Semiautomatic Epicardial Fat Segmentation Based on Fuzzy c-Means Clustering and Geometric Ellipse Fitting

Table 1

Comparison of methods using Dice similarity coefficient with manually segmented ground truth images (approved by a cardiovascular surgeon). Our proposed method outperforms the LAMT method, the classical region growing, and the snake methods.

Patient	1	2	3	4	5	6	7	8
Proposed method	0.76	0.65	0.6	0.7	0.73	0.54	0.73	0.82
LAMT	0.3	0.32	0.38	0.24	0.5	0.2	0.3	0.07
Region growing	0.39	0.15	0.34	0.34	0.28	0.24	0.39	0.17
Snake	0.3	0.15	0.16	0.32	0.36	0.29	0.38	0.3