Table 8.

AUC, accuracy, sensitivity and specificity related to the dark-lesion detection (microaneurysms and hemorrhages) on DIARETDB1 database for each classification method studied in this work using a decision threshold $\delta =0.5$.

	Accuracy	Sensitivity	Specificity	AUC
Random Forests	0.9016 ± 0.0311	0.1718 ± 0.0800	0.9818 ± 0.0035	0.8150 ± 0.0356
Linear-SVM	0.7404 ± 0.0134	0.6969 ± 0.0865	0.7462 ± 0.0245	0.7975 ± 0.0367
RBF-SVM	0.7603 ± 0.0223	0.7299 ± 0.0821	0.7576 ± 0.0311	0.8209 ± 0.344
Gaussian Processes	0.7612 ± 0.0244	0.7489 ± 0.0844	0.7630 ± 0.0350	0.8344 ± 0.0330