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

Table 4

The comparison with the state-of-the-art methods (Barbosa et al [28], Milletari et al. [30], van Stralen et al. [31], Smistad et al. [40], and Keraudren et al. [41]) on correlation between the segmentation results and the ground truth in test set.

Method	EDV			ESV			EF
	Corr	bias	std	Corr	bias	std	Corr	bias	std
	val	ml.	ml.	val	ml.	ml.	val	val	val
[28]	0.035	-5	17.7	0.033	-6.8	13.9	0.111	2.9	5.2
[30]	0.047	5.1	19.0	0.040	-16.8	15.2	0.255	15.2	7.6
[31]	0.034	-15.4	16.0	0.036	-13.2	14.4	0.389	3.7	8.8
[40]	0.049	-10.1	19.4	0.036	-11.3	14.6	0.121	3.7	5.2
[41]	0.079	15.9	24.6	0.048	-6.2	16.6	0.281	12.1	10.6
The proposed method	0.018	1.41	12	0.021	-3.59	11.3	0.208	3.52	0.07