Brain Injury Prediction on dHCP (5-fold cross-val)
81.82AccuracyGloResNet
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| GloResNetParams=5.4 M, GFLOPs=3.9, Category=Lightweight CNNs, Aggregation=Best (peak fold performance)2026.06 | 81.82 | 81 | 87 | 73.3 | 0.534 | 0.861 | |
| GloResNetParams=5.4 M, GFLOPs=3.9, Category=Lightweight CNNs, Aggregation=Avg (mean across 5 folds)2026.06 | 75.18 | 76 | 81 | 78 | 0.45 | 0.83 | |
| 3D DenseNet-121Params=11.2 M, GFLOPs=5.8, Category=Deep CNNs (Over-Parameterized)2026.06 | 64.1 | 52 | 75 | 55 | 0.28 | 0.66 | |
| 3D ResNet-50Params=46.2 M, GFLOPs=8.2, Category=Deep CNNs (Over-Parameterized)2026.06 | 62.45 | 45 | 78 | 48 | 0.245 | 0.61 | |
| Swin-UNet (Enc)Params=28.1 M, GFLOPs=6.4, Category=Transformers (Data-Hungry)2026.06 | 61.2 | 41 | 79 | 46 | 0.215 | 0.59 | |
| ViT-Base (3D)Params=86.4 M, GFLOPs=16.8, Category=Transformers (Data-Hungry)2026.06 | 58.33 | 30 | 85 | 38 | 0.182 | 0.54 | |
| 3D ResNet-18Params=33.1 M, GFLOPs=6.1, Category=Lightweight CNNs2026.06 | 50 | 20 | 80 | 32 | 0 | 0.55 |