Retinal OCT Classification on NEH (five-fold patient-level cross-validation)
0.926AccuracyKD-OCT (ConvNeXtV2-Large)
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| KD-OCT (ConvNeXtV2-Large)Pram(mil)=196.42025.12 | 0.926 | 0.929 | 0.981 | |
| KD-OCT (EfficientNet-B2)Pram(mil)=7.72025.12 | 0.9246 | 0.9215 | 0.9604 | |
| FPN-VGG16Pram(mil)=21.62025.12 | 0.92 | 0.918 | 0.958 | |
| VGG16Pram(mil)=28.32025.12 | 0.916 | 0.914 | 0.956 | |
| FPN-DenseNet121Pram(mil)=14.32025.12 | 0.909 | 0.905 | 0.952 | |
| FPN-ResNet50Pram(mil)=31.12025.12 | 0.901 | 0.898 | 0.948 | |
| DenseNet121Pram(mil)=72025.12 | 0.9 | 0.897 | 0.947 | |
| FPN-EfficientNetB0Pram(mil)=12.72025.12 | 0.878 | 0.866 | 0.933 | |
| ResNet50Pram(mil)=23.62025.12 | 0.868 | 0.864 | 0.93 | |
| EfficientNetB0Pram(mil)=42025.12 | 0.854 | 0.845 | 0.921 | |
| MedSigLIPPram(mil)=430.42025.12 | 0.845 | 0.8181 | 0.9442 | |
| Kermany et al.Pram(mil)=0.022025.12 | 0.839 | 0.829 | 0.914 | |
| SF netPram(mil)=29.22025.12 | 0.826 | 0.804 | 0.962 | |
| Kaymak et al.Pram(mil)=58.32025.12 | 0.802 | 0.8 | 0.894 | |
| Thomas et al.Pram(mil)=2.52025.12 | 0.685 | 0.691 | 0.838 | |
| HOG + SVM2025.12 | 0.672 | 0.6699 | 0.743 |