CharlieCheckpt/msintuit

MSIntuit - Chowder implementation

This is an official implementation of Chowder model used in MSIntuit CRC^TM. MSIntuit is a tool for pre-screening of microsatellite instability (MSI) using only an H&E slide of a patient with colorectal cancer (CRC).

The model shared here is a variant of Chowder model, originally published by Courtiol et al., 2018: Classification and Disease Localization in Histopathology Using Only Global Labels: A Weakly-Supervised Approach.

Install

Python: 3.7+

pip install .

How to use

import torch
from msintuit.chowder import MSIntuitChowder

dummy_features = torch.ones((16, 100, 2048))  # (batch_size, n_tiles, dimension)
model = MSIntuitChowder(in_features=2048, n_extreme=10)
logits, extreme_scores = model(dummy_features)