Table 5 The classification results of the similar studies reported in the literature and the model proposed in this study.
From: A deep learning model for diagnosis of inherited retinal diseases
References | Image type | Disease | Model | Results |
|---|---|---|---|---|
Ta-Ching Chen et al.40 | Color Fundus (1670) | RP | Xception | AUC-ROC: 96.74%/ Sensitivity: 95.71% Accuracy: 96.00 |
Guo et al.17 | Color Fundus (250) | Glaucoma, maculopathy, pathological myopia, RP | MobileNetV2 | Accuracy: 96.2% Sensitivity: 90.4% Specificity: 97.6% |
Fujinami -Yokokawa et al. 41 | Color Fundus and FAF (417) | Stargardt, RP, occult macular dystrophy | InceptionV3 | Fundus: Accuracy: 88.2% Sensitivity/Specificity: 88.3%/97.4% FAF: Accuracy: 81.3% Sensitivity/Specificity: 81.8%/95.5% |
Shah et al.19 | OCT (102) | STGD | CNN | Accuracy: 99.6% Sensitivity: 99.8% Specificity:98.0% |
Masumoto et al.18 | ultrawide-field pseudocolor and ultrawide-field autofluorescence (373) | RP | CNN | Pseudocolor : Sensitivity: 99.3% Specificity: 99.1% Autofluorescence: Sensitivity: 100% Specificity: 99.5% |
This study | Color Fundus and IR (233) | RP, Stargardt | Multi-Input MobileNetV2 | Accuracy: 96.3% Sensitivity: 96.3% Specificity: 97.92% AUC-ROC:99.31% |
