| --- |
| license: cc-by-nc-4.0 |
| tags: |
| - brain-age |
| - neuroimaging |
| - mri |
| - 3d-cnn |
| - pytorch |
| --- |
| |
| # BrainAge — Dual-Branch Brain Age Predictor (resnet18) |
|
|
| 3D RESNET18 + tabular MLP trained on **6,050 healthy T1w brain MRIs** (0–86 years). |
|
|
| ## Checkpoints |
|
|
| | File | Description | |
| |------|-------------| |
| | `brainage_resnet18.pt` | Phase A: lifespan pretraining (all ages) | |
| | `brainage_resnet18_finetune.pt` | Phase B: fine-tuned for pediatric (0–25 y) | |
|
|
| ## Training data |
|
|
| [bilalEthizo/BrainAge-Golden-Preprocessed](https://huggingface.co/datasets/bilalEthizo/BrainAge-Golden-Preprocessed) |
|
|
| ## Architecture |
|
|
| - Image: resnet18 3D CNN, input 128×144×112 |
| - Tabular: MLP over 86-dim vector (70 regions + sex + site) |
| - Fusion: concat → 64 → 1 (regression in years) |
| - Loss: HuberLoss(delta=3.0) |
| - Training: fp16 mixed precision, cosine LR, age-bin weighted sampler |
|
|
