Project · Healthcare AI
AutismRI
A working AI prototype that analyzes brain MRI scans to detect autism spectrum disorder earlier and more objectively than behavior-only assessments allow. Built on The Brain. Next step: a dedicated clinical study and FDA clearance.
- Status
- Working prototype
- Raise
- $2.0M · Seed
- Market
- $4B+ healthcare AI
- Next step
- Clinical partner
Where we are
AutismRI is a functioning platform trained on the Autism Brain Imaging Data Exchange (ABIDE) — a publicly available, multi-site U.S. research dataset covering 1,008 subjects across 19 international sites. The model analyzes structural MRI and fMRI together with patient metadata, returning a risk assessment in seconds with Grad-CAM visual explanations that show clinicians which brain regions drove the result.
The public dataset got us a real, working prototype with strong performance metrics. It cannot get us to a clinical product or FDA clearance on its own. The next step is a dedicated clinical study that generates proprietary training data under a controlled protocol — the same data a hospital or imaging center already produces when they run an ADOS-2 assessment.
We are not pre-prototype. We are pre-clinical. There is a meaningful difference.
Prototype performance
Already exceeds the ADOS-2 gold standard
Trained on ABIDE — 1,008 subjects, 520 confirmed autism cases and 488 healthy controls, ages 6–56 — the prototype’s performance exceeds the behavioral assessment benchmark on every metric. These results are from the public-data prototype, not a clinical trial.
| Metric | AutismRI (prototype) | ADOS-2 Benchmark | Improvement |
|---|---|---|---|
| Sensitivity | 86% | 84% | +2% |
| Specificity | 89% | 70% | +19% |
| AUC-ROC | 0.93 | 0.90 | +0.03 |
Sensitivity: correctly identifies autism. Specificity: correctly excludes healthy controls. AUC-ROC: overall discrimination ability. ADOS-2 is the current gold-standard behavioral assessment. Results are from the public-data prototype, not a clinical study.
What we need
Two partners unlock the next stage
The clinical study is the critical path. Everything else — FDA process, product launch, revenue — follows from it. The study requires two things we don’t yet have.
Partner 1
A clinical partner
A hospital, imaging center, or clinical site to run the study, generate the MRI and ADOS-2 data, and help us scope it accurately. The clinical partner also gives us the real study cost — today that number is still an estimate.
Partner 2
A funding partner
A source that funds the clinical study and the associated FDA and operating costs. We build the funding proposal around the clinical partner’s confirmed scope — so the clinical partner comes first.
What the study produces
Four inputs per subject define the whole study. This is exactly what a clinical site already generates when they run an ADOS-2 assessment alongside routine imaging.
- 01
Age
Patient age at time of scan
- 02
Sex
Patient sex, as clinical metadata
- 03
MRI image
Brain MRI / fMRI scan
- 04
ADOS-2 result
The clinical gold-standard assessment outcome
Technical architecture
- AI model
- Multi-branch deep learning · 3D CNN + ResNet · Trained on structural MRI and fMRI together with patient metadata
- Training data
- ABIDE dataset · 1,008 subjects · 520 confirmed autism cases + 488 healthy controls · Ages 6–56 · 19 international research sites
- Explainability
- Grad-CAM heatmaps highlight the specific brain regions that influenced each result — full transparency for the reviewing clinician
- Speed
- Analysis and clinician report delivered in seconds, versus 30–60+ minutes for behavioral assessments
- Compliance
- HIPAA-aligned · Data encryption in transit and at rest · Regulatory-grade documentation
- Infrastructure
- The Brain — Rethink's proprietary AI platform, running on-premise or hybrid inside the clinical partner's environment
The path forward
Five steps from here to the study
- 1
Identify and secure a clinical partner — the first domino
- 2
Confirm the study scope and firm up the true cost
- 3
Build the funding proposal around the confirmed scope
- 4
Secure funding through grants, foundations, and mission-aligned partners
- 5
Run the study and begin the FDA process in parallel
The raise
$2.0M · Seed
Three components. The clinical study cost is an estimate until we confirm scope with a clinical partner — that number gets firmed up before we commit to it publicly.
Clinical study
Estimate — confirmed with the clinical partner before committing
$1.2M
FDA process
Consultant fees through government fees
$200K
Equipment and operating costs
Working estimate
$600K
Total
$2.0M
Business model post-study: SaaS subscription for clinics and hospital systems — per-seat or per-scan pricing. Each additional diagnostic module (ADHD, Alzheimer’s) adds revenue on the same infrastructure without rebuilding the clinical integration.
Clinical partnerships
Interested in becoming our first clinical partner?
If your hospital, imaging center, or clinical site works in pediatric neurology, child psychiatry, or radiology and you’re open to a conversation about running the study — or if you’re a funder aligned with healthcare AI or autism research — we’d like to hear from you.