AD vs. CN binary classification on Tau-PET (5-fold cross-val)
91.9AccuracyResNet-18
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| ResNet-18Architecture=3D CNN2026.05 | 91.9 | 76.6 | 57.14 | 65.45 | 90.96 | 0.6182 | |
| Agent EnsembleStrategy=Stratified stacking of modality-specific predictions2026.05 | 91.47 | 75.56 | 53.97 | 62.96 | 90.92 | 0.5935 | |
| DenseNet-121Architecture=3D CNN2026.05 | 89.13 | 56.98 | 77.78 | 65.77 | 91.59 | 0.6051 | |
| EfficientNet-B0Architecture=3D CNN2026.05 | 82.52 | 39.08 | 53.97 | 45.33 | 75.1 | 0.3589 |