Open-source screening-support over a public ABIDE I reference space. It shows how one subject's brain-imaging features deviate from a normal reference, which evidence markers contribute, and how uncertain the read is.
Decision-first triage for one subject: a verdict band, the recommended next step, and a behavior-first anchor. Covers all 1257 subjects.
Open summary →
Seven interactive tabs: functional-network deviation map, 3D brain, age/sex norms, cortical surface, evidence markers, model facts. Each opens with a plain-language clinical lens.
Ships a 40-subject offline sample — opens with no install, no server.
Open in-depth →
Does NOT rule out autism.
Defer — rely on behavioral assessment.
A statistical signal, not a diagnosis.
✅ Behavioral assessment (ADOS · SRS · developmental history) stays the clinical standard. Imaging is an adjunct.
| Number | What it is |
|---|---|
| AUROC 0.91 | Internal validated model — public release pending. The headline target; its data/weights are not in this repo, and the live demo does not use it. |
| AUROC 0.626 | The model this demo actually runs: ABIDE I, leave-one-site-out (honest cross-site). Per-site AUC 0.31–0.73 across 20 sites. |
| AUROC 0.743 | Clinic pooled 5-fold (in-sample, optimistic vs LOSO). |
Leakage guard: any pooled cross-site AUC above ~0.78 on ABIDE is a red flag. The deployed honest number is 0.626 LOSO.
Warm / red = ASD-leaning (higher / excess) · Cool / blue = Normal-leaning (lower / deficit) · deeper color = larger deviation. A result "toward ASD" is a statistical deviation, not a diagnosis; "toward Normal" does not rule autism out.