EAD-Net: A Novel Lesion Segmentation Method in Diabetic Retinopathy Using Neural Networks

Table 1

Evaluation of exudate detection on e_ophtha_EX dataset.

Model	Lesion-level results
	SE	SP	PR	ACC	F1
U-net	79.86	99.97	78.77	99.95	79.31
Playout et al. [23]	80.02	—	78.50	—	79.25
Zheng et al. [22]	94.12	99.98	91.25	99.96	92.66
Fraz et al. [33]	81.20	94.60	90.91	89.25	—
Zhang et al. [29]	74	—	72	—	—
Imani and Pourreza [34]	80.32	99.83	77.28	—	—
Javidi et al. [25]	80.51	99.84	77.30	—	—
Guo et al. [35]	84.17	—	83.45	—	83.81
Proposed EAD-Net	92.77	99.98	89.06	99.97	90.87

SE: sensitivity; SP: specificity; PR: precision; ACC: accuracy; F1: F1 score. are methods based on U-net.