BioPrevention: AI-Powered Addiction Recovery Platform

Developed at Junction Hack 2023

An intelligent biohacking platform that leverages machine learning and real-time biometric monitoring to provide personalized intervention strategies for opioid addiction recovery. The system combines physiological data analysis with reinforcement learning to predict optimal intervention timing and intensity.

🎯 Overview

BioPrevention uses a multi-modal approach to addiction recovery support:

• Real-time Biometric Monitoring: Tracks heart rate, HRV, sleep patterns, and mood indicators
• AI-Driven Predictions: Deep Q-Network (DQN) model predicts intervention needs
• Personalized Interventions: Three-tier support system from gentle notifications to professional help
• Interactive Dashboard: Real-time visualization of physiological state and recommendations

🏗️ Architecture

The platform consists of three integrated components:

1. R Shiny Dashboard (/R/)

• Interactive web dashboard built with golem framework
• Real-time biometric data visualization
• Integration with ML prediction API
• bs4Dash modern UI components

2. Python ML Backend (/app/)

• FastAPI server with PyTorch DQN model
• Processes 6 biometric features: heart rate, HRV bands, sleep, mood
• Returns intervention recommendations (1: notification, 2: activity, 3: professional help)

3. JavaScript Visualization (/jsEx/)

• Chart.js-based real-time monitoring dashboard
• Standalone biometric visualization interface

🚀 Quick Start

Prerequisites

• R (≥ 4.0) with packages: shiny, bs4Dash, golem, echarts4r
• Python (≥ 3.8) with pip
• Modern web browser

Installation

1. Clone the repository

   git clone https://github.com/yourusername/junctionHack2023.git
   cd junctionHack2023

2. Setup Python environment

   pip install -r requirements.txt

3. Install R dependencies

   # In R console
   install.packages(c("shiny", "bs4Dash", "golem", "echarts4r", "ggplot2"))

Running the Application

Option 1: Full Stack (Recommended)

1. Start Python API server

   cd app
   python main.py

   API will be available at http://localhost:8000

2. Launch R Shiny Dashboard

   # In R console from project root
   junctionHack2023::run_app()

Option 2: Individual Components

Python API only:

uvicorn app.main:app --host 0.0.0.0 --port 8000

JavaScript Dashboard only:

cd jsEx
python -m http.server 8080
# Open http://localhost:8080

📊 API Usage

Prediction Endpoint

POST /prevent

Predicts intervention level based on current biometric state.

curl -X POST "http://localhost:8000/prevent" \
     -H "Content-Type: application/json" \
     -d '{
       "heartrate": 0.7,
       "lowFreq": 0.3,
       "highFreq": 0.8,
       "ratio": 0.4,
       "sleep": 0.6,
       "mood": 0.5
     }'

Response:

{
  "action": 2
}

Action Codes:

• 1: Light intervention (positive reinforcement notification)
• 2: Moderate intervention (social activity, exercise)
• 3: Strong intervention (contact healthcare provider)

Input Parameters

Parameter	Type	Range	Description
heartrate	float	0.0-1.0	Normalized heart rate
lowFreq	float	0.0-1.0	HRV low frequency component
highFreq	float	0.0-1.0	HRV high frequency component
ratio	float	0.0-1.0	LF/HF ratio
sleep	float	0.0-1.0	Sleep quality score
mood	float	0.0-1.0	Mood indicator

🧠 Machine Learning Model

The core ML component uses a Deep Q-Network (DQN) architecture:

• Input Layer: 6 biometric features
• Hidden Layers: 16 neurons each with ReLU activation
• Output Layer: 3 action classes
• Framework: PyTorch
• Training: Reinforcement learning for optimal intervention timing

Model Architecture

Input(6) → Linear(16) → ReLU → Linear(16) → ReLU → Linear(3)

🛠️ Development

Project Structure

junctionHack2023/
├── R/                  # Shiny application modules
│   ├── app_ui.R       # UI definition
│   ├── app_server.R   # Server logic
│   └── mod_mainDash.R # Main dashboard module
├── app/               # Python ML backend
│   ├── main.py        # FastAPI server
│   ├── DQN.py         # Neural network model
│   └── modelInit.py   # Model initialization
├── jsEx/              # JavaScript visualization
│   ├── index.html     # Dashboard HTML
│   ├── script20.js    # Chart.js logic
│   └── styles.css     # Styling
├── biohacker.pt       # Pre-trained model weights
└── requirements.txt   # Python dependencies

Adding New Features

1. New Biometric Input: Update State model in app/main.py and retrain DQN
2. UI Modifications: Edit R Shiny modules in /R/ directory
3. Visualization: Extend Chart.js implementation in /jsEx/

Testing

# Test API endpoint
python -c "
import requests
response = requests.post('http://localhost:8000/prevent',
                        json={'heartrate': 0.5, 'lowFreq': 0.3, 'highFreq': 0.7, 'ratio': 0.4, 'sleep': 0.8, 'mood': 0.6})
print(response.json())
"

🚀 Deployment

Heroku Deployment

The application is configured for Heroku deployment with the included Procfile:

# Deploy Python API
heroku create your-app-name
git push heroku main

Alternative Deployment

• R Shiny: Deploy to shinyapps.io or RStudio Connect
• Python API: Deploy to AWS Lambda, Google Cloud Run, or any container platform

🤝 Contributing

1. Fork the repository
2. Create a feature branch (git checkout -b feature/AmazingFeature)
3. Commit your changes (git commit -m 'Add some AmazingFeature')
4. Push to the branch (git push origin feature/AmazingFeature)
5. Open a Pull Request

📜 License

This project was developed for Junction Hack 2023. Please ensure responsible use of addiction recovery technology.

🙏 Acknowledgments

• Junction Hack 2023 organizers and participants
• PyTorch and R Shiny communities
• Healthcare professionals who provided domain expertise

---

For technical questions or collaboration opportunities, please open an issue or contact the development team.