Table 2.

Performance metrics of the deep learning classifier for retinal vascular diseases on ultra-widefield colour fundus photographs

	Accuracy	AUC-ROC	Sensitivity	Specificity	Precision	F1 score
Diabetic retinopathy	85.2	90.5	67.6	92.4	83.1	75.4
Retinal vein occlusion	88.4	91.2	78.7	91.0	77.2	83.3
Sickle cell retinopathy	93.8	96.7	94.7	93.4	91.2	94.2
Healthy controls	86.2	88.5	67.2	92.3	77.8	72.3

Metrics are expressed as percentage (%).

AUC-PR, area under the precision-recall curve; AUC-ROC, area under receiver operating characteristics curve.