FUZZY CLUSTERING OF ELLIPTIC RING-SHAPED CLUSTERS

A new method for the detection and parameter estimation of elliptic ri ng-shaped clusters is developed. The algorithm is based on optimizatio n of an objective function in which the ellipse prototype parameters ( 2 foci and a radius) are included in the optimization scheme. The perf ormance of the Fuzzy k-Ellipses (FKE) algorithm has been evaluated on various forms of synthetic data, such as multiple intersecting ellipse s, concentric ellipses, incomplete ellipses, ellipses with hidden line s, and combinations of the above, and has been found to be very good. The algorithm was employed for the detection of the outer and inner co ntours of the heart's left ventricle from MRT images of transverse sec tions of the thorax. Reconstruction of 3-D dynamic images (systole-dia stole) of the left ventricle was obtained from the 2-D images, taken a t different points in time during the cardiac cycle.