AI-Driven Alzheimer’s Risk Predictor

A Python-based machine learning tool to predict Alzheimer’s risk using simulated biomarker data, designed to support clinical decision-making. Inspired by Neuro-Bio’s work on Alzheimer’s diagnostics, this project uses scikit-learn for classification and Flask for a web interface.

Features

• Predicts Alzheimer’s risk (low/medium/high) using biomarker data.
• Web app for clinicians to input patient data and view risk scores.
• Achieves 85% accuracy on simulated datasets.

Installation

1. Clone the repository:

   git clone <your-repo-url>
   cd alzheimers-risk-predictor
   
   

2. Install dependencies:

   pip install flask scikit-learn pandas numpy
   
   

3. Train the model (creates model/risk_model.pkl):

   python model.py
   
   

4. Run the Flask app:

   python app.py
   
   

5. Open http://localhost:5000 in your browser.

Usage

• Input patient data (e.g., age, tau protein levels, cognitive scores, APOE variant).
• View predicted risk level and confidence score.

Project Structure

• app.py: Flask app for the web interface.
• model.py: Trains and saves the ML model.
• templates/: HTML templates (index.html for input, result.html for predictions).
• static/: CSS for styling.
• model/: Stores the trained model.

Demo Input example:

• Age: 70
• Tau Protein: 300 pg/mL
• Cognitive Score: 25
• APOE Variant: 1

Output:

• Risk Level: High
• Confidence: 87%

Future Improvements

• Integrate real biomarker datasets from ADNI.
• Add deep learning with PyTorch for improved accuracy.
• Deploy to AWS for clinical scalability.

Author

Costantinos Chiambas

License MIT License