DocGPT (Doctor GPT) - AI-Powered Medical Diagnosis System

Ahsan Umar

ML Engineer

Software Engineer

AI Developer

Python

PyTorch

scikit-learn

DocGPT (Doctor GPT) - AI-Powered Medical Diagnosis System

DocGPT (Doctor GPT) is an advanced medical diagnosis system that combines Vision Transformer (ViT) based deep learning models with LangChain agents to provide comprehensive medical image analysis and detailed diagnostic reports. The system leverages the power of PyTorch for deep learning and Groq's LLM for generating human-like medical insights.

Core Technology

LangChain Agents: Intelligent agents that coordinate between different disease detection models and the LLM to provide comprehensive medical analysis

Vision Transformer (ViT): State-of-the-art transformer architecture for medical image analysis

Deep Learning Models: Specialized PyTorch models trained for different medical conditions:

ResNet-based architecture for Eye Disease detection

Vision Transformer for Skin Cancer classification

Custom CNN architecture for Pneumonia detection

Features

Multi-Disease Detection: Supports multiple medical conditions:

Eye Diseases (Cataract, Glaucoma, Diabetic Retinopathy)

Skin Cancer (Melanoma Detection)

Pneumonia (X-ray Analysis)

Brain Tumor (Coming Soon)

Heart Disease (Coming Soon)

See the DocGPT models Code

AI-Powered Analysis:

Deep learning models for accurate disease detection

Groq LLM integration for detailed medical reports

LangChain agents for orchestrating the analysis pipeline

Modern Architecture:

FastAPI backend with automatic OpenAPI documentation

Streamlit frontend for testing (React.js interface planned)

Modular design for easy extension to new disease types

Tech Stack

Backend Framework: FastAPI

Deep Learning:

PyTorch

torchvision

Vision Transformer (ViT)

AI Integration:

LangChain for agent orchestration

Groq API for medical report generation

Image Processing:

PIL

torchvision transforms

Development:

Python 3.12+

pydantic for data validation

uvicorn for ASGI server

Installation

Clone the repository:

git clone https://github.com/codewithdark-git/DocGPT.git
cd DocGPT

Create and activate a virtual environment:

python -m venv venv
source venv/bin/activate  # Linux/Mac
# OR
venv\Scripts\activate  # Windows

Install dependencies:

pip install -r requirements.txt

Set up environment variables in .env:

GROQ_API_KEY=your_groq_api_key
GROQ_MODEL_NAME=llama-3.2-11b-vision-preview

Usage

Start the FastAPI backend:

uvicorn app.main:app --reload

Start the Streamlit frontend (in a new terminal):

streamlit run streamlit_app.py

Access the applications:

API Documentation: http://localhost:8000/docs

Streamlit Interface: http://localhost:8501

API Endpoints

Health Check

GET /api/v1/health: Check API health status

Disease Prediction

POST /api/v1/predict: Submit an image for disease prediction

Parameters:

disease_type: Type of disease to predict

file: Image file

Project Structure

DocGPT/
├── app/
│   ├── main.py           # FastAPI application
│   ├── config.py         # Configuration settings
│   ├── routers/          # API routes
│   ├── schemas/          # Pydantic models
│   └── services/         # ML models and business logic
├── models/               # Trained model files
├── streamlit_app.py      # Streamlit frontend
├── requirements.txt      # Project dependencies
└── .env                  # Environment variables

Disease Types

Eye Disease Detection:

Normal

Cataract

Glaucoma

Diabetic Retinopathy

Skin Cancer Detection:

Melanoma

Non-Melanoma

Pneumonia Detection:

Normal

Pneumonia

Contributing

Fork the repository

Create a feature branch

Commit your changes

Push to the branch

Create a Pull Request

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments

Medical datasets providers

PyTorch team

Groq API team

FastAPI and Streamlit communities

Like this project

Posted Feb 4, 2025

DocGPT (Doctor GPT) is an advanced medical diagnosis system that combines Vision Transformer (ViT) based deep learning models with LangChain agents to provide …

Likes

Views