Left ventricular segmentation in MR using hierarchical multi-class multi-feature fuzzy connectedness

Abstract

In this paper, we present a new method for data-driven automatic extraction of endocardial and epicardial contours of the left ventricle in cine bFFE MR images. Our method employs a hierarchical, multi-class, multi-feature fuzzy connectedness framework for image segmentation. This framework combines image intensity and texture information with anatomical shape, while preserving the topological relationship within and between the interrelated anatomical structures. We have applied this method on cine bFFE MR data from eight asymptomatic and twelve symptomatic volunteers with very encouraging qualitative and quantitative results. © Springer-Verlag Berlin Heidelberg 2004.