Fully Convolutional Network and Visual Saliency-Based Automatic Optic Disc Detection in Retinal Fundus Images

Table 1

Comparison of the evaluation indexes with different methods.

Methods	Dataset	Dice	Jaccard	Recall	Accuracy
SLIC-Hough	DRISHTI-GS	0.866 3	0.846 7	0.953 4	0.992 2
	RIM-ONE r3	—	—	—	—
LICE	DRISHTI-GS	0.920 0	0.854 1	0.935 2	0.967 9
	RIM-ONE r3	0.905 6	0.810 7	0.896 4	0.924 1
Modified U-net	DRISHTI-GS	0.949 0	0.904 2	0.926 8	0.996 9
	RIM-ONE r3	0.903 6	0.8289	0.8737	0.9922
Our method	DRISHTI-GS	0.965 8	0.934 1	0.964 8	0.996 7
	RIM-ONE r3	0.918 2	0.877 5	0.904 8	0.993 2