End-To-End Automated Mean Linear Intercept Measurement System

This repository hosts the implementation and resources for the automated Mean Linear Intercept (MLI) measurement system. The system is designed to calculate the MLI from histopathological images of lung tissue using a deep learning-based approach for semantic segmentation.

Features

• Automated MLI Measurement:

  • Supports both direct and indirect methods.
  • Accurately segments lung structures such as alveoli, alveolar septa, blood vessels, and bronchi using a multiresolution convolutional neural network (CNN).
  • Evaluates thousands of field-of-view (FOV) images for robust MLI scoring.

• Graphical User Interface (GUI):

  • Intuitive and user-friendly interface for parameter adjustment and batch processing.
  • Real-time feedback on progress and results.

• Customizable Parameters:

  • FOV size and stride.
  • Guideline length in µm.
  • Save FOVs and MLI method options.

System Workflow

1. Image Preprocessing: Normalizes color variations to ensure consistent segmentation.
2. Semantic Segmentation: Identifies lung structures using a trained CNN model.
3. Field-of-View Extraction: Generates FOV images from WSIs using a sliding window approach.
4. Field-of-View Screening: Accepts or rejects FOVs based on specified criteria.
5. MLI Measurement: Calculates MLI using direct and/or indirect method.

Requirements

The system uses avalible GPU for the semantic segmentation and CPU to compute the MLI score.

• Anaconda Framework: To manage the Python environment and dependencies.
• Python: Version 3.9.0
• CUDA: Version 11.2
• cuDNN: Version 8.1
• TensorFlow: Version 2.10
• Required Python Packages: Listed in requirements.txt

Installation and Usage

1. Clone the repository:

   git clone https://github.com/atallahmadi/e2emli.git
   cd e2emli

2. Download the pre-trained models from:

   • Link 1
   • Link 2

3. Set up the Python 3.9 environment:

   conda create --name MLI Python==3.9
   conda activate MLI

4. Install CUDA Toolkit and cuDNN:

   conda install -c conda-forge cudatoolkit=11.2 cudnn=8.1.0

5. Install required dependencies:

   pip install -r requirements.txt

6. Launch the GUI:

   python app.py

Citation

A. Madi and A. . Chan, “End-To-End Automated Mean Linear Intercept Measurement System”, CMBES Proc., vol. 47, no. 1, May 2025.

• Article

Acknowledgements

This project was supported by the Natural Sciences and Engineering Research Council of Canada (NSERC) and clinical collaborators at the University of Ottawa and the Ottawa Hospital Research Institute.

Contact

For questions or collaboration inquiries, please contact:

• Atallah Madi - [email protected]
• Dr. Adrian D.C. Chan - [email protected]