Skip to main content

Table 5 Results comparison with existing methods on complete DRIVE database

From: A framework for retinal vasculature segmentation based on matched filters

No Type Methods Year Sen Spe ACC AUC
1   2nd human observer 0.7796 0.9717 0.9470
2 Unsupervised methodology Zana et al. [27] 2001 0.6971 0.9377 0.8984
3 Jiang et al. [18] 2003 0.9212 0.9114
4 Mendonca et al. [29] 2006 0.7344 0.9764 0.9452
5 Al-Diri et al. [13] 2009 0.7282 0.9551
6 Lam et al. [12] 2010 0.9472 0.9614
7 Miri et al. [26] 2011 0.7352 0.9795 0.9458
8 Fraz et al. [28] 2011 0.7152 0.9759 0.9430
9 You et al. [39] 2011 0.7410 0.9751 0.9434
10 Zhao et al. [32] 2014 0.7354 0.9789 0.9477
11 Proposed method 2015 0.7489 0.9818 0.9529 0.9664
12 Supervised methodology Niemeijer et al. [34] 2004 0.9416 0.9294
13 Soares et al. [35] 2006 0.7332 0.9782 0.9461 0.9614
14 Staal et al. [33] 2004 0.9441 0.9520
15 Ricci et al. [11] 2007 0.9595 0.9558
16 Lupascu et al. [36] 2010 0.7200 0.9597 0.9561
17 Marin et al. [37] 2011 0.7067 0.9801 0.9452 0.9588
18 Fraz et al. [41] 2012 0.7406 0.9807 0.9480 0.9747
19 Vega et al. [38] 2015 0.7444 0.9600 0.9412
  1. –, means the value is not answered in the reference paper
  2. Italic numbers denote the best cases among unsupervised methods and the comparatively better results among supervised ones in terms each evaluation criterion