README.md

metadata

title: MRI-JEPA
emoji: 🧠
colorFrom: blue
colorTo: indigo
sdk: gradio
sdk_version: 6.14.0
app_file: app.py
pinned: false
license: mit

MRI-JEPA — Brain MRI Analysis

Drop a NIfTI scan (.nii / .nii.gz) for:

• Skull stripping — multi-Otsu + morphological operations
• I-JEPA embedding — ViT-H/14 patch mean (facebook/ijepa_vith14_1k)
• Pathology classification — KNN + MLP (balanced acc. 0.965)
• Age regression — KNN + MLP (MAE 6.2 yrs, R² 0.866)

Trained on FOMO60K (~48K brain MRI scans across 16 datasets).

By Shunya Research