. Author manuscript; available in PMC: 2021 Aug 1.

Published in final edited form as: IEEE Trans Med Imaging. 2021 Feb 2;40(2):699–711. doi: 10.1109/TMI.2020.3035253

Table II.

Quantitative Comparison of Different Channel Attention Methods for Skin Lesion Segmentation. Enc, Dec and Enc&Dec Means Channel Attention Blocks Are Plugged in the Encoder, the Decoderand Both Encoder and Decoder, Respectively

Network	P_max	Para	Dice(%)	ASSD(pix)
Baseline	-	1.9M	87.77±3.51	1.23±1.07
C-A(Enc)	×	2.7M	91.06±3.17	0.73±0.56
C-A(Enc)	√	2.7M	91.36±3.09	0.74±0.55
C-A(Dec)	×	2.7M	91.56±3.17	0.64±0.45
C-A(Dec)	√	2.7M	91.68±2.97	0.65±0.54
C-A(Enc&Dec)	×	3.4M	90.85±3.42	0.92±1.40
C-A(Enc&Dec)	√	3.4M	91.63±3.50	0.58±0.41