A Segmentation of Melanocytic Skin Lesions in Dermoscopic and Standard Images Using a Hybrid Two-Stage Approach

Table 4

Comparative results of segmentation between the existing and proposed methods for images from the PH² and Dermofit atlases.

Methods	Dermofit			PH^2 data
	SEN	SPE	ACC	SEN	SPE	ACC
Otsu with RGB (MATLAB 2018b)	0.611	0.723	0.683	0.522	0.706	0.652
Level set with RGB (MATLAB 2018b)	0.712	0.878	0.805	0.719	0.800	0.784
FC-LS with RGB [28]	0.873	0.926	0.918	0.891	0.914	0.904
Adaptive thresholding with YIQ [38]	0.618	0.980	0.937	0.703	0.949	0.879
-means with CIELAB [21]	0.809	0.789	0.824	0.869	0.953	0.932
Local binary pattern clustering [39]	0.787	0.923	0.704	0.884	0.948	0.859
Proposed method (HK-LS with CIELAB)	0.919	0.944	0.942	0.923	0.964	0.946

FC-LS: fuzzy C-mean thresholding-based level set; HK-LS: hierarchical -means clustering-based level set; SEN: sensitivity; SPE: specificity; ACC: accuracy.