Segmentation of ventricular angiographic images using fuzzy clustering

Describes a fuzzy based segmentation algorithm for the estimation of left ventricular contours in angiographic images. The proposed approach proceeds in two stages. Firstly, a fuzzy c-mean classification algorithm is used to provide a fuzzy partition of the image. For that purpose, a membership function is computed for each pixel and allows its classification as belonging to the ventricle or to the image background. The second stage of the method is devoted to a decision process, applying a global analysis followed by a fine segmentation which is only focused on ambiguous points. First results on real images are then presented and discussed.