🏥 Diabetic Retinopathy Detection System

🌐 Live Demo: https://dr.duminduthushan.com

An AI-powered multi-modal system for diabetic retinopathy risk assessment combining clinical data analysis with retinal image processing

📋 Table of Contents

Overview
Features
Architecture
Tech Stack
Installation
Usage
API Documentation
Model Performance
Deployment
Project Structure
Contributing
License
Contact

🔍 Overview

The Diabetic Retinopathy Detection System is a comprehensive healthcare AI solution that combines clinical data analysis with retinal image processing to provide accurate risk assessments for diabetic retinopathy. The system employs ensemble machine learning models and deep learning techniques to achieve high accuracy in early detection.

🎯 Key Objectives

Early Detection: Identify DR risk before symptoms appear
Multi-Modal Analysis: Combine clinical and imaging data for better accuracy
Accessible Healthcare: Provide rapid, automated screening
Clinical Decision Support: Assist healthcare providers with AI-powered insights

✨ Features

🏥 Clinical Analysis

13+ Clinical Parameters: Comprehensive health metrics analysis
Real-time Risk Assessment: Instant prediction results
Ensemble Models: XGBoost + Random Forest for robust predictions
87.8% Accuracy: Validated on PIMA Indian Diabetes Dataset

🔬 Image Analysis

Retinal Image Processing: Advanced preprocessing with CLAHE
Transfer Learning: MobileNetV2 architecture
Multi-class Classification: 5-stage DR severity detection
Real-time Processing: Fast inference for clinical use

🔄 Fusion System

Multi-Modal Integration: Combines clinical and image predictions
Weighted Risk Scoring: Intelligent risk aggregation
Confidence Metrics: Prediction reliability indicators
Feature Importance: Shows key risk factors

💻 User Interface

Responsive Design: Works on desktop, tablet, and mobile
Real-time Validation: Immediate input feedback
Interactive Visualizations: Risk meters and progress bars
Professional Medical UI: Healthcare-grade interface

🏗️ Architecture

graph TB
    subgraph "Frontend - Vercel"
        A[React App] --> B[Form Validation]
        B --> C[API Client]
    end
    
    subgraph "Backend - Contabo VPS"
        D[FastAPI Server] --> E[Model Pipeline]
        E --> F[Clinical Model]
        E --> G[Image Model]
        E --> H[Fusion Model]
    end
    
    subgraph "ML Pipeline"
        I[Data Preprocessing] --> J[Feature Engineering]
        J --> K[Model Inference]
        K --> L[Risk Calculation]
    end
    
    C -->|HTTPS| D
    F --> L
    G --> L
    H --> L
    L --> M[Results]

🛠️ Tech Stack

Backend

Technology	Purpose
	Core programming language
	REST API framework
	Deep learning framework
	Machine learning library
	Image processing
	Numerical computing
	Data manipulation

Frontend

Technology	Purpose
	UI framework
	Programming language
	Styling
	HTTP client

Deployment

Technology	Purpose
	Containerization
	Frontend hosting
	Reverse proxy
	Server OS

🚀 Installation

Prerequisites

Python 3.9+
Node.js 16+
Git

Clone Repository

git clone https://github.com/yourusername/diabetic-retinopathy-ai.git
cd diabetic-retinopathy-ai

Backend Setup

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

# Install dependencies
pip install -r requirements.txt

# Train models (optional - pre-trained models included)
python ml-pipeline/src/train_clinical_model.py
python ml-pipeline/src/train_image_model.py

# Start backend server
python backend/main.py

Frontend Setup

# Navigate to frontend directory
cd frontend

# Install dependencies
npm install

# Start development server
npm start

Docker Setup

# Build and run with Docker Compose
docker-compose up --build

# Access application
# Frontend: http://localhost:3000
# Backend: http://localhost:8000

📊 Model Performance

Clinical Model (XGBoost Ensemble)

Metric	Score
Accuracy	87.8%
Precision	84.5%
Recall	89.2%
F1-Score	86.8%
AUC-ROC	0.921

Image Model (MobileNetV2)

Metric	Score
Accuracy	85.0%*
Precision	82.3%*
Recall	87.6%*
F1-Score	84.9%*
AUC-ROC	0.904*

*Note: Image model currently trained on synthetic data for demonstration. Production deployment would use APTOS 2019 or EyePACS datasets for real clinical accuracy.

Fusion Model

Metric	Score
Combined Accuracy	92.0%
Clinical Weight	0.6
Image Weight	0.4

📁 Project Structure

diabetic-retinopathy-ai/
├── 📂 backend/
│   ├── main.py              # FastAPI application
│   ├── models.py             # Pydantic models
│   ├── utils.py              # Utility functions
│   └── requirements.txt      # Python dependencies
│
├── 📂 frontend/
│   ├── 📂 src/
│   │   ├── App.js           # Main React component
│   │   ├── App.css          # Styles
│   │   └── 📂 components/   # React components
│   ├── package.json         # Node dependencies
│   └── public/              # Static assets
│
├── 📂 ml-pipeline/
│   ├── 📂 src/
│   │   ├── train_clinical_model.py
│   │   ├── train_image_model.py
│   │   └── train_fusion_model.py
│   ├── 📂 models/           # Trained models
│   └── 📂 data/             # Training data
│
├── 📂 docker/
│   ├── Dockerfile
│   ├── docker-compose.yml
│   └── nginx.conf
│
└── README.md

🌐 API Documentation

Base URL

https://api.dr.duminduthushan.com

Endpoints

Health Check

GET /health

Clinical Prediction

POST /predict/clinical
Content-Type: application/json

{
  "glucose": 120,
  "blood_pressure": 70,
  "bmi": 25.5,
  "age": 35,
  ...
}

Combined Prediction

POST /predict/combined
Content-Type: multipart/form-data

Form Data:
- data: JSON clinical data
- image: Retinal image file

Model Information

GET /models/info

🚢 Deployment

The system is deployed using a distributed architecture:

Frontend: Vercel (Global CDN)
Backend: Contabo VPS (4 vCPU, 8GB RAM)
Domain: Cloudflare DNS
SSL: Let's Encrypt

Deployment Commands

# Frontend (Vercel)
vercel --prod

# Backend (VPS)
ssh user@vps-ip
cd /opt/dr-detection
docker-compose up -d

🤝 Contributing

Contributions are welcome! Please follow these steps:

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

⚠️ Disclaimer

This system is designed for educational and research purposes only. It should not be used as a substitute for professional medical diagnosis or treatment. Always consult qualified healthcare providers for medical decisions.

📄 License

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

Name		Name	Last commit message	Last commit date
Latest commit History 75 Commits
backend		backend
docs		docs
frontend		frontend
ml-pipeline/src		ml-pipeline/src
nginx		nginx
scripts		scripts
.dockerignore		.dockerignore
.env.example		.env.example
.gitignore		.gitignore
LICENSE		LICENSE
README.md		README.md
docker-compose.yml		docker-compose.yml
requirements.txt		requirements.txt
test_image_model_performance.py		test_image_model_performance.py

License

codedbydumi/diabetic-retinopathy-ai

Folders and files

Latest commit

History

Repository files navigation