Dominique Delaere

Continuous voxel classification by stochastic relaxation: theory and application to MR imaging and MR angiography

Abstract In this paper we present a stochastic relaxation method for voxel classification in magnetic resonance (MR) images. This method is based on Bayesian decision theory. In this framework, the optimal classification corresponds to the minimum of an objective function, which is here defined as the expected number of misclassified voxels. The objective function encodes constraints according to two a priori models: the scene model and the camera model. The scene model reflects a priori knowledge of anatomy and morphology; the camera model relates observed MR-image intensities to anatomical objects. Both models are described using the concept of Markov random fields (MRF). This allows continuity and local contextual constraints to be easily modelled via the associated Gibbs Potential Functions. The minimum of the objective function is approximated asymptotically by stochastically sampling the associated Gibbs posterior joint probability distribution. The method is applied to brain tissue classification in MRI and blood vessel classification in MR angiograms. Each application contains a novel aspect: in the former, we introduce topological constraints on neighbouring tissues; in the latter, we incorporate shape constraints on cylindrical structures.

Experimental Teleradiology. Novel Telematics Services Using Image Processing, Hypermedia and Remote Cooperation to Improve Image-Based Medical Decision Making

We illustrate the possible impact of information technology on digital radiology using two pilot projects. The first aimed to improve the transfer of radiological information and the interaction between radiologist and referring physician, using hypermedia documents and hypermedia electronic mail. During a 12-month evaluation period, approximately 100 hypermedia reports were generated and distributed to hepatologists and neurosurgeons. The second project aimed to improve planning of orthopaedic surgery, using an image processing service for three-dimensional visualization and a conferencing system to support active cooperation. Over two months, 12 routine cases were reconstructed in three dimensions and conferencing was carried out between surgeons and radiologists in hospitals 15 km apart. Initial results were encouraging. Telematics may be valuable in reconciling the growing need for multidisciplinary cooperation with the growing geographical and organizational separation of different experts. We conclude that to benefit from information technology, the focus should not be on a direct translation of traditional working methods, but rather on possibilities that were not available in the film-based environment.

Local filtering and global optimization methods for 3-D magnetic-resonance angiography image enhancement

In this presentation, we discuss the visualization of cerebral blood vessels in 3-D MR angiography images. Two techniques for an improved visualization are investigated: 3-D non- linear morphological filters that enhance the contrast of blood-vessel-like structures and a global stochastic optimization framework incorporating shape constraints. The resulting filtered images are combined into a novel hybrid volume rendering visualization method for the integrated viewing of brain structures and cerebral vasculature.

Automated contour detection of the left ventricle in short axis view and long axis view on 2D echocardiograms

An algorithm is presented to automatically delineate the endocardium in echocardiograms. The algorithm uses an optimization technique to overcome the poor image quality of existing algorithms and to use dynamic information. The information used in the process consists of grey level gradient, texture gradient, smoothness of the contour, and deviation of the present contour with respect to one of the previous video frames. The optimization algorithm locates the best compromise between image information and continuity in space and time. The automatic delineation is used for wall motion analysis and calculation of volumes and volume changes of the left ventricle (LV). An assessment of the accuracy of the volume calculations and indirectly the delineation was made by comparing the actual and calculated volume of the LV of excised dog hearts filled with a known amount of water. It is demonstrated that the dynamic programming technique is promising in detecting the myocardial borders.<<ETX>>

Model-Based Image Segmentation: Methods and Applications

We discuss different methods and applications of model-based segmentation of medical images. In this paper model-based segmentation is defined as the assignment of labels to pixels or voxels by matching the a priori known object model to the image data. Labels may have probabilities expressing their uncertainty. Particularly we compare optimization methods with the knowledge-based system approach.

Knowledge-based system for the fully automatic quantification of coronary stenotic lesions from two angiographic projections

ABSTRACTIn this paper, we describe work-in-progress regarding a fully automatic reporting system for coronary arterystenotic lesions. In a first step all blood vessel segments are assigned their anatomical label according to a coronaryanatomy model. Segment labeling is done using a constraint satisfaction technique, because most anatomical coronaryartery knowledge can be formulated as umary constraints only depending on local segment attributes, and binaryrelational constraints such as thicker than, left ofand above. In a second step, we perform an automatic quantificationof all artery trajectories. Therefore, we calculate a stenosis severity score for each segment, which is not only based on local properties, like per cent diameter or per cent area stenosis, but also takes into account the anatomical significance of the vessel. For example, a stenotic lesion proximal on the Left Anterior Descending (LAD) branch ismuch more significant than one on its distal side branches. Results are presented on clinical coronary angiograms.1. INTRODUCTIONThe primary reason to visualize the coronary arteries is for the localization and assessment of the atheroscleroticlesions. Although the exact causes of the disease are not yet fully understood, in (1) it is described as follows:Atherosclerosis is a variable combination of changes of the intima consisting of the focal accumulationof lipids, complex carbohydrates, blood and blood products, fibrous tissue and calcium deposits, andassociated with medial changes.With coronary arteriography, the disease can be recognized as obstructive artery lesions usually accompanied witha decrease in artery lumen. To be able to quantify these obstructive lesions, various groups have developed contourdetection methods. Usually, these methods make use of a few manually defined points on the center line of the artery.Then, the artery trajectory is extracted and a diameter or area function is calculated, based on densitometry or edgedetection (2, 3, 4).Criticisms to these semi-automatic systems can be summarized as follows:. The main disadvantage is the (sometimes) large amount of operator interaction in analyzing a large number ofvessels, thus impending a frequent and flexible use in clinical practice.. These systems are unable to give a qualitative estimate of stenosis severity since they take only into account thegeometrical significance of a stenotic lesion and not the anatomical relevance. To be able to do so, anatomicalcoronary artery knowledge has to be incorporated in order to obtain a structural object description.However, stenosis severity also depends on the anatomical interpretation of the coronary segment. For example,a 50% stenotic lesion proximal on the left anterior descending branch is much more important than a similarlesion in one of its distal side branches! Therefore, an objective severity score has to take into account therelative position and anatomical significance of the specific artery.

FUZZY VOXEL LABELING OF MAGNETIC RESONANCE IMAGES BY GLOBAL OPTIMIZATION

In this paper, we present a probabilistic method that is devised to improve the visual representation and the quantitative analysis of magnetic resonance (MR) images. The result of this procedure can be seen as a “fuzzy” labeling of the image voxels (where the voxel intensity represents the degree of belief that the voxel belongs to a certain object class).

A knowledge-based system for the automatic quantification of stenotic lesions on angiograms

Work-in-progress is described regarding a fully automatic reporting system for coronary artery stenotic lesions. A knowledge-based system is developed for the automatic labeling of the left coronary artery (LCA) on standard RAO and LAO projections. By means of a hierarchical search strategy, using several heuristic blood vessel models, the most important coronary arteries are given a correct anatomical label. After this labeling phase, an automatic quantification of all artery trajectories is performed. A stenosis severity score is calculated for each segment, which is not only based on local properties, like percent diameter or percent area stenosis, but also takes into account the anatomical significance of the vessel. Results are presented on clinical coronary angiograms.<<ETX>>