AutiScan

AutiScan uses advanced eye-tracking to identify autism traits non-invasively, analyzing visual patterns for a scientifically-backed assessment in just seconds.

Comment

Inspiration

1.** High cost of autism detection in India**: ₹50,000–₹2,00,000 per evaluation (private clinics like Apollo or NIMHANS).

  1. Delayed diagnosis: Average age of autism diagnosis in India is 5–6 years (vs. signs at 18 months).
  2. Limited access: Only 1 child psychiatrist per 2 lakh children in rural areas (Indian Journal of Psychiatry, 2022).
  3. Early intervention gap: 70% of autistic children miss early therapy due to cost and awareness (AIIMS study).

What it does

  1. Eye-tracking test: Tracks gaze while following a moving dot on-screen.
  2. Real-time analysis: Detects autism-linked patterns (e.g., less focus on social cues). 3.** Visual report**: Shows gaze path and risk level (low/medium/high).
  3. Risk assessment: Based on tracking accuracy (e.g., <70% linked to ASD traits).
  4. Secure login: Uses Clerk Authentication for privacy.

How we built it

  1. Framework: Next.js for fast, scalable web app development.
  2. Eye-tracking: WebGazer.js captures gaze via webcam.
  3. Algorithm: Custom code compares gaze accuracy to ASD research (e.g., 75% threshold). 4.** Design**: Tailwind CSS for responsive, user-friendly UI.
  4. Authentication: Clerk for secure sign-up/login.
  5. Visualization: Chart.js for interactive gaze reports. 7.Hosting: Deployed on Vercel for accessibility.

Challenges we ran into

  1. Lighting issues: Indian homes’ uneven lighting affects gaze tracking.
  2. Algorithm tuning: Hard to balance sensitivity for ASD traits vs. false positives.
  3. Next.js conflict: Server-side rendering clashed with webcam access.

Accomplishments that we're proud of

  1. Affordable tool: Cuts ₹50,000+ diagnostic costs to near-zero.
  2. Working prototype: Real-time eye-tracking on a web app.
  3. Accuracy: Matches 75–80% sensitivity of research studies.
  4. Secure Authentication: Clerk Authentication protects user data.
  5. Impact potential: Reaches India’s 1.5 million autistic kids.

What we learned

  1. Next.js tricks: Client-side focus solved rendering issues. 2.Autism insights: ASD kids show 20–30% less gaze accuracy (research). 3.Early detection value: 70% outcome boost before age 5 (Indian studies). 4.User needs: Privacy and simplicity matter in health apps.

What's next for AutiScan

1.Boost accuracy: Add ML models trained on Indian eye-tracking data. 2.Multi-language: Support Hindi, Tamil, etc., for rural reach. 3.Scale up: Free/low-cost rollout via NGOs like Action for Autism. 4.Regulatory goal: Seek CDSCO approval as a screening tool.

Built With

  • blazeface
  • blazeface-for-face-detection-data-visualization:-canvas-api-for-eye-movement-visualization-testing:-jest
  • canvas
  • clerk
  • frontend:-next.js-15
  • jest
  • react
  • react-19
  • tailwind
  • tailwind-css-authentication:-clerk-state-management:-zustand-ai/ml-components:-tensorflow.js
  • tensorflow.js
  • typescript
  • zustand
Share this project:

Updates