Table 7.

Algorithms for the detection of exudates.

Authors	Method	Database(s) used	No. of images	Performance measure
Ravishankar et al. [22]	Mathematical morphology	Non-public dataset, DIARETDB0, DRIVE, STARE	516	SE 0.957, SP 0.942
Walter et al. [100]	Mathematical morphology	Non-public dataset	30	SE 0.928, PPV 0.924
Sopharak et al. [101]	Optimally adjusted morphological operators	Non-public dataset	60	SE 0.80, SP 0.995
Welfer et al. [102]	Mathematical morphology	DIARETDB1	89	SE 0.7048, SP 0.9884
Sopharak et al. [103]	Fuzzy c-means clustering, morphological operators	Non-public dataset	40	SE 0.8728, SP 0.9924
Sopharak et al. [104]	Naive Bayes and SVM classification	Non-pubic dataset	39	SE 0.9228, SP 0.9852
Sánchez et al. [105]	Linear discriminant classification	Non-public dataset	58	SE 0.88 (FPI 4.83)
Niemeijer et al. [106]	kNN and linear discriminant classification	Non-public dataset	300	SE 0.95, SP 0.86
García et al. [107]	1:MLP, 2:RBF, and 3:SVM classification	Non-public dataset	67	1:SE 0.8814, PPV 0.8072; 2:SE 0.8849, PPV 0.7741; 3:SE 0.8761, PPV 0.8351
Harangi et al. [108]	Active contour fusion, region-wise classification	1:DIARETDB1; 2:HEI-MED	258	1:SE 0.86, PPV 0.84 (lesion level); 1:SE 0.92, SP 0.68 (image level); 2:SE 0.87, SP 0.86 (image level)
Nagy et al. [114]	Majority voting-based ensemble	DIARETDB1	89	SE 0.72, PPV 0.77