AI-Based Universal Lesion Segmentation

Overview

This repository contains the code for the AI-Based Universal Lesion Segmentation application, based on my master's thesis titled AI-Based Universal Lesion Segmentation Application for Thoracic-Abdominal in Computed Tomography Scans. The project focuses on developing an advanced deep learning framework to automate lesion segmentation in CT scans, aiming to improve diagnostic accuracy and workflow efficiency.

The model leverages state-of-the-art techniques, including:

• U-Mamba Architecture: An encoder-decoder network with skip connections optimized for capturing fine details in CT scans.
• LION Optimizer: An advanced optimization strategy to enhance training performance.

The deployment website is built using Vite + React to provide a modern, responsive interface for users to interact with the segmentation model.

Thesis Summary

The thesis presents:

• A novel deep learning framework for universal lesion segmentation targeting thoracic-abdominal CT scans.
• Integration of meticulous preprocessing (including VOI cropping), a robust segmentation model, and effective post-processing methods.
• Comprehensive experimental evaluation showcasing the model's accuracy and potential clinical impact.
• A demonstration platform (this website) to visualize segmentation results and interact with the model.

Model Deployment

The web application serves as an interactive platform where users can:

• Upload CT Scans: Submit images for segmentation.
• View Segmentation Results: See the segmented lesions overlaid on the original scans.
• Monitor Model Performance: Access real-time results and intermediate outputs for diagnostic insights.

The deployment leverages a modern stack (Vite + React) for a fast and responsive user experience.

Screenshots

Main Dashboard

This screenshot shows the main dashboard of the web application. Users can upload CT scans and quickly access segmentation outputs through an intuitive interface.

Processing and Results

This view highlights the real-time segmentation process, displaying how the input scans are processed and segmented.

Detailed Segmentation Output

This view provides a detailed look at the segmentation results, with overlays on the CT scans clearly showing the identified lesions.

Installation

1. Clone Repository and Install Node.js Dependencies

1. Clone the repository:

   git clone https://github.com/yourusername/your-repo.git
   cd your-repo

2. Install Node.js dependencies:

   npm install

2. Set Up and Run the Flask Server

1. Activate the Python Environment:

   • Navigate to the Flask server's virtual environment directory:

     cd flask-server/server_flask/Scripts

   • For Windows, activate the environment by running:

     activate

   • For Unix-based systems, run:

     source activate

2. Run the Flask Server:

   • Navigate back to the root directory (if needed):

     cd ../../

   • Start the Flask server by executing:

     python flask-server/server.py

3. Run the Web Application

• In a separate terminal window, run the development server:

  npm run dev

4. Build for Production

• To build the project for production, run:

  npm run build

Usage

After installation, you can use the application to:

• Choose CT scan images from the examples or Upload your own for lesion segmentation.
• Interact with the web interface for further analysis and diagnostics.
• View real-time segmentation results and explore model performance.

For more detailed usage instructions, please refer to the in-app documentation.

Contributing

Contributions are welcome! Please see the CONTRIBUTING.md file for guidelines on how to contribute to the project.

License

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

Acknowledgements

• Supervisor: Dr. BOUALLEG Yaakoub for invaluable guidance throughout this research.
• Committee Members: For their insightful feedback and support.
• Family & Friends: For their constant encouragement and support.