Coronary Segmentation from Echocardiography using Fuzzy Connectedness

In this paper we study the segmentation of coronary arteries from bidimensional echocardiography simulated images (phantoms) using Fuzzy Connectedness concepts implemented using Image Foresting Transform (IFT). This approach transforms the image into an oriented and weighted graph; therewith a graph-based algorithm can be applied to process the image segmentation. In echocardiographic scenario, it is necessary a preprocessing step, a filtering, to reduce noise degradation in order to improve accuracy to the segmentation algorithm. So far, we have tested two filters: linear scaling mean variance (LSMV) and another one with edgepreserving feature. Finally, an evaluation is proposed (for each filter) using phantoms simulating echocardiography images and results are discussed. Using Watershed 2D by IFT and Generalized Fuzzy Connectedness 2D, preliminary results are reached and they will be compared to the proposed 2D segmentation by IFT.