A Novel Approach to Skin Lesion Segmentation: Multipath Fusion Model with Fusion Loss

Table 1

Comparative analysis of proposed and the single methods (results of the 20% test set from the training set).

Method	Performance metrics
	Dice	Jaccard	Acc	Sens	Spec
U-net + Resnet2D (with dice loss)	0.75	0.67	0.86	0.85	0.86
U-net + Resnet2D (with FTL)	0.87	0.79	0.88	0.87	0.88
V-Net2d (with dice loss)	0.88	0.81	0.91	0.90	0.91
V-Net2D (with FTL)	0.90	0.83	0.93	0.92	0.93
Fusion model with fusion loss (ours)	0.92	0.84	0.95	0.95	0.96