Deep metric learning for otitis media classification

In this study, we propose an automatic diagnostic algorithm for detecting otitis media based on otoscopy images of the tympanic membrane. A total 1336 were assessed by a medical specialist into three groups: acute media, with effusion, and no effusion. To provide proper treatment care limit use unnecessary antibiotics, it is crucial to correctly detect membrane abnormalities, distinguish between The proposed approach classification task deep metric learning, study compares performance different distance-based loss functions. Contrastive loss, triplet multi-class N-pair are employed, compared standard cross-entropy class-weighted networks. Triplet achieves high precision highly imbalanced data set, methods useful insight decision making neural network. results comparable best clinical experts paves way more accurate operator-independent diagnosis media.