🤖 Autism Screening Agent — Clinical Behavioral Diagnostic Pipeline

An intelligent, accessible clinical screening pipeline and interactive diagnostic workbench designed for early developmental screening in children. Combines validated psychometric questionnaire scoring (AQ-10 / M-CHAT-R) with multi-modal machine learning and human-in-the-loop clinical review.

Developed by Jaswanth Reddy (@Jaswanth1902) to bridge clinical evaluation waitlists and bring compassionate, objective screening tools to parents and healthcare workers.

---

💡 Why Autism Screening Agent?

Early identification of Autism Spectrum Disorder (ASD) dramatically improves childhood developmental interventions. However, families often face 6 to 18-month waitlists for specialized clinical evaluations.

This agent was built to provide an accessible, objective, and privacy-preserving first line of triage:

• Validated Questionnaires: Digitized, human-friendly AQ-10 and M-CHAT-R scoring protocols.
• Ensemble Machine Learning: Pre-trained stacking classifiers analyzing behavioral trait clusters.
• 3-Tier Compassionate Results:
  • 🟢 Low Risk: Reassuring feedback and developmental milestone tracking.
  • 🟡 Moderate Risk: Monitoring suggestions and pediatric follow-up prompts.
  • 🟠 High Risk: Direct, compassionate guidance to accredited developmental pediatricians.
• 100% Patient Privacy: Local-first processing; questionnaire answers are never monetized or exfiltrated.

---

🏗️ Diagnostic Architecture

flowchart LR
    Caregiver([Caregiver / Clinical User]) --> UI[React 18 Interactive Assessment]
    UI --> Sanitizer[Input Validator & Anonymizer]
    Sanitizer --> Scorer[Deterministic Scoring Engine\n(AQ-10 & M-CHAT-R Algorithms)]
    
    subgraph MLPipeline["Machine Learning Assessment Engine"]
        Scorer --> FeatureExtractor[Trait Cluster Vectorizer]
        FeatureExtractor --> StackingModel[Stacking Classifier\nXGBoost + Random Forest]
        StackingModel --> RiskClassifier[Probabilistic ASD Risk Score]
    end
    
    RiskClassifier --> ClinicalReport[Actionable Diagnostic Report\n+ Local Care Resources]
    ClinicalReport --> UI

Loading

---

🚀 Quick Start & Setup

1. Backend Server (Flask)

pip install -r requirements.txt
python app.py

2. Frontend Interface (React)

cd autism_web/autism-website
npm install
npm start

Access the screening workbench at http://localhost:3000.

---

🛡️ Patient Privacy & Clinical Hardening

1. Zero Personally Identifiable Information (PII): The application does not collect child names, addresses, or phone numbers. Only anonymous questionnaire integers are evaluated.
2. Local Machine Execution: The entire Flask and ML scoring stack runs strictly on 127.0.0.1.
3. Clinical Boundaries: This tool is an initial behavioral screening instrument, not a formal medical diagnosis. It is designed to assist, never replace, certified developmental pediatricians.

See SECURITY.md for vulnerability reporting and privacy disclosures.

---

👤 Author & Maintainer

Jaswanth Reddy

• GitHub: @Jaswanth1902
• Email: jaswanthreddy1537@gmail.com

---

📄 License

Licensed under the MIT License.