Robust Vessel Segmentation in Fundus Images

Table 2

Comparison of the results using the DRIVE [26] public database. The proposed methods achieved the best accuracy (Acc) compared to the state-of-the-art solutions.

Algorithm	Se	Sp	Acc
Proposed	0.644	0.987	0.9572
Frangi et al. [1]	0.660	0.985	0.9570
Marín et al. [34]	0.706	0.980	0.945
Human observer	0.776	0.972	0.947
Dizdaroglu et al. [24]	0.718	0.974	0.941
Soares et al. [35]	0.7283	0.9788	0.9466
Mendonça and Campilho [36]	0.7344	0.9764	0.9452
Staal et al. [37]	0.7194	0.9773	0.9442
Niemeijer et al. [38]	—	—	0.9416
Zana and Klein [39]	—	—	0.9377
Martinez-Perez et al. [40]	0.7246	0.9655	0.9344
Odstrčilík et al. [7]	0.7060	0.9693	0.9340
Espona et al. (subpixel accuracy) [41]	0.7313	0.9600	0.9325
Chaudhuri et al. [4]	0.6168	0.9741	0.9284
Al-Diri and Hunter [22]	—	—	0.9258
Espona et al. (pixel accuracy) [41]	0.6615	0.9575	0.9223
Jiang and Mojon [42]	—	—	0.9212
All background	—	—	0.8727

“—” indicates that this information was not available.