IEF-MIL

Instance and Embedding Fused Multiple Instance Learning

Table of Contents

• Overview
• Installation
• Whole Slide Image Preprocessing
  • Segmentation and Patching
  • Feature Extraction
• Training
• Testing
• Multi-Scale Inference

Acknowledgements

This repository is sourced from CLAM. The CLAM repository provided foundational work and inspiration for the development of this project. We thank the authors for their contributions to the field.

Overview

This project implements Instance and Embedding Fused Multiple Instance Learning (IEF-MIL) for whole slide image analysis. The primary functionalities include segmentation, patching, and feature extraction.

Installation

1. Clone the repository:

   git clone https://github.com/username/repo-name.git
   

2. Install the required dependencies:

   pip install -r requirements.txt
   

Whole Slide Image Preprocessing

1. Segmentation and Patching:

   python create_patches_fp.py --source path/to/data_dir --save_dir path/to/save_dir --patch_size 224 --seg --patch --stitch
   

--source: Path to the directory containing the whole slide images. --save_dir: Directory where the generated patches will be saved. --patch_size: Size of the patches to be created (e.g., 224). --seg: Optional flag for segmentation. --patch: Optional flag to create patches. --stitch: Optional flag to stitch the patches back together.

2. Feature Extraction

   python extract_features_fp.py --data_h5_dir path/to/features_dir --data_slide_dir path/to/data_dir --csv_path path/to/features_dir/process_list_autogen.csv --feat_dir path/to/features_dir --batch_size 224 --slide_ext .ndpi
   

--data_h5_dir: Directory where the extracted features will be saved. --data_slide_dir: Directory containing the whole slide images. --csv_path: Path to the CSV file listing the images for feature extraction. --feat_dir: Directory for storing the extracted features. --batch_size: Number of images to process in each batch (e.g., 224). --slide_ext: File extension of the slide images (e.g., .ndpi).

Training

Detailed instructions for training models will be provided here. Currently, refer to the scripts and configuration files for information on training procedures.

Testing

Information on how to test models and evaluate performance will be added here. Please check the relevant scripts and configuration files for testing guidelines. Used metrics: AUC-ROC, AUC-PR, per class PR curves, F1 score and accuracy

Multi-Scale Inference

python create_heatmaps.py --config config_template.yaml