A Combined Fully Convolutional Networks and Deformable Model for Automatic Left Ventricle Segmentation Based on 3D Echocardiography

Table 2

The segmentation results of different initialization methods with traditional snake on 3DE in test set.

Method	ED						ES
					dice						dice
	mean	std	mean	std	mean	std	mean	std	mean	std	mean	std
Standard initialization + traditional snake	4.9	1.7	15.9	5.1	0.23	0.18	5.8	1.35	13.5	7.9	0.35	0.15
FCN + traditional snake	4.6	1.58	13.8	4.7	0.29	0.17	5.5	1.67	12.9	7.2	0.38	0.19
FCN+T + traditional snake	3.8	1.34	11.7	4.5	0.26	0.15	4.9	1.27	11.4	5.8	0.34	0.16
FCN+T++ traditional snake	3.35	0.89	9.8	4.2	0.23	0.13	3.48	0.91	10.1	4.9	0.30	0.14
FCN+T++ traditional snake	3.15	0.78	9.4	3.7	0.18	0.11	3.1	0.83	9.87	4.2	0.21	0.11
FCN+T++ traditional snake	2.9	0.7	8.9	3.1	0.13	0.008	2.8	0.75	9.3	3.9	0.16	0.05