Automatic Segmentation Measuring Function for Cardiac MR-Left Ventricle (LV) Images

Abstract

Automatic segmentation approaches are a desirable solution for Endocardium (inner) and Epicardium (outer) contours delineation using cardiac magnetic resonance left ventricle (CMR-LV) short axis images. The Level Set Model (LSM) and Variational LSM (VLSM) is the state-of-the-art in detecting the inner and outer contour for medical images. However, in CMR-LV images segmentation the LSM and VLSM are facing with the issue of re-initialisation because of irregular circle shape. In this paper, we developed an automatic segmentation measuring function based on statistical formulation to solve the re-initialisation issues in huge set of data images. The sign Euclidean distance function successfully classified the negative (inner contour) and positive (outer contour) features. The Fuzzy C mean interaction operator intersects the high membership degree that initialises the centre point. The experiments were conducted using the Sunnybrook and Pusat Juntung Hospital Umum Sarawak (PJHUS) cardiac datasets. This paper aims at developing a distance function to guide the automatic segmentation for LV contours and also to reduce segmentation error.