TABLE 2.

Comparison of the segmentation performance of the proposed method and several competing FCNs on the brain cortical plate and hippocampus datasets.

Dataset	Method	DSC	HD95 (mm)	ASSD (mm)
Brain cortical plate	Proposed	0.878 ± 0.037 *	0.871 ± 0.141 *	0.271 ± 0.068 *
	UNet++	0.862 ± 0.054	0.938 ± 0.430	0.258 ± 0.082
	Attention UNet	0.831 ± 0.060	1.125 ± 0.337	0.536 ± 0.128
	DSRnet	0.846 ± 0.050	0.978 ± 0.229	0.492 ± 0.135
	SEFCN	0.786 ± 0.076	2.095 ± 0.801	0.641 ± 0.115
Hippocampus	Proposed	0.895 ± 0.020 *	1.035 ± 0.203 *	0.416 ± 0.064 *
	UNet++	0.874 ± 0.027	1.479 ± 1.427	0.533 ± 0.208
	Attention UNet	0.820 ± 0.036	5.850 ± 5.150	1.192 ± 0.597
	DSRnet	0.821 ± 0.038	3.438 ± 3.065	1.395 ± 0.316
	SEFCN	0.716 ± 0.052	8.614 ± 5.230	1.795 ± 0.761

Better results for each dataset/criterion have been marked using bold type. We used paired t-tests to find statistically significant differences;

asterisks denote significantly better results at p < 0.01.)