LGAI-EXAONE/EXAONEPath-CRC-MSI-Predictor

Description

MSI classification of CRC tumors using EXAONEPath 1.0.0 - a patch-level foundation model for pathology

Model Overview

This model serves as a reference for predicting MSI status using CRC (colorectal cancer) tumor images as input. When the model receives an H&E-stained whole slide image as input, it removes artifacts observed in the image and extracts only tissue-related objects. These objects are then reconstructed into a set of tiles with a size of 256 by 256 pixels at an mpp (micron per pixel) of 0.5.

The tiles pass through the EXAONEPath v1.0 patch-level foundation model (https://huggingface.co/LGAI-EXAONE/EXAONEPath), which converts them into a set of features. These features are then integrated into a slide-level feature representation through an aggregator(see the figure below). Finally, a linear classifier predicts the MSI status (MSS or MSI-H/L).

The model achieves an average performance of AUROC 0.93 on TCGA-COAD + TCGA-READ data and 0.84 on in-house data.

This open-source release aims to demonstrate that the combination of EXAONEPath and the aggregator can effectively perform pathological tasks. It is hoped that this source code will serve as an important reference not only for CRC MSI prediction but also as an image-based solution for various disease-related problems, including molecular subtyping, tumor subtyping, and mutation prediction.

This open-source code is designed to be compatible with the MONAI framework (https://monai.io/) and all users have to understand how MONAI works to use the codes appropriately.

Also, all users must check the accompanying license before use.

Input and Output Formats

The input is any whole slide images (WSIs) having filename extention such as .svs, .rmxs, .tif and etc. The output is an array with probabilities for each of the two classes (MSS / MSI(H/L)).

Setup

pip install -r requirements.txt

Execute Inference

The inference can be executed as follows

1. Copy your WSI files into samples directory
2. Run inference

   python -m monai.bundle run inference --meta_file configs/metadata.json --config_file configs/inference.yaml
   

Contact

LG AI Research Technical Support: contact_us1@lgresearch.ai

License

Copyright (c) LG AI Research

The model is licensed under EXAONEPath AI Model License Agreement 1.0 - NC.