Medical Diagnostics AI Agent

Sharan G S

An advanced AI-powered system for analyzing complex medical cases using specialized LLM-based agents. This project uses free, open-source APIs (Ollama and Hugging Face) to provide intelligent medical case analysis, symptom evaluation, and diagnostic suggestions.

⚠️ Medical Disclaimer ⚠️

This system is for educational and informational purposes only. It is NOT a substitute for professional medical advice, diagnosis, or treatment. Always seek the advice of qualified healthcare providers. If you have a medical emergency, call emergency services immediately.

✨ Features

• 🏥 Comprehensive Case Analysis - Multi-step reasoning for complex medical cases
• 🔍 Symptom Extraction - Automatic categorization and analysis of symptoms
• 💊 Differential Diagnosis - AI-generated diagnostic suggestions with confidence levels
• 💬 Interactive Chat - Conversational interface for medical questions
• 📚 PubMed Integration - Access to peer-reviewed medical literature and research citations
• 🌐 Multi-Language Support - Interface and responses in Tamil and English
• 📄 PDF Export - Download analysis reports as professional PDF documents
• 🎤 Voice Input Support - Dictate medical cases using speech-to-text
• 🖼️ Medical Imaging Analysis - Analyze medical images (X-rays, CT scans) for quality and structure
• 🧠 Enhanced RAG System - Vector-based medical knowledge retrieval with semantic search
• 🌐 Web Interface - Modern, responsive web UI
• 🖥️ CLI Interface - Rich command-line interface for power users
• 📚 Knowledge Base - Medical knowledge integration with disease-symptom mappings
• 🔄 Automatic Fallback - Seamless switching between Ollama and Hugging Face APIs

🚀 Quick Start

Prerequisites

• Python 3.8 or higher
• Ollama (recommended) or Hugging Face API key

Installation

1. Clone or download the project

   cd /Users/sharan/Downloads/AI-Agent-for-Medical-Diagnostics

2. Create a virtual environment

   python3 -m venv venv
   source venv/bin/activate  # On macOS/Linux

3. Install dependencies

   pip install -r requirements.txt

4. Set up Ollama (Primary LLM - Recommended)

   # Install Ollama
   brew install ollama  # macOS
   
   # Start Ollama service
   ollama serve
   
   # In a new terminal, pull a model
   ollama pull llama3.2
   # or
   ollama pull mistral

5. Configure environment variables

   cp .env.example .env
   # Edit .env and add your Hugging Face API key (optional, for fallback)

Running the Application

Web Interface (Recommended)

python app.py

Then open your browser to: http://localhost:5000

CLI Interface

# Interactive mode
python cli.py interactive

# Chat mode
python cli.py chat

# Analyze from file
python cli.py analyze-file case.txt

# Check system status
python cli.py status

📖 Usage Examples

Web Interface

1. Navigate to http://localhost:5000
2. Enter a medical case description in the "Case Analysis" section
3. Click "Analyze Case" to get comprehensive analysis
4. Use the "Interactive Consultation" chat for follow-up questions

CLI Interface

# Start interactive session
python cli.py interactive

# Example case input:
Patient: 45-year-old male
Symptoms: Persistent cough for 3 weeks, fever (101°F), night sweats, fatigue
Duration: 3 weeks
Medical History: No significant past medical history

🏗️ Project Structure

AI-Agent-for-Medical-Diagnostics/
├── agents/
│   ├── medical_agent.py      # Main medical AI agent
│   ├── llm_provider.py        # LLM abstraction layer
│   └── prompts.py             # Specialized medical prompts
├── config/
│   └── settings.py            # Configuration management
├── knowledge/
│   ├── medical_kb.py          # Medical knowledge base
│   └── sample_data.json       # Disease and symptom data
├── static/
│   ├── css/style.css          # Modern UI styling
│   └── js/app.js              # Frontend JavaScript
├── templates/
│   └── index.html             # Web interface
├── utils/
│   ├── validators.py          # Input validation
│   └── logger.py              # Logging utilities
├── app.py                     # Flask web application
├── cli.py                     # Command-line interface
├── requirements.txt           # Python dependencies
└── README.md                  # This file

🔧 Configuration

Environment Variables

Edit .env file:

# Ollama (Primary)
OLLAMA_BASE_URL=http://localhost:11434
OLLAMA_MODEL=llama3.2

# Hugging Face (Fallback)
HUGGINGFACE_API_KEY=your_key_here
HUGGINGFACE_MODEL=microsoft/BioGPT

# Flask
FLASK_PORT=5000
FLASK_DEBUG=True

Getting a Free Hugging Face API Key

1. Go to https://huggingface.co/join
2. Create a free account
3. Navigate to Settings → Access Tokens
4. Create a new token
5. Add it to your .env file

🎯 Key Components

Medical Agent

The core AI agent (agents/medical_agent.py) provides:

• Case analysis with multi-step reasoning
• Symptom extraction and categorization
• Differential diagnosis generation
• Treatment suggestions
• Interactive chat capabilities

LLM Provider

Abstraction layer (agents/llm_provider.py) supporting:

• Ollama - Local, unlimited, free inference
• Hugging Face - Cloud-based fallback
• Automatic failover between providers

Knowledge Base

Medical knowledge system (knowledge/medical_kb.py) with:

• Disease-symptom mappings
• Red flag detection
• Symptom categorization

🧪 Testing

Run the application with sample cases from examples/sample_cases.md:

# Web interface
python app.py

# CLI
python cli.py interactive

🛠️ Troubleshooting

Ollama Connection Issues

# Check if Ollama is running
curl http://localhost:11434/api/tags

# Start Ollama
ollama serve

# Pull a model if not already done
ollama pull llama3.2

Hugging Face API Issues

• Verify your API key in .env
• Check rate limits (free tier has limits)
• Ensure model name is correct

Port Already in Use

# Change port in .env
FLASK_PORT=5001

📝 Sample Cases

See examples/sample_cases.md for example medical cases to test with.

🤝 Contributing

This is an educational project. Feel free to:

• Report issues
• Suggest improvements
• Add more medical knowledge to the knowledge base
• Improve prompts for better analysis

Gmail- sharangs08@gmail.com

📄 License

This project is for educational purposes. Please ensure compliance with medical regulations in your jurisdiction.

🔗 Resources

• Ollama Documentation
• Hugging Face API
• Flask Documentation

---

Remember: This tool is for educational purposes only. Always consult with qualified healthcare professionals for medical advice.

---

Made with love 💚 from Sharan G S