MediCure-AI: Advanced Medical Image Analysis System

This project implements a cutting-edge AI-powered medical image analysis system that leverages multiple fine tuned YOLO Nano segmentation model and Google's Gemini AI to detect various medical conditions from radiological images. The system provides real-time analysis, detailed reports, and professional medical insights for liver, brain, eye, kidney, and lung conditions.

Introduction

MediCure-AI revolutionizes medical imaging with AI-powered diagnostic precision. Our system combines state-of-the-art fine tuned YOLO Nano segmentation model with Google's Gemini AI to provide comprehensive analysis of medical scans including CT, MRI, X-ray, and ultrasound images. The platform is designed to assist healthcare professionals in making faster, more accurate diagnoses while generating detailed PDF reports for clinical documentation.

Features

Multi-Modal Medical Analysis: Support for liver, brain MRI, eye, kidney, and lung imaging
Real-time Detection: Instant analysis with confidence scoring and bounding box visualization
AI-Powered Insights: Integration with Google Gemini AI for professional medical interpretations
Automated Report Generation: PDF reports with findings, confidence scores, and treatment recommendations
Pseudo-Coloring: Advanced visualization techniques for better anomaly identification
Web-Based Interface: User-friendly Flask web application
Batch Processing: Support for multiple model analysis on single images
Secure File Handling: Robust file upload and processing with validation

Supported Medical Conditions

🫀 Liver Analysis

Conditions Detected: Liver tumors, normal liver tissue
Applications: Hepatocellular carcinoma detection, liver health assessment

🧠 Brain MRI Analysis

Conditions Detected: Brain tumors, anatomical structures
Applications: Neurological disorder diagnosis, tumor detection

👁️ Eye Examination

Conditions Detected:
- Cataract
- Diabetic Retinopathy
- Glaucoma
- Normal eye conditions
Applications: Ophthalmological screening, preventive care

🫘 Kidney Analysis

Conditions Detected: Kidney stones
Applications: Urological assessment, stone detection

🫁 Lung Analysis

Conditions Detected: Lung tumors/cancer
Applications: Pulmonary oncology, early cancer detection

System Architecture

Workflow Diagram

graph TD
    A[Upload Medical Image] --> B[File Validation]
    B --> C[Model Selection]
    C --> D{Single Model or Multi-Model?}
    D -->|Single| E[YOLO Model Processing]
    D -->|Multi| F[All Models Processing]
    E --> G[Detection & Visualization]
    F --> G
    G --> H[Pseudo-Coloring Application]
    H --> I[Gemini AI Analysis]
    I --> J[Generate PDF Report]
    J --> K[Return Results]

Model Inventory

graph TD
    L[YOLO Models]
    L --> M[Liver Model - liver_tumour.pt]
    L --> N[Brain Model - brain_mri.pt]
    L --> O[Eye Model - eye.pt]
    L --> P[Kidney Model - kidney.pt]
    L --> Q[Lung Model - lung cancer.pt]

Key Components

Image Upload & Validation: Secure file handling with format verification
Multi-Model Detection: Parallel processing across specialized YOLO models
Visualization Engine: Bounding box rendering with pseudo-coloring for anomalies
AI Analysis: Google Gemini integration for medical insights
Report Generation: Automated PDF creation with comprehensive findings

Installation

Prerequisites

Python 3.8 or later
pip (Python package installer)
CUDA-compatible GPU (recommended for faster inference)

Step 1: Clone the Repository

git clone https://github.com/kknani24/medicure-ai-.git
cd medicure-ai-

Step 2: Create Virtual Environment

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

Step 3: Install Required Packages

pip install -r requirements.txt

Step 4: Install Additional Dependencies

# Install PyTorch with CUDA support (optional, for GPU acceleration)
pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118

# Install Ultralytics YOLO
pip install ultralytics

# Install additional packages
pip install flask opencv-python google-generativeai reportlab python-dotenv

Configuration

Step 1: Environment Setup

Create a .env file in the project root:

GEMINI_API_KEY=your_google_gemini_api_key_here

Step 2: Model Files

Ensure all YOLO model files are present:

├── liver_tumour.pt
├── brain_mri.pt
├── eye.pt
├── kidney.pt
└── lung cancer.pt

Step 3: Directory Structure

medicure-ai-/
├── app.py
├── .env
├── requirements.txt
├── static/
│   └── uploads/
├── templates/
│   └── index.html
├── models/
│   ├── liver_tumour.pt
│   ├── brain_mri.pt
│   ├── eye.pt
│   ├── kidney.pt
│   └── lung cancer.pt
└── README.md

Usage

Step 1: Start the Application

python app.py

Step 2: Access the Web Interface

Open your browser and navigate to http://localhost:5000

Step 3: Upload and Analyze

Select a medical image (PNG, JPG, JPEG formats supported)
Choose a specific model or leave blank for multi-model analysis
Click "Analyze Image" to process
View results, AI insights, and download the PDF report

Command Line Usage

For programmatic access, you can also use the API endpoints directly:

import requests

# Upload and analyze image
files = {'file': open('medical_scan.jpg', 'rb')}
data = {'model': 'liver'}  # Optional: specify model
response = requests.post('http://localhost:5000/detect', files=files, data=data)
result = response.json()

API Endpoints

POST /detect

Analyzes uploaded medical images and returns detection results.

Parameters:

file (required): Medical image file (PNG, JPG, JPEG)
model (optional): Specific model to use ('liver', 'brain_mri', 'eye', 'kidney', 'lung')

Response:

{
    "message": "Detection completed",
    "original_image": "/static/uploads/scan.jpg",
    "processed_image": "/static/uploads/processed_scan.jpg",
    "detections": {
        "liver": [
            {
                "class": "tumor",
                "confidence": 0.87,
                "bbox": [150, 200, 300, 350]
            }
        ]
    },
    "gemini_response": "AI medical analysis...",
    "pdf_report": "/static/uploads/report_scan.pdf"
}

Technologies Used

YOLO (You Only Look Once)

YOLO is a state-of-the-art real-time object detection system. In this project, we use specialized YOLO models trained on medical datasets for accurate detection of various pathological conditions. The models provide fast inference with high accuracy, making them ideal for clinical applications.

Google Gemini AI

Google's Gemini AI provides advanced multimodal understanding capabilities. In our system, it analyzes medical images alongside detection results to generate professional medical insights, treatment recommendations, and diagnostic interpretations with the expertise of a seasoned clinician.

Flask

Flask serves as the web framework powering our medical analysis platform. Its lightweight nature and extensive ecosystem make it perfect for creating responsive web applications that can handle medical image processing workflows efficiently.

OpenCV

OpenCV handles all computer vision operations including image loading, processing, visualization, and pseudo-coloring. Its optimized algorithms ensure fast image manipulation and enhancement, crucial for real-time medical image analysis.

Code Structure

medicure-ai/
├── app.py                     # Main Flask application
├── static/
│   ├── uploads/              # Uploaded and processed images
│   └── css/                  # Stylesheets
├── templates/
│   └── index.html            # Web interface template
├── models/                   # YOLO model files
├── requirements.txt          # Python dependencies
├── .env                      # Environment variables
└── README.md                # Project documentation

Key Functions

allowed_file(): Validates uploaded file formats
process_image(): Main image processing pipeline
process_multiple_model_results(): Handles multi-model analysis
generate_response_with_image(): Interfaces with Gemini AI
create_pdf_report(): Generates comprehensive PDF reports

Contributing

We welcome contributions to improve MediCure-AI! Here's how you can contribute:

Fork the Repository

git fork https://github.com/kknani24/medicure-ai-.git

Create a Feature Branch

git checkout -b feature/amazing-feature

Make Your Changes

Add new medical models
Improve detection accuracy
Enhance UI/UX
Add new features

Commit Your Changes

git commit -m "Add amazing feature"

Push to the Branch

git push origin feature/amazing-feature

Open a Pull Request

Areas for Contribution

Model Enhancement: Improve existing models or add new medical condition detection
UI/UX Improvements: Enhance the web interface and user experience
Performance Optimization: Optimize inference speed and memory usage
Documentation: Improve documentation and add tutorials
Testing: Add comprehensive test coverage

License

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

Disclaimer

⚠️ Important Medical Disclaimer: This AI system is designed to assist healthcare professionals and should not be used as a substitute for professional medical diagnosis or treatment. Always consult with qualified healthcare providers for medical decisions. The system is intended for educational and research purposes and should undergo proper clinical validation before use in medical practice.

For questions, support, or collaboration opportunities, please open an issue or contact the development team.

Name		Name	Last commit message	Last commit date
Latest commit History 32 Commits
README.md		README.md
app.py		app.py
brain_mri.pt		brain_mri.pt
eye.pt		eye.pt
index.html		index.html
kidney.pt		kidney.pt
liver_tumour.pt		liver_tumour.pt
lung cancer.pt		lung cancer.pt
requirements.txt		requirements.txt

kknani24/medicure-ai-

Folders and files

Latest commit

History

Repository files navigation