BrainConnect-ASD / README.md
Yatsuiii's picture
Point Space to org-owned interpreter model
352e076
metadata
title: BrainConnect ASD
emoji: 
colorFrom: red
colorTo: gray
sdk: gradio
sdk_version: 6.14.0
python_version: '3.13'
app_file: app.py
pinned: true
license: apache-2.0
short_description: ASD detection from brain connectivity
tags:
  - amd
  - amd-hackathon-2026
  - medical-ai
  - brain-connectivity
  - gradio

BrainConnect-ASD

Site-invariant Autism Spectrum Disorder classification from resting-state fMRI, built for the AMD Developer Hackathon 2026.

Models

Model Description
Yatsuiii/brain-connect-gcn Adversarial Brain Mode Network — 20-model LOSO GCN ensemble, AUC 0.7298 cross-site
lablab-ai-amd-developer-hackathon/asd-interpreter-merged Qwen2.5-7B fine-tuned clinical interpreter — generates natural language reports from saliency scores

Pipeline

fMRI (.1D) → preprocessing → FC matrix → GCN ensemble (20 models) → p(ASD)
                                                    ↓
                                         gradient saliency → Qwen2.5-7B → clinical report
  • 60 total models trained across 3 atlases (CC200, AAL, Harvard-Oxford) × 20 LOSO folds
  • AMD MI300X (ROCm 7.0) used for all training and LLM inference
  • ~20ms end-to-end inference per subject (preprocessing + 20-model ensemble)
  • 1,102 subjects · 20 acquisition sites · cross-site evaluation only