A Hybrid Unsupervised Approach for Retinal Vessel Segmentation

Table 10

Performance assessments of existing techniques on the four datasets.

Method	Year	DRIVE		STARE		HRF		CHASE_DB1
		MCC	CAL	MCC	CAL	MCC	CAL	MCC	CAL
Unsupervised techniques
Chauduri [31]	1989	0.420	0.208	—	—	—	—	—	—
Hoover [53]	2000	—	—	0.615	0.534	—	—	—	—
Fraz [68]	2011	0.733	—	0.700	—	—	—	—	—
Fraz [69]	2013	0.736	—	0.691	—	—	—	—	—
B-COSFIRE [5]	2015	0.719	0.721	0.698	0.709	0.686	0.577	0.656	0.608
RUSTICO [58]	2019	0.729	0.728	0.698	0.709	0.691	0.587	0.663	0.620
Proposed	2020	0.739	0.696	0.707	0.566	0.710	0.656	0.629	0.547
Supervised techniques
Yang [70]	2019	0.736	—	0.704	—	0.712	—	—	—
Yang [71]	2018	0.725	—	0.662	—	0.682	—	—	—
FC-CRF [73]	2016	0.756	0.731	0.727	0.658	0.690	0.541	0.704	0.622
UP-CRF [73]	2016	0.740	0.675	0.726	0.665	0.677	0.475	0.689	0.571
Vega [72]	2015	0.662	—	0.640	—	—	—	—	—
Niemeijer [11]	2004	0.722	0.659	—	—	—	—	—	—