TiM-Net: Transformer in M-Net for Retinal Vessel Segmentation

Table 7

The corresponding coarse-grained ablation analysis results. The best value of each metric is shown as 0.9726.

Data set	Backbone1	Backbone2	DA	TransL2	Acc	Se	Sp	AUC
DRIVE	✓				0.9604	0.7042	0.9854	0.9130
	✓		✓		0.9638	0.7787	0.9817	0.9358
	✓			✓	0.9617	0.7136	0.9858	0.9345
	✓		✓	✓	0.9641	0.7523	0.9847	0.8858
		✓			0.9640	0.7514	0.9846	0.9619
		✓	✓		0.9638	0.7869	0.9810	0.9634
		✓		✓	0.9639	0.7330	0.9862	0.9620
		✓	✓	✓	0.9638	0.7805	0.9816	0.9682
CHASEDB1	✓				0.9684	0.7430	0.9842	0.8902
	✓		✓		0.9713	0.7303	0.9874	0.9172
	✓			✓	0.9693	0.7553	0.9838	0.9312
	✓		✓	✓	0.9681	0.7617	0.9821	0.9062
		✓			0.9711	0.7523	0.9860	0.9643
		✓	✓		0.9719	0.7692	0.9856	0.9679
		✓		✓	0.9712	0.7635	0.9854	0.9670
		✓	✓	✓	0.9711	0.7697	0.9865	0.9648
STARE	✓				0.9674	0.7371	0.9878	0.8855
	✓		✓		0.9726	0.8132	0.9875	0.9626
	✓			✓	0.9700	0.7681	0.9878	0.9677
	✓		✓	✓	0.9697	0.7351	0.9911	0.8970
		✓			0.9686	0.7665	0.9871	0.9633
		✓	✓		0.9707	0.7848	0.9878	0.9680
		✓		✓	0.9700	0.7759	0.9876	0.9700
		✓	✓	✓	0.9711	0.7867	0.9880	0.9670