🧠 AI Tool for Early Stage Dementia Detection

An AI-powered web application that detects early signs of dementia through cognitive screening tests, scored by an ensemble machine learning model.

⚠️ Important Disclaimer

This tool is NOT a medical diagnosis system. It does not diagnose, treat, or cure dementia or any other disease. Results are for screening and awareness purposes only and should never replace a professional medical consultation. If you or someone you know is experiencing cognitive difficulties, please consult a qualified medical professional.

📌 Overview

MindGuard screens users for early dementia indicators through a series of cognitive tasks — memory recall, clock drawing, math problems, shape recognition, and story comprehension. Responses are scored by a trained ensemble ML model (Random Forest + others) to produce a risk assessment.

✨ Features

• 🧪 6-variant cognitive screening test sets (randomized per session)
• 🤖 Ensemble ML model for dementia risk prediction
• 🎙️ Voice assistant for accessibility
• 📊 Dashboard with results history and visual reports
• 🎮 Brain training games (Chess, Sudoku, Duck game)
• 🔐 Auth system with protected routes
• 📄 PDF report generation

🛠️ Tech Stack

Layer	Technology
Backend	Python, Flask, Flask-CORS
ML	scikit-learn, joblib, NumPy, pandas, librosa
Frontend	React 18, TypeScript, Vite, TailwindCSS, shadcn/ui
Database	SQLite
State	TanStack Query, React Hook Form, Zod

⚙️ Setup & Installation

Prerequisites

• Python 3.10+
• Node.js 18+

Backend

git clone https://github.com/CodeKatanaa/Ai-Early-Stage-Dementia-Detection.git
cd Ai-Early-Stage-Dementia-Detection
python -m venv venv
venv\Scripts\activate
pip install -r requirement.txt
python App.py

Frontend

cd frontend
npm install
npm run dev

Backend runs on http://localhost:5000 — Frontend on http://localhost:5173

📁 Project Structure

├── App.py                  # Flask backend & ML inference API
├── tr1.py                  # Model training script
├── ev1.py                  # Model evaluation script
├── ts1.py                  # Testing script
├── requirement.txt         # Python dependencies
├── models/
│   ├── ensemble_model.pkl  # Trained ensemble model
│   ├── random_forest.pkl   # Random forest model
│   ├── scaler.pkl          # Feature scaler
│   └── config.json         # Model config & thresholds
├── visuals/                # Training plots & metrics
├── csv files/              # Dataset files
├── datapreparation codes/  # Data preprocessing scripts
└── frontend/               # React + TypeScript frontend
    └── src/
        ├── pages/          # Route pages
        ├── components/     # Reusable UI components
        ├── services/       # API layer
        ├── contexts/       # Auth context
        └── hooks/          # Custom hooks

📊 ML Model Performance

Training visuals available in the /visuals folder:

• Model comparison
• Confusion matrices
• ROC curves
• Precision-recall curves
• Feature importance

🤝 Contributing

Pull requests are welcome! Please open an issue first for major changes.

📄 License

MIT © CodeKatanaa