as candidates for an epilepsy surgery using the gold standard (more costly and painful) methods of EEG phase II study. Experimental results show that the texture features may predict the candidacy for epilepsy surgery. If successful in large population studies, the proposed non-invasive method can replace invasive and costly EEG studies.

Multiple initiatives in the pharmaceutical and beauty care industries are directed at identifying therapies for weight management. Body composition measurements are critical for such initiatives. Imaging technologies that can be used to measure body composition noninvasively include DXA (dual energy x-ray absorptiometry) and MRI (magnetic resonance imaging). Unlike other approaches, MRI provides the ability to perform localized measurements of fat distribution. Several factors complicate the automatic delineation of fat regions and quantification of fat volumes. These include motion artifacts, field non-uniformity, brightness and contrast variations, chemical shift misregistration, and ambiguity in delineating anatomical structures. We have developed an approach to deal practically with those challenges. The approach is implemented in a package, the Fat Volume Tool, for automatic detection of fat tissue in MR images of the rat abdomen, including automatic discrimination between abdominal and subcutaneous regions. We suppress motion artifacts using masking based on detection of implicit landmarks in the images. Adaptive object extraction is used to compensate for intensity variations. This approach enables us to perform fat tissue detection and quantification in a fully automated manner. The package can also operate in manual mode, which can be used for verification of the automatic analysis or for performing supervised segmentation. In supervised segmentation, the operator has the ability to interact with the automatic segmentation procedures to touch-up or completely overwrite intermediate segmentation steps. The operator's interventions steer the automatic segmentation steps that follow. This improves the efficiency and quality of the final segmentation. Semi-automatic segmentation tools (interactive region growing, live-wire, etc.) improve both the accuracy and throughput of the operator when working in manual mode. The quality of automatic segmentation has been evaluated by comparing the results of fully automated analysis to manual analysis of the same images. The comparison shows a high degree of correlation that validates the quality of the automatic segmentation approach.

Continuing with our previous work of the segmentation and delineation of upper airway, the purpose of this work is to segment and delineate soft tissue organs surrounding the upper airway, such as adenoid, tonsils, fat pads and tongue, with the further goal of studying the relationship among the architectures of these structures, for understanding upper airway disorders in children. We use two MRI protocols, Axial T2 (used for adenoid, tonsil, and fat pads) and sagittal T1 (for tongue), to gather information about different aspects of the tissues. MR images are first corrected for background intensity variation and then the intensities are standardized. All segmentations are achieved via fuzzy connectedness algorithms with only limited operator interaction. A smooth 3D rendition of the upper airway and its surrounding tissues is displayed. The system has been tested utilizing 20 patient data sets. The tests indicate a 95% or better precision and accuracy for segmentation. The mean time taken per study is about 15 minutes including operator interaction time and processing time for all operations. This method provides a robust and fast means of assessing sizes, shapes, and the architecture of the tissues surrounding the upper airway, as well as providing data sets suitable for use in modeling studies of airflow and mechanics.

The positions of the lobar fissures are of growing interest as computer-based quantitative measures to detect early pathologies and to predict or measure outcomes emerge. While we have developed a semi-automatic fissure detection method in our previous work, in this paper we describe the use of an anatomic pulmonary atlas with a priori knowledge about lobar fissures to automatically segment the lobar fissures. 16 volumetric CT scans from 16 subjects are used to construct the pulmonary atlas. After deforming the fissures onto a template image, the average fissure and variability between different subjects can be obtained by local statistical measures. The probabilistic analysis for the atlas shows that the atlas can provide an initialization for the fissure detection in certain regions with a predictable variation, although the initialization may not be close and complete. A ridgeness measure is applied on original images to enhance the fissure contrast. The fissure detection is accomplished by the initial fissure search and the final fissure search. While only parts of the initial search results are correctly delineated, a regional statistic analysis of ridgeness selects the most "reliable" initial search results, which are then used to initialize the final search. Our method has been tested in 22 volumetric thin-slice CT images from 12 subjects, and the results are compared to manual tracings. The mean of the similarity indices between the manual and computer defined lobes is 0.988. The results indicate a strong agreement between the automatic and manual lobe segmentations.

Conventional BOLD fMRI has limited use in overt speech paradigms, due to movement and susceptibility artifacts. Our study used an arterial spin-tagging (AST) sequence to quantify focal brain activation in a continuous speech task. Furthermore, we compared the results to conventional BOLD fMRI. The ASSIST sequence was used to obtain transverse perfusion images of the brain, acquired on a 1.5T GE-Signa scanner. Three conditions were alternated in a block design: generation of complete sentences, nonsense syllables and rest with continuous and overt speech production. For 4 normal volunteers, task-related perfusion maps with quantified rCBF and rCBV values were calculated and activations were mapped to the MNI brain. The same paradigm was scanned with BOLD contrast fMRI in separate, independent scans and data from 6 subjects were analyzed using SPM99. Using the AST sequence, we could reliably identify focal brain activation in an overt continuous speech paradigm, and the activations observed were consistent with previous PET studies. We found differential activation at increasing levels of speech production with a focus in the left insula and opercular IFG related to the production of sentences at a syntactic level as opposed to nonsense syllable production. The BOLD technique failed to identify some of these activation foci, possibly due to decreased SNR and artifacts.

The advent of event-related functional magnetic resonance imaging (fMRI) has resulted in many exciting studies that have exploited its unique capability. However, the utility of event-related fMRI is still limited by several technical difficulties. One significant limitation in event-related fMRI is the low signal-to-noise ratio (SNR). In this work, a new non-parametric technique for noise suppression in event related fMRI data is proposed based on spectrum subtraction. The new technique is based on generalized spectral subtraction that allows correlated noise components to be treated robustly. Moreover, it adaptively estimates a nonparametric model for random and physiological components of noise from the acquired data in a simple and computationally efficient manner. This allows the new method to overcome the limitations of previous methods while maintaining a robust performance given its fewer assumptions and suggests its value as a useful preprocessing step for fMRI data analysis.

The application of independent components analysis (ICA) to functional magnetic resonance imaging data has been proven useful to decompose the signal in terms of its basic sources. The main advantage is that ICA requires no prior assumption about the neuronal activity or the noise structure, which are usually unknown in fMRI. This enables the detection of true activation components free of random and physiological noise. Hence, this technique is superior to other techniques such as subspace modeling or canonical correlation analysis, which have underlying assumptions about the signal components. Nevertheless, this technique suffers from a fundamental limitation of not providing a consistent ordering of the signal components as a result of the whitening step involved in ICA. This mandates human intervention to pick out the relevant activation components from the outcome of ICA, which poses a significant obstacle to the practicality of this technique. In this work, a simple yet robust technique is proposed for ranking the resultant independent components. This technique adds a second step to ICA based on canonical correlation analysis and the prior information about the activation paradigm. This enables the proposed technique to provide a consistent and reproducible ordering of independent components. The proposed technique was applied to real event-related functional magnetic resonance imaging data and the results confirm the practicality and robustness of the proposed method.

Functional Magnetic Resonance Imaging (fMRI) is a powerful technique for studying the working of the human brain. This overall goals of the project are to devlop a novel method for the analysis of fMRI data in order to discover the activation of a network of regions involving most likely the hippocampus, parietal cortex and cerebellum as a person is navigating in a virtual environment. Spatially sensitive voxels are extracted by selecting voxels that have high mutual information. Each of these extracted voxels is then used to create a response curve for the stimulus of interest, in this case spatial location. Following the voxel extraction stage, the set of extracted voxel time series would be treated as a population and used to predict the location of the subject at any randomly selected time in the experiment. The population of voxels essentially "votes" with their current activity. The approach used for prediction is the Bayesian reconstruction method. The ability to predict the location of a subject in the virtual environment based on brain signals will be useful in developing a physiological understanding of spatial cognition in virtual environments.

Brain imaging and particular functional MRI (fMRI), which acquires brain volumes in time, reveals new understanding of the functional/structural relation in neuroscience. During fMRI imaging physiological state changes occur in the brain regions activated from the task paradigm which the subject performs in the scanner. These state changes can be depicted in the small veins of the activated region due to the blood oxygen level dependent (BOLD) effect. For each brain voxel in the fMRI experiment one accumulates a time series vector which has to be analyzed for similarity to the original task paradigm vector and its characteristic hemodynamic response function (HRF). Various analysis methods have been discussed for fMRI analysis, model-based statistical or unsupervised data-driven techniques. The purpose of this paper is to introduce a new method which combines two different approaches. We use an unsupervised self-organizing map (SOM) neural network to reduce the time series vector space by non-linear pattern recognition into a 2D table of representative time series wave-forms. Using a-priori knowledge of the HRF, either derived from a theoretical wave-form model or estimated from a brain region of interest (ROI), one can use correlation analysis between the time series patterns of the SOM table and the HRF to depict regions of activation specific to the HRF. An optional second SOM training with a reduce number of neurons of the best-matching time series to the HRF classification refines the second neural network pattern table. The learned time series pattern of each neuron and the corresponding brain voxels are superimposed onto the subject's brain image for visual investigation.

We have developed a new method employing the Canny edge detector and Radon transformation to segment images of polyp candidates for CT colonography (CTC) computer aided polyp detection and obtain features useful for distinguishing true polyps from false positive detections. The technique is applied to two-dimensional subimages of polyp candidates selected using various 3-D shape and curvature characteristics. We detect boundaries using the Canny operator. The baseline of the colon wall is detected by applying the Radon transform to the edge image and locating the strongest peak in the resulting transform matrix. The following features are calculated and used to classify detections as true positives (TP) and false positives (FP): polyp boundary length, polyp base length, polyp internal area, average intensity, polyp height, and inscribed circle radius. The segmentation technique was applied to a data set of 15 polyps larger than 3 mm and 617 false positives taken from 80 CTC studies (supine and prone screening of 40 patients). The sensitivity was 100% (15 of 15). 58% of the FP's were eliminated leaving an average of 3 false positives per study. Our method is able to segment polyps and quantitatively measure polyp features independently of orientation and shape.

An automatic method to segment colonic polyps from CT colonography is presented. The method is based on a combination of knowledge-guided intensity adjustment, fuzzy c-mean clustering, and deformable models. The input is a set of polyp seed points generated by filters on geometric properties of the colon surface. First, the potential polyp region is enhanced by a knowledge-guided adjustment. Then, a fuzzy c-mean clustering is applied on a 64*64 pixel sub-image around the seed. Fuzzy membership functions for lumen air, polyp tissues and other tissues are computed for each pixel. Finally, the gradient of the fuzzy membership function is used as the image force to drive a deformable model to the polyp boundary. The segmentation process is first executed on the 2D transverse slice where the polyp seed is located, and then is propagated to neighboring slices to construct a 3D representation of the polyp. Manual segmentation is performed on the same polyps and treated as the ground truth. The automatically generated segmentation is compared with the ground truth segmentation to validate the accuracy of the method. Experimental results showed that the average overlap between the automatic segmentation and manual segmentation is 76.3%. Given the complex polyp boundaries and the small size of the polyp, this is a good result both visually and quantitatively.

We developed a method for generating the centerline of a colon in CT Colonography that is computationally fast, and robust to collapsed regions. Patients underwent CT Colonography after standard pre-colonoscopy cleansing. The colonic lumen was segmented using an existing anatomy-based approach, and a distance map of the colonic lumen was computed using a distance transform. The centerline was computed as follows: Local maxima representative for the centerline were sparsely extracted from the distance map. Iteratively, each pair of maxima satisfying a set of connection criteria were connected, creating a graph-like structure containing a main centerline with additional branches. Branches were later removed and the resulting centerline was stored. Centerlines of the colon were computed, and also manually and independently drawn by two radiologists, for 33 CT Colonographic data sets. The data sets were chosen to give a wide spectrum of colons, ranging from cases with good segmentation and extension to cases with collapsed regions and numerous extra-colonic components such as small bowel. On average, 94% of the human-generated centerlines were correctly identified by the computer-generated centerlines. The average displacement between the human- and computer-generated centerlines was 4.0 mm. Average centerline computation time was less than 4 seconds.

The purpose of this paper is to assess colonic distension and length and to evaluate a polyp matching algorithm on prone and supine studies in CT colonography (CTC) computer aided diagnosis (CAD). Our CTC CAD software was used to reconstruct the colonic surface, to compute the colon centerline and to detect spatial positions of polyps. Normalized distance along the centerline (NDAC) and colon distension at each polyp were compared in prone and supine positions. 8 patients presented 12 polyps detected in both supine and prone positions and confirmed by colonoscopy. NDAC for all polyps were very similar in supine/prone positions. The average NDAC differences were 0.01 ± 0.01 which indicates excellent agreement. Colon profiles (distension vs. DAC) were not similar in supine/prone positions due to high variability of colonic distension and mobility of the colon when the patient was repositioned. However the final colon lengths and average distension were very similar (average difference was 4.5%). Also, as could be expected from normal anatomy, the cecum and rectum tended to have the greatest distension and the rectum tended to be better distended on the prone exam. The colon centerline provides a natural coordinate reference and is a useful tool for CTC CAD.

Recently, CT colonography has been recognized as an effective option for evaluating colorectal polyps in the USA. We have applied this technique to preoperative staging of colorectal cancer patients with a contrast-enhanced multi-detector row CT (MDCT). The use of manipulated multi-planar reconstruction (MPR) views in contrast-enhanced MDCT colonography proved advantageous for detecting lymph node metastases. Furthermore, 3-dimensional (3D) endoluminal images with Hansfield-transparency settings allowed vascular views of the colorectal wall for identification of invasive colorectal cancers. Using endoluminal images, increase in flow and pooling of blood related to angiogenesis of invasive cancer could be demonstrated, not only in the lymph nodes but also in the colorectal wall. Both MPR views and 3D endoluminal images can be acquired from the same 3D volumetric data generated by helical scanning in MDCT colonography, and both have great potential as modalities for computer-aided diagnosis (CAD) using blood flow information. Therefore the use of CAD can be expected to improve radiologists' diagnostic performance with regard to colorectal cancer.

Currently, virtual colonoscopy examinations require extensive bowel preparation because residual materials can occlude lesions or can be misinterpreted as polyps. Our goal is to investigate a probabilistic method to segment contrast enhanced residual materials and remove them from the rendering. The region around a sample position is modeled to contain mixtures of air, tissue and tagged intraluminal remains. For each image sample a probability vector is calculated expressing the probability that the materials of interest are present. A probability space is defined using the probabilities for pure materials as base vectors. Mixture vectors are constructed at 45-degree angles between the pure material vectors. The probability vectors are compared to the base vectors and the mixture vectors to classify them into material mixtures. Consider the layer between air and tagged fluid. Image intensities are similar to tissue. The scale at which the Gaussian averaged probability is calculated is increased until convergence: two successive scales result in the same classification. The Bayesian classification method shows good results with relatively large objects. However, edges of small or thin objects are likely to be misclassified: a too large environment is needed for convergence.

To study the application of diffusion weighted imaging and image post processing in the diagnosis of stroke, especially in acute stroke, 205 patients were examined by 1.5 T or 1.0 T MRI scanner and the images such as T1, T2 and diffusion weighted images were obtained. Image post processing was done with "3D Med System" developed by our lab to analyze data and acquire the apparent diffusion coefficient (ADC) map. In acute and subacute stage of stroke, the signal in cerebral infarction areas changed to hyperintensity in T2- and diffusion-weighted images, normal or hypointensity in T1-weighted images. In hyperacute stage, however, the signal was hyperintense just in the diffusion weighted imaes; others were normal. In the chronic stage, the signal in T1- and diffusion-weighted imaging showed hypointensity and hyperintensity in T2 weighted imaging. Because ADC declined obviously in acute and subacute stage of stroke, the lesion area was hypointensity in ADC map. With the development of the disease, ADC gradually recovered and then changed to hyperintensity in ADC map in chronic stage. Using diffusion weighted imaging and ADC mapping can make a diagnosis of stroke, especially in the hyperacute stage of stroke, and can differentiate acute and chronic stroke.

To study the technique and application of perfusion weighted imaging (PWI) in the diagnosis and medical treatment of acute stroke, 25 patients were examined by 1.5 T or 1.0 T MRI scanner. The Data analysis was done with "3D Med System" developed by our Lab to process the data and obtain apparent diffusion coefficient (ADC) map, cerebral blood volume (CBV) map, cerebral blood flow (CBF) map as well as mean transit time (MTT) map. In accute stage of stroke, normal or slightly hypointensity in T1-, hyperintensity in T2- and diffusion-weighted images were seen in the cerebral infarction areas. There were hypointensity in CBV map, CBF map and ADC map; and hyperintensity in MTT map that means this infarct area could be saved. If the hyperintensity area in MTT map was larger than the area in diffusion weighted imaging (DWI), the larger part was called penumbra and could be cured by an appropriate thrombolyitic or other therapy. The CBV, CBF and MTT maps are very important in the diagnosis and medical treatment of acute especially hyperacute stroke. Comparing with DWI, we can easily know the situation of penumbra and the effect of curvative therapy. Besides, we can also make a differential diagnosis with this method.

Bubble emphysema is a disease characterized by the presence of air bubbles within the lungs. With the purpose of identifying pulmonary air bubbles, two alternative methods were developed, using High Resolution Computer Tomography (HRCT) exams. The search volume is confined to the pulmonary volume through a previously developed pulmonary contour detection algorithm. The first detection method follows a slice by slice approach and uses selection criteria based on the Hounsfield levels, dimensions, shape and localization of the bubbles. Candidate regions that do not exhibit axial coherence along at least two sections are excluded. Intermediate sections are interpolated for a more realistic representation of lungs and bubbles. The second detection method, after the pulmonary volume delimitation, follows a fully 3D approach. A global threshold is applied to the entire lung volume returning candidate regions. 3D morphologic operators are used to remove spurious structures and to circumscribe the bubbles. Bubble representation is accomplished by two alternative methods. The first generates bubble surfaces based on the voxel volumes previously detected; the second method assumes that bubbles are approximately spherical. In order to obtain better 3D representations, fits super-quadrics to bubble volume. The fitting process is based on non-linear least squares optimization method, where a super-quadric is adapted to a regular grid of points defined on each bubble. All methods were applied to real and semi-synthetical data where artificial and randomly deformed bubbles were embedded in the interior of healthy lungs. Quantitative results regarding bubble geometric features are either similar to a priori known values used in simulation tests, or indicate clinically acceptable dimensions and locations when dealing with real data.

In clinical research, non-invasive MR perfusion imaging is capable of investigating brain perfusion phenomenon via various hemodynamic measurements, such as cerebral blood volume (CBV), cerebral blood flow (CBF), and mean trasnit time (MTT). These hemodynamic parameters are useful in diagnosing brain disorders such as stroke, infarction and periinfarct ischemia by further semi-quantitative analysis. However, the accuracy of quantitative analysis is usually affected by poor signal-to-noise ratio image quality. In this paper, we propose a hemodynamic measurement method based upon sub-band denoising and spline curve fitting processes to improve image quality for better hemodynamic quantitative analysis results. Ten sets of perfusion MRI data and corresponding PET images were used to validate the performance. For quantitative comparison, we evaluate gray/white matter CBF ratio. As a result, the hemodynamic semi-quantitative analysis result of mean gray to white matter CBF ratio is 2.10 ± 0.34. The evaluated ratio of brain tissues in perfusion MRI is comparable to PET technique is less than 1-% difference in average. Furthermore, the method features excellent noise reduction and boundary preserving in image processing, and short hemodynamic measurement time.

Functional medical imaging, such as PET or SPECT, is capable of revealing physiological functions of the brain, and has been broadly used in diagnosing brain disorders by clinically quantitative analysis for many years. In routine procedures, physicians manually select desired ROIs from structural MR images and then obtain physiological information from correspondent functional PET or SPECT images. The accuracy of quantitative analysis thus relies on that of the subjectively selected ROIs. Therefore, standardizing the analysis procedure is fundamental and important in improving the analysis outcome. In this paper, we propose and evaluate a normalization procedure with a standard 3D-brain model to achieve precise quantitative analysis. In the normalization process, the mutual information registration technique was applied for realigning functional medical images to standard structural medical images. Then, the standard 3D-brain model that shows well-defined brain regions was used, replacing the manual ROIs in the objective clinical analysis. To validate the performance, twenty cases of I-123 IBZM SPECT images were used in practical clinical evaluation. The results show that the quantitative analysis outcomes obtained from this automated method are in agreement with the clinical diagnosis evaluation score with less than 3% error in average. To sum up, the method takes advantage of obtaining precise VOIs, information automatically by well-defined standard 3-D brain model, sparing manually drawn ROIs slice by slice from structural medical images in traditional procedure. That is, the method not only can provide precise analysis results, but also improve the process rate for mass medical images in clinical.

The purpsoe was to evaluate the influence of a right-sided pleural effusion on the lung aeration dynamics in the respiratory cycle during pressure controlled ventilation. Pleural effusion was simulated by infusion of 3% gelatin into the pleural cavity in steps of 300ml totaling 1200ml in four anesthetized pigs. After each step, volume scans and respirator gated 50ms scans at a constant table position (carina niveau) were taken. The dynamic changes of the previously defined air-tissue ratios (in steps of 100HU) were evaluated in three separate regions of left and right lung: a ventral, an intermediate and a dorsal area. The affected side revealed dramatic alveolar collapse. There was a shift of the lung density to higher air-tissue ratios (+200HU) but showing the same air-tissue ratio dynamics. A slight lateral shift of 32mm (±14mm) the mediastinum was measured. The unaffected side showed no increase in the air-tissue ratios caused by hyperinflation but an increase of density due to mediastinal shift. Air-tissue ratio dynamics remained unchanged on the unaffected side compared to baseline measurements. We visualized the ventilation mismatch caused by pleural effusion. The contra-lateral lung is not affected by unilateral pleural effusion. Pressure controlled ventilation prevents hyper-inflation of non-dependent lung areas.

Quantitative assessment of MR examinations in 37 survivors of childhood cancer treated with central nervous system prophylaxis revealed that normal appearing white matter (NAWM) volume is associated with attention-related problems, localized specifically in the right prefrontal region. T1-, T2-, and PD-weighted images were segmented and divided into pre-frontal, frontal, parietal/temporal, and parietal/occipital regions for each hemisphere. These eight regions were analyzed in five slices centered at the level of the basal ganglia. The patient's age at diagnosis and time elapsed from diagnosis were used as covariates in the regressions. Attentional measures showed significant deficiency when compared to age and gender normative values. Total, frontal and/or prefrontal NAWM volumes from the range of slices examined were significantly associated with 5 of the 8 attentional measures. The frontal/prefrontal region of the brain is associated with executive functioning tasks and could potentially be spared as much as possible during therapy planning. The results of the present study further support the contention that NAWM is an important substrate for treatment-induced neurocognitive problems among survivors of malignant brain tumors of childhood.

Electrical impedance tomography (EIT), can produce a cross-sectional image that allows non-invasive assessment of resistivity distribution in a measured object. Since its introduction more than two decades ago, EIT has attracted considerable interests. However, its clinical application has been constrained, to some extent, by the difficulties in interpreting the reconstructed images, due to their low resolution. To facilitate the application of EIT to modeling human physiological functions, in this study, we proposed a method to quantify the changes of EIT images relative to anatomical structure. Based on ECG-gated MRI images, a 3D human thorax model was developed. EIT measurements of the thorax model were simulated by finite difference method and images were reconstructed using the filtered back projection algorithm. By varying the resistivities of the lungs, the ventricles and the atria in the thorax model, changes in the images were derived by computing the average resistivity change in the regions of interest. The results show there is a very strong relationship between organ volme and the magnitude of the observed resistivity change in the image. All the organs show a nearly linear change in the observed resistivity as a function of the resistivity change in the model.

In current clinical practice, the noninvasive assessment of left ventricular deformation can be determined using all the principal imaging modalities, including contrast angiography, echocardiography, cine computed tomography, single photon emission tomography and magnetic resonance imaging. However, since the heart undergoes complex motion, proper characterization of its motion still remains an open and challenging research problem. A number of approaches for nonrigid motion analysis have been studied in the literature. Much of the effort has confined to estimate the displacement vector for each image point or optical flow. This is a challenging problem in image analysis because of a wide range of possible motions and the presence of noise in the image sets. In this work, we present an algorithm for computation of optical flow based on a signal-dependent radially Gaussian kernel that adapts over time. The adaptive kernel obtained from the proposed algorithm is used to estimate a 3D-frequency spectrum for a given pixel in a series of images. The orientation of the spectrum in the frequency domain is totally governed by the pixel velocity. In a recent contribution, a linear regression model is used over the spectrum to obtain the velocity components that are proportional to the pixel movement.

A manual and a semi-automatic image registration method were compared with retrospective ECG (rECG) gating to correct for cardiac motion in myocardial perfusion (MBF) measurement. 5 beagles were used in 11 experiments. For each experiment a 30 s cine CT scan of the heart was acquired after contrast injection. For the manual method, a reference end-diastole (ED) image was selected from the first cardiac cycle. ED images in subsequent cardiac cycles were manually selected to match the shape of the reference ED image. For each cardiac cycle in the semi-automatic method, the image with the maximum area and the most similar shape to the selected image of the previous cardiac cycle was chosen as ED image. MBFs were calculated from the images registered by the three methods and compared. The averages of the difference of MBF_manual and MBF_semi-auto and MBF_rECG in the lateral free wall of LV were 3.6 and 3.4 ml/min/100g respectively. The corresponding standard deviations from the mean were 9.1 and 28.3 ml/min/100g respectively. We concluded from these preliminary results that image registration methods were better than rECG gating for correcting heart, which should facilitate more precise measurement of MBF.

Rapid Prototyping (RP) is a technique used in industry for manufacturing prototypes. Its capability to physically reproduce geometrical complex shapes is getting increasing interest in many fields of medicine. In the field of vascular surgery, replicas of artery lumen have utility in complex cases or when standard imaging is felt to be equivocal. Replicas can also facilitate experimental studies of computational vascular fluid-dynamics permitting in-vitro reproductions of blood flow in living subjects before and after surgery. The VIrtual VAscular (VIVA) project at CRS4, developed a system able to process three-dimensional (3D) datasets extracted from a Computer Tomography (CT) apparatus, visualize them, reconstruct the geometry of arteries of specific patients, and simulate blood flow in them. In this paper, the applicability of RP techniques to VIVA's real size replicas of an autoptic carotid vessel lumen is presented and an overview of the RP based system developed is provided. The techniques used in our prototype are discussed and experimental results for the creation of a human carotid lumen replica are analyzed. We discuss in detail the pipeline of manufacturing process: 3D geometric reconstruction from segmented points, geometry tessellation, STL (Stereo Lithography format) conversion. Moreover we illustrate some technical details of the specific RP technique used to build the lumen replicas, which is called Fused Deposition Modelling (FDM), the materials used for prototypes, throughput time and costs of the FDM models realized. The system is totally based on open-source software. This enables us to control each step of the pipeline, from data acquisition to STL export file. In this context, we present main sources of error encountered during all manufacturing process stages.

We present an automatic and robust tagged-residue detection technique using vector quantization based classification. This technique enables electronic cleansing even on poorly tagged datasets, leading to more effective virtual colonoscopy. In order to reduce the sensitivity towards intensity variation among the tagged residual material, we use a multi-step technique. First, we apply classification using an unsupervised and self-adapting vector quantization algorithm. Then, we sort the resultant classes by their average intensities. We apply thresholding on these classes based on a conservative threshold. This helps us in differentiating soft tissue inside tagged material from poorly tagged region or noise.

A hyperspectral imaging spectrograph has been used to measure the fluorescence and reflectance of cervical tissue in vivo. The instrument was employed in a clinical trial in Vilnius, Lithuania, where 111 patients were examined. The patients were initially screened by Pap smear, examined by colposcopy and a tissue sampling procedure was performed. Detailed histopathological assessments were performed on the biopsies, and these assessments were correlated with spectra and images. The results of the spectroscopic investigations show that different tissue types within one biopsy region exhibit different spectral signatures. A spectral analysis of the entire image localizes dysplastic regions in both fluorescence and reflectance, suggesting that the hyperspectral imaging technique is useful in the management of cervical malignancies.

Using Statistical Parametric Mapping (SPM), we have implemented methods to investigate the relation between findings from structural MRI and regional cerebral blood flow (rCBF) SPECT performed in groups of subjects with neuropsychiatric disorders and healthy controls. Using a MATLAB program developed and integrated into SPM, suited means of gray matter values (MGM) were calculated in regions relevant to the disorder in question, and linearly correlated with rCBF values from the same patients. In a study of Obsessive Compulsive Disorder (OCD), patients showed a focus of increased MGM in the right putamen relative to controls. MGM from the putamen were significantly inversely correlated with rCBF measures in regions critical to OCD, mainly the anterior cingulate gyrus. These results are consistent with the notion that an imbalance of cortico-striatal circuits is relevant to OCD pathophysiology. In a study of Alzheimer's disease (AD), foci of decreased gray matter in patients relative to controls were identified bilaterally in the hippocampus. In AD patients, reduced hippocampal MGM correlated significantly with decreased rCBF during a memory task, in a network of regions usually involved in memory processes. An inverse correlation with frontal rCBF was also observed, suggesting compensatory efforts of executive regions during the memory task.

Patent ductus arteriosus (PDA), a common condition among preterm infants, increases the risk of intraventricular hemorrhage, bronchopulmonary dysplasia, and death in afflicted individuals. Current clinical treatment of PDA relies on use of the drug indomethacin to close the ductus arteriosus. In the present study, we have investigated the effect of indomethacin on cerebral blood flow (CBF), cerebral blood volume (CBV), and cerebral mean transit time (MTT) in newborn piglets using computed tomography (CT) perfusion. Twenty newborn piglets divided by age into two groups, less than 12 hours of age (n = 10) and greater than 12 hours of age (n = 10) were studied. Five piglets in each group received indomethacin treatment (0.2 mg/kg infused over 30 min) while remaining piglets served as controls. No significant changes in CBF were observed in control groups. In both indomethacin treated groups, average CBF decreased 32.3% and 34.3% (P > 0.05) below baseline immediately post infusion in piglets less than and greater than 12 hours of age respectively. Piglets less than 12hours of age treated with indomethacin also exhibited a delayed increase in CBF, maximum average increase of 41.7% (P > 0.05) above baseline at 210 min post infusion, a response not observed in the corresponding group of piglets greater than 12 hours of age. The observed age dependent response may be due to functional/anatomical closure of the PDA.

A new three-dimensional (3D) method of evaluating the joint space from fast GRE MRI has been developed that allows the reconstruction of the two dimensional (2D) distance map between the femur and the tibia bone plates. This method uses the MRI data, an automated 3D segmentation, and an unsupervised joint space extraction algorithm that identify the medial and lateral compartments of the knee joint. The extracted medial and lateral compartments of the tibia-femur joint space were analyzed by 2D distance maps, where visual as well quantitative information was extracted. This method was applied to study the dynamic behavior of the knee joint space under axial load. Three healthy volunteers' knees were imaged using fast GRE sequences in a clinical scanner under unloaded (normal) conditions and with an axial load that mimics the person's standing load. Furthermore, one volunteer's knee was imaged at four regular time intervals while the load was applied and at four regular intervals without load. The results show that changes of 50 microns in the average distance between bones can be measured and that normal axial loads reduce the joint space width significantly and can be detected by this method.

Conventional SPECT Tc-99m sestamibi scintimammography (STSM) has limited clinical utility due to fairly low radiopharmaceutical uptake in the breast tissue as compared to the heart and the liver. We investigated the use of a cone-beam collimator (CBC) to STSM. Each detector on a multi-headed gamma camera can be equipped with parallel-beam (PBC) or cone-beam collimators (CBC). PBC can provide truncation-free SPECT projection sets, while CBC offers increased sensitivity in a limited field-of-view (FOV). Combined PBC and CBC SPECT ddata acquisition may provide improved lesion contrast and overall better imaging performance within CBC FOV with significantly reduced truncation artifacts in the reconstructed images. In this paper we evaluate the combined CBC&PBC SPECT method using a limited number of confirmed breast cancer patients and female chest phantoms with simulated breast lesions. We envision the combined CBC&PBC SPECT as a useful clinical tool in scintimammography.

For cancer polyp detection based on CT colonography we investigate the sample variance of two methods for estimating the sensitivity and specificity. The goal is the reduction of sample variance for both error estimates, as a first step towards comparison with other detection schemes. Our detection scheme is based on a committee of support vector machines. The two estimates of sensitivity and specificity studied here are a smoothed bootstrap (the 632+ estimator), and ten-fold cross-validation. It is shown that the 632+ estimator generally has lower sample variance than the usual cross-validation estimator. When the number of nonpolyps in the training set is relatively small we obtain approximately 80% sensitivity and 50% specificity (for either method). On the other hand, when the number of nonpolyps in the training set is relatively large, estimated sensitivity (for either method) drops considerably. Finally, we consider the intertwined roles of relative sample sizes (polyp/nonpolyp), misclassification costs, and bias-variance reduction.