Skip to main content
Fig. 7 | BioMedical Engineering OnLine

Fig. 7

From: Automatic diagnosis of imbalanced ophthalmic images using a cost-sensitive deep convolutional neural network

Fig. 7

Performance comparison of the CS-ResCNN method and various conventional methods. Two sets of deep learning methods and 16 sets of conventional methods were evaluated using accuracy, sensitivity and specificity indicators. ad The four conventional methods WT, LBP, SIFT and COTE, respectively, compared with three data-level methods; e the CS-ResCNN method and five representative conventional methods (ResCNN, SIFT-UNDER, COTE-UNDER, WT-UNDER and LBP-UNDER). CS-ResCNN, cost-sensitive residual convolutional neural network; ResCNN, native residual convolutional neural network; WT, wavelet transformation; LBP, local binary pattern; SIFT, scale-invariant feature transform; COTE, color and texture features; SMOTE, synthetic minority over-sampling technique; BSMOTE, borderline-SMOTE; UNDER, under-sampling

Back to article page