DRDtechAI - Full-Stack Diabetic Retinopathy Detection System.

An end-to-end AI-powered Medical Detection Platform designed for automated Diabetic Retinopathy (DR) grading, Grad-CAM explainability heatmaps, and retinal blood vessel segmentation from fundus photographs.

---

🌟 System Overview

The DRDtechAI ecosystem consists of three integrated microservices:

graph TD
    User["👨‍⚕️ Clinician / User"] -->|Interacts with| Frontend["🌐 Frontend (React 19 + Vite)"]
    Frontend -->|REST API & Auth| Backend["⚙️ Backend (Node.js + Express + TypeScript)"]
    Backend -->|Stores Scans & Data| DB[("🗄️ MongoDB")]
    Backend -->|Uploads Images| Cloudinary["☁️ Cloudinary CDN"]
    Backend -->|Inference Request| AI["🧠 DR_AI_SERVER (FastAPI + TensorFlow)"]
    AI -->|Returns DR Stage, Grad-CAM & Vessel Masks| Backend
    Backend -->|Aggregated Report| Frontend

Loading

---

1. Frontend: Modern, responsive React 19 web application for patient intake, scan upload, and Detection report visualization.
2. Backend: Node.js / Express / TypeScript API server handling user authentication, patient records, Cloudinary storage, and ML workflow orchestration.
3. ML: Python / FastAPI deep learning microservice hosting fine-tuned ResNet-101 (APTOS dataset) and U-Net (DRIVE dataset) models.

---

🚀 Key Features

• Automated Diabetic Retinopathy Classification: 5-stage clinical severity grading (No DR, Mild, Moderate, Severe, Proliferative DR) using fine-tuned ResNet-101.
• Explainable AI (Grad-CAM Heatmaps): Visualizes the anatomical regions in retinal fundus images that influenced the Detection decision.
• Retinal Blood Vessel Segmentation: High-precision vascular tree extraction using a deep U-Net architecture.
• Secure Authentication & Patient Records: End-to-end patient management with JWT-based security and MongoDB database.
• Cloud Image Pipeline: Automated retinal image upload, optimization, and storage with Cloudinary.
• Comprehensive Detection Reports: Interactive visual report comparing original fundus scans, Grad-CAM overlays, and vessel segmentation masks with clinical action items.

---

📋 Prerequisites

Before setting up the project, make sure you have installed:

• Git
• Node.js (v18.0.0 or higher with npm)
• Python (v3.10 or v3.11 with pip)
• MongoDB (Local instance or MongoDB Atlas)
• Cloudinary Account (For image hosting credentials)

---

💻 How To Use & Quickstart

To clone and run the full stack application, follow the step-by-step instructions below.

1. Clone the Repository

# Clone this repository
$ git clone https://github.com/manojsargam-dev/DRDtech-AI.git
# Wait little time for downloading model

# Navigate into the project root
$ cd "DRDtechAI"

---

2. Start the AI Server (DR_AI_SERVER)

# Go into the AI server directory
$ cd ML

# Create and activate virtual environment
$ python -m venv venv

# Windows (PowerShell):
$ .\venv\Scripts\Activate.ps1
# Windows (cmd):
$ venv\Scripts\activate
# Linux / macOS / Bash:
$ source venv/bin/activate

# Install Python dependencies
$ pip install -r requirements.txt.utf8

# Run the AI server (starts on http://localhost:8000)
$ uvicorn app.main:app --reload --host 0.0.0.0 --port 8000

---

3. Start the Backend API (Backend)

Open a new terminal window:

# Go into the Backend directory
$ cd "DRDtechAI/Backend"

# Install Node dependencies
$ npm install

# Copy environment variables and fill your credentials
$ cp .env.example .env

# Run the backend development server (starts on http://localhost:3000)
$ npm run dev

---

4. Start the Frontend Application (Frontend)

Open a third terminal window:

# Go into the Frontend directory
$ cd "DRDtechAI/Frontend"

# Install dependencies
$ npm install

# Copy environment variables
$ cp .env.example .env

# Run the Vite development server (starts on http://localhost:5173)
$ npm run dev

Open http://localhost:5173 in your browser to access the application.

---

⚙️ Environment Configuration

Important

Populate your .env keys with their respective values in both Backend/ and Frontend/ directories before starting the services.

Backend/.env

PORT=3000
MONGO_URI=your_mongodb_cluster_uri
SECRET=your_jwt_secret_key
EXPIRES=7d
CLOUDINARY_CLOUD_NAME=your_cloudinary_cloud_name
CLOUDINARY_API_KEY=your_cloudinary_api_key
CLOUDINARY_API_SECRET=your_cloudinary_api_secret
CLIENT_URL=http://localhost:5173
ML_SERVICE_URL=http://localhost:8000

Frontend/.env

VITE_API_URL=http://localhost:3000
VITE_ML_API_URL=http://localhost:8000

Note

If you are using Linux Bash for Windows (WSL), make sure MongoDB is running via sudo service mongodb start or connect using your remote MongoDB Atlas URI.