Dynamic computed tomography (CT) provides the advantages of sampling in many regions within an organ where a sampling catheter could not be introduced and eliminates superposition. However, dynamic CT is generally slow. An important aspect of the indicator dilution curve generation is the sampling frequency and sample duration. The indicator dilution curves generated with a CT scanner must behave like the `conventional' indicators. The contrast agent passes sequentially through the heart chambers and in doing so the duration of the bolus increases and the peak opacification of the bolus diminishes. The pattern of progressive dispersal of the dilution curve, somewhat in proportion to the duration of the curve's transit along the circulation, conveys something about the structure-to-function relationship of the circulation through which the indicator passes. The standard deviation to mean transit time relationship in the aortic root appears to be along a different regression line than does that relationship in the myocardium.

Intravascular ultrasound is a minimally invasive tomographic technique which produces 2-D cross-sectional images depicting vessel wall architecture and plaque morphology. Currently, no reliable automated approaches exist that offer segmentation of blood and vascular wall. We have developed a method for automated segmentation of intravascular ultrasound images to differentiate among plaque, wall, and blood. To achieve reliable border detection in noisy intravascular ultrasound images, a priori knowledge is incorporated in the edge detection process using heuristic graph searching. The method was validated using images from two phantoms that were imaged under several pressure conditions. In the first image set, our automated border detection method correctly identified the wall and plaque borders in 69/91 images. In the second image set, our method successfully identified external and internal wall and plaque borders in all 36 images. Lumen cross-sectional areas correlated very well with distending pressure in both sets of images. By comparison with the micrometer determined average wall thickness, mean absolute error of wall thickness was 0.02 +/- 0.01 mm.

Texture analysis has proven useful in the evaluation of many forms of digital imagery. Our group has shown that texture analysis of echocardiograms may be useful for characterization of myocardial ischemia, contusion, and cardiomyopathies. Recent work in tissue characterization of the brain has used texture analysis of magnetic resonance images to discriminate normal brian tissue from edema and tumor. Magnetic resonance images of the heart have been noted to exhibit visually apparent abnormalities in region image intensity in disease. Based upon these qualitative visual observations, our previous work in cardiac ultrasound texture analysis and the recent application of texture analysis to magnetic resonance brain images, we hypothesized that texture analysis would be useful in evaluating myocardial tissue characteristics from cardiac magnetic resonance images. As a first step towards the goal of cardiac tissue classification, we evaluated the ability of texture analysis to discriminate the left ventricular myocardium from other tissues using magnetic resonance images.

The elastic properties of arteries and plaque are difficult to assess due to the technical problems associated with obtaining the required dimensional and stress parameters in vivo. This paper describes a method to determine the biomechanical properties of blood vessels and plaque based on dimensional information obtained from the analysis of intravascular ultrasound images. As a pilot study, two models of a stenotic artery were made that included an eccentric, flexible plaque of known size. These models were submerged in water and imaged at up to 15 axial locations over a static pressure range from 0 to 200 mmHg. Plaque and wall borders were automatically detected using a heuristic graph searching technique. Cross-sectional area changes, dimensional parameters and elastic properties were determined for each location and intraluminal pressure. The incremental elastic modulus of the tube wall was within 8% of the value determined from micrometer measurements. Plaque stiffness was close to that of a fibrous arterial plaque. Cross-sectional changes indicated a radial expansion of the compliant plaque with pressure.

Electron beam computed tomography (EBCT) has provided a new tool for identification and possible quantification of coronary arterial plaque calcium. EBCT is the only imaging modality currently available which generates images of the spatial, temporal, and contrast resolution required for the identification of small foci of calcium and the potential for accurate quantification of calcium. Meanwhile, interest in quantification of coronary arterial calcium via EBCT and its correlation with severity of coronary atherosclerosis is increasing. Data remain inconclusive, but it appears that the reproducibility of quantitative grading of the extent of calcification by EBCT may be limited, in part, by the arbitrary nature of the scoring algorithm employed within the analysis tools currently provided by the EBCT manufacturer. It has not been possible to objectively determine optimum values for minimum plaque area and brightness threshold or to quantitatively determine whether single optimal values even exist. Also, although the current system tabulates the score, area, and mean attenuation for each plaque, the locations of the plaques are not reported.

Using ultrafast computed tomography (CT) for calcium quantification offers a potentially non- invasive way to evaluate the presence and severity of coronary artery disease. The currently applied index of ultrafast CT coronary calcium amount is the coronary calcium score of Agatston et al., but this score has not been thoroughly evaluated as to its accuracy and stability. In the present research we estimate the grey scale non-uniformity within the vascular blood pools in the anterior-posterior axis, and in the cephalad-caudad axis. We then estimate the effects of these non-uniformities on coronary calcium scores.

The recent emphasis on early diagnosis of coronary artery disease has stimulated research for a reliable and non-invasive screening method. Radiographically detectable coronary calcium has been shown to predict both pathologic and angiographic findings. Ultrafast computed tomography (UFCT), in quantifying coronary calcium, may become an accurate non-invasive method to evaluate the severity of coronary disease. The currently applied index of UFCT coronary calcium amount is the coronary calcium score of Agatston et al. This score has not been thoroughly evaluated as to its accuracy and dependence on scanning parameters. A potential drawback of the score is its dependence on predetermined CT number thresholds. In this investigation we used a chest phantom to determine the effects of particle size, slice thickness, and reconstruction algorithm on the coronary calcium score, and on the calcium mass estimated with a new method which is not dependent on thresholds.

A critically important component in the development of new methods for treatment of pulmonary diseases is the development of sensitive techniques for assessing alterations in regional lung structure and function. We describe an automated method for segmentation of airway trees from 3-D sets of CT images. The method is based on a combination of conventional 3-D seeded region growing that is used to identify large airways, knowledge- based 2-D segmentation of individual CT slices to identify probable locations of smaller diameter airways, and merging of airway regions across the 3-D set of slices resulting in a tree-like airway structure. The preliminary validation of the method was done in eighty 3 mm thick CT sections from two 40 slice data sets of a canine thorax scanned with lungs held at 1.5 kPa and 2.5 kPa. The method's performance was compared with that of the conventional 3-D region growing method. The knowledge-based approach to identification of potential airways in individual image slices substantially outperforms the conventional method and promises to be applicable to in vivo pulmonary CT images.

Multidimensional cardiac imaging no doubt has many origins and approaches. In reviewing the topic I would like to elaborate on one of those origins and analyze the technique addressed by the following points: (1) Did multidimensional imaging achieve the original goal (which, in our case, was more accurate estimation of the left ventricular chamber volume)? (2) Is this original goal still important? (3) Has multidimensional cardiac imaging provided new insights, diagnostic, or therapeutic capabilities? Indeed, the development of multidimensional imaging as we know it today provides an example of how a physiological measurement problem led to new measurement approaches that helped answer longstanding questions and that the new approach itself led to new directions of research using both the x ray-based as well as other imaging modalities.

In the latter 1970s, largely due to the development of echocardiography and the ready availability of invasive contrast ventriculography, clinicians noted that distinct and serial changes occurred in the heart after infarction where cardiac enlargement could progress long after completion of scarring in the infarct region. Similar changes have also been observed in the non-infarcted or non-ischemic myocardial regions commencing within days of infarction. This process, which involves both viable and infarcted heart muscle, has been termed `post- infarction cardiac remodeling.' A major obstacle to further comprehension of post-infarction cardiac remodeling in man has been related to limitations in applications of conventional cardiac imaging methods and conventional cardiac image processing. Electron beam computed tomography (EBCT) has emerged recently as an alternative means to image the heart and has been extensively validated for studies of ventricular volumes, function, muscle mass, shape, characterization of regional mechanics, and affords 3-D image registration. A research investigation was designed which involved a total of 55 patients entered prospectively following an index myocardial infarction. Each was imaged using EBCT at hospital discharge, six-weeks, six-months and one-year after the event.

We present a dynamic reconstruction of the left ventricle (LV) of the human heart. LV surface is represented by a set of points. The coordinates of these points are iterated by an artificial neural network while optimizing the match between the reconstruction based on these coordinates and the signal data. The input for the network are the segment's positions which represent the surface within the original data. The output is a set of real-valued coordinates quantifying the location of the LV surface points. The reconstruction is simultaneously developed in 3-D space and temporal domain. A topological constraint during training of the network gives corresponding vertices in space and time with global correctness. At any phase of the heart beat the network develops a map among the surface points which is highly ordered. This results in very regular wire-frames, that can be displayed rapidly on even small graphic workstations. Without time and third dimension this is very similar to Durbin's algorithm for solving the traveling salesman problem (TSP). To achieve a smooth representation we keep our network from developing the full TSP optimal solution.

This paper describes efforts aimed at more accurately and objectively determining and quantifying the local, regional, and global function of the left ventricle (LV) of the heart under both normal and ischemic conditions. These measurements and evaluations are made using non-invasive, 3-D, cardiac diagnostic imaging sequences (i.e., 4-D data) and rely on an approach that follows the shape properties of the endocardial and epicardial surfaces of the LV over the entire cardiac cycle. Our efforts involve the development of an acute infarct animal model that permits us to establish the validity of our noninvasive image analysis algorithms, as well as permits us to study the efficacy of using in vivo, image-derived measures of function for predicting regional myocardial viability (immediately post mortem). We first describe the experimental setup for the animal model, including the use of implanted imaging-opaque markers that assist in setting up a gold standard against which image-derived measurements can be evaluated. Next, the imaging techniques are described, and finally the image analysis methods and their comparison to the validation technique are discussed.

When analysis of global and regional left ventricular function is utilized to assess ischemic zone size, it is assumed: (1) a predictable relationship exists between the two; and (2) this relationship is not significantly affected by changes in loading conditions. This study examined the relationship between ischemic zone size and ventricular performance in 11 anesthetized paced dogs subjected to acute coronary occlusion. Cardiac cine-computed tomography was used to quantify regional and global left ventricular function. Ischemic zone size was quantified autoradiographically following left atrial microsphere injection. Both global and regional left ventricular function were strongly influenced by changes in ventricular loading. We conclude: (1) both global and regional left ischemic dysfunction are only modestly correlated with ischemic zone size during acute coronary occlusion; and (2) estimates of ischemic zone size using measurements of ventricular function are highly load-sensitive.

Anatomic delineation of the heart and great vessels is a necessity when managing children with congenital heart disease. Spatial orientation of the vessels and chambers in the heart and the heart itself may be quite abnormal. Though magnetic resonance imaging provides a noninvasive means for determining the anatomy, the intricate interrelationships between many structures are difficult to conceptualize from a 2-D format. Taking the 2-D images and using a volumetric analysis package allows for a 3-D replica of the heart to be created. This model can then be used to view the anatomy and spatial arrangement of the cardiac structures. This information may be utilized by the physicians to assist in the clinical management of these children.

We have recently embarked on a systematic evaluation of the regional and global mechanical processes of the systemic, morphologic right ventricle (RV) which is in either a single or dual chambered circulation as well as single left ventricles (LV). An MRI tagging technique which lays down 2 sets of parallel stripes perpendicular to each other on the myocardium as well as standard cine MRI were utilized. Finite strain analysis was applied to the grid lines to derive principle strains and the motion of the intersection points were tracked through systole to determine regional radial shortening and twist. Cine sequences were used to derive the various parameters of ventricular geometry and performance as well as visualizing flow profiles in the aorta. We noted a marked decrease in vol, EF, and CO in the Fontan group of patients when compared to other surgical subgroups. It is hypothesized that atrial stiffening by surgical placement of baffles may contribute to the observed changes in ventricular mechanics. Aortic flow profiles in the reconstructed aorta were noted to be heterogenous across the aortic diameter.

Recently, fuzzy c-Shells algorithms have been proposed for spherical boundary detections. Only a few applications of the Fuzzy c-Shells algorithm have been explored. This paper proposes a new algorithm utilizing the Fuzzy c-Shells approach for detection of the approximate location of the left ventricle. The proposed technique has been successfully applied to 25 2-D magnetic resonance cardiac images and has located the left ventricles.

Spatial registration of functional and structural data is a vital first step in the regional analysis of images obtained from functional imaging techniques. We present a set of procedures that produce accurate, objective, and robust registrations between PET and MR images that are highly automated and user-friendly. These techniques have been shown to be both reliable and valid. Because the method is automated and therefore efficient, each injection of the PET experiment can be individually fit to overcome errors introduced because of subject movement between PET injections.

Recent advances in high-resolution EEG imaging methods have made it advantageous to decrease inter-electrode distance to approximately 1 - 2 cm. To take full advantage of this increased recording density, it has become imperative to consider inter-subject anatomical variability and even intra-subject anatomical asymmetry. The present study used anatomical information from MRI to augment functional data obtained through EEG. Specifically, acrylic helmets made for each subject and normally used during EEG were utilized to orient NMR sample tubes filled with a marker medium (H₂O(DOT)Cu₂SO₄) radially from the scalp at selected EEG recording sites during MRI. Using the software package AVS, the MRI data could then be volumetrically 3-D rendered, 3-D isosurface rendered, or arbitrarily sliced. The tubes appeared in the 3-D renderings as pointers from recording sites to underlying cortical anatomy. Our task was simplified by our focus on a limited area of the cortex. The renderings provide subject-specific anatomical templates for mapping of EEG topographic patterns and clearly reveal individual variations of cortical surface topography that are usually unaccounted for in EEG analysis.

A common medical diagnostic problem is the determination of physiological function. MRI and CT are somewhat dissimilar but complementary imaging technologies. While CT provides excellent information regarding internal bony structures, MRI has proven to be superior in the production of high contrast images of soft tissue. The integration of these two modalities will ameliorate solutions to problems which require highly accurate mappings of anatomical features. PET imagery presents an accurate view of physiological function but little anatomical information. The ability to integrate quantitative information from these complementary modalities will result in improved medical analysis and subsequent improvements in patient care. We consider the problem of multimodality data fusion as an extension of regularization theory. Directionally controlled continuity stabilizers are utilized in the reconstruction process. The fusion method presented in this paper can deal with surface discontinuities of an arbitrary order. The fusion methodology presented can handle with data sets belonging to different visual cues.

Magnetic resonance imaging (MRI) can be used for functional brain studies. The identification of areas with changed blood oxygenation level dependent (BOLD) signal is usually done by visually inspecting maps of different kinds created through different post-processing procedures of the acquired images. It is desirable that the maps have as good an image quality as possible, and principal component analysis (PCA) can be used for this task. PCA is a data- driven method which does not use information about the timing of the experiment, instead the variance-covariance structure of the image data set is analyzed. PCA results in linear combinations of the analyzed MR images called score images, and the possibility to use score images as functional maps is investigated and compared to another commonly used method.

An angiographic method using first pass distribution analysis (FPA) has been investigated for determining instantaneous absolute volumetric blood flow in an angiographic perfusion phantom and in a rabbit animal model following intra-arterial injection of contrast material. The method is based on the concept of conservation of contrast material in successive angiographic images, utilizing the videodensitometric information in the arterial bed. The volume of contrast material entering the perfusion bed between two successive images was determined using videodensitometric and entrance vessel calibration techniques. In phantom studies, measured (M) and known (K) mean flow rates were related with videodensitometric and entrance vessel calibration techniques, respectively. In vivo, measured and known flow rates in the left common carotid artery of two rabbits were related with the videodensitometric and the entrance vessel calibration techniques, respectively. The results of this study demonstrated the potential utility of the FPA algorithm in conjunction with digital substraction angiography for measuring phasic blood flow.

Human studies were conducted to determine the feasibility of functional MRI at 1.5 T during auditory stimulation in the presence of gradient sound generated during a conventional gradient-echo sequence. The gradient-echo pulse sequence was optimized to minimize the effect of gradient sound. Stimuli comprised a simple 1 kHz tone as well as spoken words to investigate possible localization during language comprehension. Also, functional images in response to somatosensory and visual stimulation were generated, albeit with a different gradient-echo sequence to optimize contrast and temporal resolution. The resulting functional images indicate a 2 - 9% increase in signal intensity localized to the corresponding auditory, somatosensory or visual cortical regions during stimulation.

We have developed a knowledge-based approach to analyzing dynamic nuclear medicine data sets using factor analysis. Prior knowledge is used as constraints to produce factor images and their associated time functions which are physically and physiologically realistic. These methods have been applied to both planar and tomographic image sequences acquired using various single-photon emitting and positron emitting radiotracers. Computer-simulated data, non-human primate studies, and human clinical studies have been used to develop and evaluate the methodology. The organ systems studied include the kidneys, heart, brain, liver, and bone. The factors generated represent various isolated aspects of physiologic function, such as tissue perfusion and clearance. In some clinical studies, the factors have indicated the potential to isolate diseased tissue from normally functioning tissue. In addition, the factor analysis of data acquired using newly developed radioligands has shown the ability to differentiate the specific binding of the radioligand to the targeted receptors from the non-specific binding. This suggests the potential use of factor analysis in the development and evaluation of radiolabeled compounds as well as in the investigation of specific receptor systems and their role in diagnosing disease.

By applying non-conventional statistical analysis and visualization techniques to PET data obtained from a combined group of patients and normals, we are able to illustrate topographic covariance profiles unique to the disease at various stages of progression. Each profile represents a neuroanatomical regional network that is not discernible in the unprocessed data sets using standard analytical methods. The magnitude of a profile's manifestation in a given subject is expressed as a subject score which can correlate with independent clinical disease severity measures such as quantitative rigidity and bradykinesia ratings in Parkinson's disease. To create representations of these profiles a semi-automated routine is used which first generates a 2D pseudocolor map of the network where each region is weighted in accordance with its relative contribution to the overall profile. This representation is then transformed to a 3D isometric form so that the metabolic topography becomes more visually apparent. To fully perceive the evolving topographical pattern from initial to final stages of the disease, intermediate stages of disease progression are derived by interpolation to create a smooth progression of images that are displayed in an animated sequence.

Patients with pulmonary fibrosis frequently smoke cigarettes. The cause of dyspnea in these patients is often complex because of the coexistence of multiple disease processes. We investigated 10 cigarette smokers with pulmonary fibrosis who were referred for evaluation of new onset or worsening dyspnea. Chest radiographs and pulmonary function tests were obtained in addition to high-resolution computed tomography (HRCT). In those patients with HRCT evidence of both diseases, spirometry and lung volumes were most often normal. Although plain films provided a reasonable assessment of fibrosis, they underestimated the severity of emphysema. Quantitation of both emphysema and fibrosis by HRCT was reproducible and correlated with key pulmonary function tests. Our findings indicate that the HRCT scan is a useful diagnostic test in patients with pulmonary fibrosis who are also cigarette smokers.

The airway has always been a central focus for respiratory pathology in infants and children. Imaging of the larynx, trachea, and the central bronchi can be readily accomplished by radiographic or conventional CT techniques. Newer high resolution CT (HRCT) techniques have extended our view of the bronchi peripherally to the limits of scanner resolution, i.e., to bronchial generations 7 - 9, and rapid volumetric CT data acquisitions have made it possible to follow the same lung anatomic level through the rapidly occurring changes in a series of experimental protocols. These techniques together with a custom designed computer software program for image display and analysis have enabled us to objectively study changes in airway caliber and lung density that occurred in an animal mode of airway reactivity and thereby relate structure with function in the airways.

The bronchial tree is one of the most well known fractal structures in the human body. Fractal objects like the bronchial tree have complex structures with self-similar properties over different scales. In this paper, we present a 3-D bronchial tree computer model constructed using fractal growth rules and actual morphometric data. Given the important role that the chest walls play in lung morphogenesis, the growth of the model was constrained within a 3-D boundary extracted from high resolution CT (HRCT) lung data. The model was built within a space of (512 X 512 X 512) pixels, and it currently models only the right lung. The parameters of the model are: (1) the diameter, d, of a branch; (2) the length, l, of a branch; (3) the angle, (theta) , between a branch and its parent. Utilizing this model as a testground, we developed a functional relationship between different changes in the morphology of the bronchial tree and the values of its 3-D fractal dimension and 3-D lacunarity.

This report focuses on preliminary experiments designed to determine regional blood flows and air, blood, and tissue contents at end expiratory lung volume in anesthetized, paralyzed, normal, sham-operated, and pneumonectomized (left lung removed) rabbits with and without wax plombage. High temporal resolution measurements were made with an EBCT scanner during the mechanical injection of a bolus of radiopaque contrast material into the pulmonary vasculature. The time-intensity curves of selected lung regions were analyzed with VIDA^R using a modification of the myocardial blood flow model proposed by Wolfkiel et al. The resulting data provided an estimate of regional blood flow and total and regional air, blood and `tissue' contents, where `tissue' represents intracellular and interstitial water, i.e., lung water exclusive of blood. The estimates of mean lung air, blood and tissue contents were similar across groups and consistent with anticipated results.

Spiral computed tomography (CT) is a new technique for rapidly acquiring volumetric data within the body. By combining a continuous gantry rotation and table feed, it is possible to image the entire thorax within a single breath-hold. This eliminates the ventilatory misregistration seen with conventional thoracic CT, which can result in small pulmonary lesions being undetected. An additional advantage of a continuous data set is that axial sections can be reconstructed at arbitrary intervals along the spiral path, resulting in the generation of overlapping sections which diminish partial volume effects resulting from lesions that straddle adjacent sections. The rapid acquisition of spiral CT enables up to a 50% reduction in the total iodinated contrast dose required for routine thoracic CT scanning. This can be very important for imaging patients with cardiac and renal diseases and could reduce the cost of thoracic CT scanning. Alternatively, by combining a high flow peripheral intravenous iodinated contrast injection with a spiral CT acquisition, it is possible to obtain images of the vasculature, which demonstrate pulmonary arterial thrombi, aortic aneurysms and dissections, and congenital vascular anomalies in detail previously unattainable without direct arterial access.

We have recently developed methodologies to assess the local distributions of alveolar ventilation and pulmonary perfusion using positron emission tomography (PET) with ¹³NN gas as a tracer. In order to quantify the true regional heterogeneity in lung function from these images, it was important to assess the contributions of noise caused by finite count statistics and by imaging artifacts. To characterize these artifacts we collected multiple images with different total number of counts from a uniform phantom labeled with ¹¹CO₂ and assessed their heterogeneity as the mean normalized variance of the pixel by pixel data. We developed a novel disc phantom made of open cell foam with a density comparable to that of the lungs. Images of this phantom were reconstructed with a Hanning filter set for different resolution lengths (L). The mean normalized variance of these images was found to closely follow a linear relationship with the inverse of the average number of counts per pixel and L^-3 having an intercept that represented the heterogeneity caused by imaging and reconstruction artifacts.

We reanalyzed regional pulmonary blood flow (PBF) and lung water concentration (LWC) data obtained by positron emission tomography to calculate the spatial correlation, p(d), among picture elements (pixels) as a function of distance. Animals were studied in the supine position. Data were obtained both at baseline and after oleic acid induced acute lung injury. With this new analytical approach, we confirmed that p(d) for PBF is strongly positive for adjacent regions but values fall off steadily, reaching strong negative values, at the greatest distances measured. Acute lung injury affected the relationship between p(d) and distance, with less negative correlation at greater distances when perfusion redistribution occurred after injury. However, the magnitude of this effect was not great. The relationship between p(d) and distance for LWC was similar to that observed for PBF, but lung injury caused more, not less, negative correlation at greater distances. Since the relationship between p(d) and distance does not seem to be very sensitive to changes in perfusion distribution, better descriptors of regional heterogeneity are still needed.

To help characterize the determinants of the spatial distribution of regional pulmonary structure and function and to characterize a spatial autocorrelation (SAC) approach, we have applied SAC statistics to our pulmonary cine x-ray CT data of regional pulmonary blood flow and to various computer derived models (cubes and pyramids, 3-D wedges, and lung shapes in which pure `flow' gradients in either the x, y, or z directions were applied). To generate graphs of correlation vs. distance, we bin the data according to distance into a user specified number of groupings and then autocorrelate the data within each bin. Only regions of pulmonary parenchyma within the same lobe were used. We present the results of our analysis which show that several regional parameters exhibit a similar negative sloping correlation vs. distance relationship. SAC statistics provide a unique tool for demonstrating the existence of underlying patterns to distribution of pulmonary function.

Accurate measurement of regional pulmonary blood flow distribution is of interest both as a research and diagnostic tool. Measurements of regional pulmonary perfusion via x-ray CT offer the possibility of detecting perfusion deficits due to pulmonary embolus while maintaining a high degree of anatomic detail. Use of bolus injection of conventional radiopaque contrast with associated short mean transit times (5 - 7 seconds), requires a high degree of temporal resolution offered clinically only by electron beam x-ray CT (Imatron). The present study was intended to characterize biodegradable radiopaque microspheres as an alternative contrast agent which would allow for measurement of regional pulmonary blood flow with scanning times associated with conventional or spiral thin slice, volumetric x-ray CT protocols. To test this, a dog was scanned at 6 slice levels and 13 time points with image acquisition gated to the cardiac cycle. Lung volumes were maintained at functional residual capacity.

A cellular automation is described which mimics cell population proliferation by replacing the site values by the age and the cycle phase of cells. The model described in this paper takes into account the size of the cells. The model is used to simulate the cell proliferation of the human breast cancer cell line MCF-7. The initial configuration of the cellular automaton is obtained from the discretization of the experimental data obtained from a light microscopic image analysis at 1 day of culture. After each day of proliferation the pattern obtained from the simulation is compared with the experimental result of the image analysis. The comparison is made from a topographical point of view through the concept of minimal spanning tree graph.

This paper summarizes three recent applications of computer vision techniques in dentistry developed at the Czech Technical University. The first one uses a special optical instrument to capture the image of the tooth arc directly in the patient's mouth. The captured images are used for visualization of teeth position changes during treatment. The second application allows the use of images for checking teeth occlusal contacts and their abrasion. The third application uses photometric measurements to study the resistance of the dental material against microbial growth.

Texture analysis of bone radiographs can play an important role in characterizing the progression of bone diseases by computing texture measures on the digitized bone radiographs. Fractal dimension is one such texture measure which has been used with success as radiographs of trabecular bone are shown to exhibit self-similar characteristics. Markov random fields (MRF) have been used successfully to classify texture by modeling it as stochastic processes. But it has been shown that MRF models do not perform well in modeling self-similar textures such as fractional Brownian motion (FBM). This limitation can be overcome by characterizing statistical properties of the incremental process which builds up a fractal object. Since we try to characterize the statistical properties of the incremental process which builds the fractal object rather than its multi scale behavior using Gaussian MRF, this approach is complimentary to using fractal dimension as a feature in characterizing texture.

Detection of subtle structural changes in trabecular bone is important in evaluating the load- bearing capability of whole bones. Microstructural changes in trabecular bone due to remodeling or resorption lead to changes in bone strength. Recently, fractal-based analyses of radiographs have demonstrated that a fractal model can describe trabecular bone patterns independent of mass density. In this case, the descriptor of choice is a scale-invariant measure of trabecular detail known as fractal dimension. The objective of this work was to compare two measures of the distribution of trabecular bone -- fractal dimension and mean gray level -- in a decalcifying environment. The fractal-based analysis relied upon the spatial distribution of trabecular material while the mean gray level measurements depended upon the average x- radiation attenuation over a region of interest. Data were produced from four separate slices of vertebral bone which demonstrated that a change in the spatial distribution of trabecular material may be expressed in terms of a concurrently changing estimate of the fractal dimension within a region of interest. This change was not necessarily reflected in the mean gray level estimate of mass density.

Based on analyses of various types of digital microcirculatory image (DMCI), we summed up the image features of DMCI, the digitizing demands for digital microcirculatory imaging, and the basic characteristics of the DMCI processing. A dynamic and still imaging separation processing (DSISP) mode was designed for developing a DMCI workstation and the DMCI processing. Original images in this study were clinical microcirculatory images from human finger nail-bed and conjunctiva microvasculature, and intravital microvascular network images from animal tissue or organs. A series of dynamic and still microcirculatory image analysis functions were developed in this study. The experimental results indicate most of the established analog video image analysis methods for microcirculatory measurement could be realized in a more flexible way based on the DMCI. More information can be rapidly extracted from the quality improved DMCI by employing intelligence digital image analysis methods. The DSISP mode is very suitable for building a DMCI workstation.

Rapid analysis of large multi-dimensional data sets is critical for the successful implementation of a comprehensive MR cardiac exam. We have developed a software package for the analysis and visualization of cardiac MR data. The program allows interactive visualization of time and space stacks of MRI data, automatic segmentation of myocardial borders and myocardial tagging patterns, and visualization of functional parameters such a motion, strain, and blood flow, mapped as colors in an interactive dynamic 3D volume rendering of the beating heart.