Fig. 4: Screening performance of the FairerOPTH on the OculoScope dataset.
From: Fairer AI in ophthalmology via implicit fairness learning for mitigating sexism and ageism
a The FairerOPTH consists of two branches, pathology and disease classification that predict 67 fundus features and 38 ophthalmic diseases, respectively. Such design of the two branches aims to enhance the disease representation in the disease classification branch, resulting in higher screening accuracy. b Comparison of FairerOPTH with the baseline model and state-of-the-art multi-label classification methods using mAP, specificity, sensitivity, and AUC (area under the curve) evaluation metrics. c, d ROC (Receiver Operating Characteristics) curves for 38 diseases of baseline model and FairerOPTH. Source data are provided as a Source Data file.
