Table 3.

AUC, accuracy, sensitivity and specificity related to the bright-lesion detection on E-OPHTHA exudates database for each classification method using a decision threshold $\delta =0.5$.

	Accuracy	Sensitivity	Specificity	AUC
Random Forests	0.9508 ± 0.0084	0.4785 ± 0.1015	0.9921 ± 0.0043	0.9256 ± 0.0173
Linear-SVM	0.8533 ± 0.0245	0.7721 ± 0.0857	0.8651 ± 0.0399	0.8948 ± 0.0351
RBF-SVM	0.8796 ± 0.0229	0.8118 ± 0.0618	0.8851 ± 0.0296	0.9240 ± 0.0161
Gaussian Processes	0.8762 ± 0.0206	0.8348 ± 0.0650	0.8795 ± 0.0266	0.9353 ± 0.0174