Binary Alzheimer's Disease Classification (CN vs. AD) on AIBL (test)
98.5AUCMVSC (I-JEPA)
Evaluation results
| Method | Links | ||
|---|---|---|---|
| MVSC (I-JEPA)Backbone=ViT-H/14, Compression strategy=MVSC2026.01 | 98.5 | 93.52 | |
| MVSC (DINOv3)Backbone=ViT-B/16, Compression strategy=MVSC2026.01 | 97.19 | 95.68 | |
| ViTBackbone=ViT-S/162026.01 | 95.75 | 93.53 | |
| RAPTOR2026.01 | 95.06 | 94.96 | |
| M3T2026.01 | 94.96 | 95.07 | |
| MERLINMode=image encoder only2026.01 | 94.49 | 91.37 | |
| 3D ResNetBackbone=ResNet502026.01 | 93.95 | 89.93 | |
| Mean (I-JEPA)Backbone=ViT-H/14, Compression strategy=Mean pooling2026.01 | 91.94 | 90.53 | |
| Max (I-JEPA)Backbone=ViT-H/14, Compression strategy=Max pooling2026.01 | 88.9 | 80.57 | |
| Mean (DINOv3)Backbone=ViT-B/16, Compression strategy=Mean pooling2026.01 | 83.28 | 82.73 | |
| Max (DINOv3)Backbone=ViT-B/16, Compression strategy=Max pooling2026.01 | 71.79 | 83.45 | |
| PPAL2026.01 | 61.89 | 90.47 | |
| AXIAL2026.01 | 51.92 | 81.34 |