🏥 AI Medical Assistant

A real-time AI-powered medical tool that transcribes patient-healthcare professional interactions and extracts key insights using OpenAI Whisper and GPT-4.

Features

✅ Phase 1: Nurse Interaction Interface (Completed)

• 🎤 Real-time Voice Recording - Click-to-record microphone interface
• 📝 Live Transcription - Powered by OpenAI Whisper
• 🤖 AI Summarization - GPT-4 generates medical summaries
• 💾 Data Persistence - Saves summaries to JSON files
• 🎨 Beautiful UI - Modern, responsive interface

🚧 Coming Next

• Doctor-patient interaction with real-time insights
• Vector similarity search for relevant medical history
• WebSocket real-time updates
• Medical record integration and updates

Tech Stack

• Backend: FastAPI + Python
• AI: OpenAI Whisper (transcription) + GPT-4 (summarization)
• Frontend: Vanilla HTML/CSS/JavaScript
• Storage: JSON files (simple, no database needed)

Setup Instructions

1. Prerequisites

• Python 3.8+ (tested with Python 3.13)
• OpenAI API account and key
• Microphone access in your browser

2. Installation

# Clone or download the project
cd hackathon

# Create and activate virtual environment
python3 -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Install dependencies
pip install fastapi "uvicorn[standard]" openai python-multipart aiofiles python-dotenv websockets
pip install git+https://github.com/openai/whisper.git

3. Configuration

Create a .env file in the project root:

# Get your API key from: https://platform.openai.com/api-keys
OPENAI_API_KEY=your_openai_api_key_here

4. Running the Application

# Start the server
python main.py

# Or alternatively:
uvicorn main:app --reload --host 0.0.0.0 --port 8000

The application will be available at: http://localhost:8000

Usage

Nurse Interface Workflow

1. Access the Interface: Open http://localhost:8000 in your browser
2. Patient Context: See the current patient info (Sarah Michelle Johnson)
3. Record Conversation:
   • Click the microphone button to start recording
   • Speak your nurse-patient conversation
   • Click again to stop recording
4. View Transcription: See the live transcription appear
5. Generate Summary: Click "Generate Summary" for AI analysis
6. Save to Record: Click "Save to Record" to persist the data

Features in Action

• Real-time Transcription: Using Whisper's "base" model for speed
• Medical Context: GPT-4 receives patient medical history for context
• Professional Summaries: AI generates clinical documentation
• Persistent Storage: All interactions saved to summaries/ directory

File Structure

hackathon/
├── main.py                 # FastAPI application
├── medical_record.md       # Patient medical history
├── requirements.txt        # Python dependencies
├── summaries/             # Generated summaries directory
│   ├── all_summaries.json # Master file with all summaries
│   └── summary_*.json     # Individual session files
└── README.md              # This file

API Endpoints

• GET / - Main nurse interface
• POST /transcribe - Upload audio for transcription
• POST /summarize - Generate medical summary
• POST /save-summary - Save summary to JSON
• GET /summaries - Retrieve all saved summaries

Sample Medical Summary

The AI generates professional medical notes like:

**Chief Complaint**: Patient reports chest discomfort and shortness of breath with exertion over the past week.

**Vital Signs**: Blood pressure elevated at 145/92 mmHg, heart rate 88 bpm.

**Assessment**: Given patient's history of diabetes, hypertension, and family cardiac history, chest symptoms warrant further cardiac evaluation.

**Plan**: ECG ordered, cardiology referral initiated, continue current medications.

Troubleshooting

Common Issues

1. Microphone not working: Ensure browser permissions for microphone access
2. OpenAI API errors: Check your API key in .env file
3. Whisper loading slowly: First model load takes time, subsequent uses are faster
4. Module not found: Ensure virtual environment is activated and dependencies installed

Performance Tips

• Use Chrome/Firefox for best microphone support
• Speak clearly for better transcription accuracy
• Keep recordings under 30 seconds for optimal processing
• The "base" Whisper model balances speed and accuracy

Contributing

This is a hackathon project! Feel free to:

• Add new features
• Improve the UI/UX
• Optimize performance
• Add error handling
• Implement additional medical workflows

Next Steps

Based on your PRD, the next features to implement:

1. Doctor interface with real-time insights
2. Vector similarity search using sentence-transformers
3. Medical record highlighting based on conversation
4. WebSocket real-time updates

License

MIT License - Built for the hackathon!