Pre-trained model to predict PD-L1 status based on IHC scans from H&E images, as described in the paper "Deep Learning Based Image Analysis Predicts PDL1 Status From H&E-Stained Histopathology Images in Breast Cancer".
The supplied code infers the PD-L1 status from JPEG images of Hematoxylin and Eosin TMA scans.
The status is given by a probability score in range [0,1].
Examples for Hematoxylin and Eosin images are available under data_examples.
Usage:
predict_on_folder.py [arguments] [options]
Arguments:
<images_root_dir> The path to the directory containing the .jpg images for inference.
Options:
--model_path The path to the pre-trained model for inference (a '.pt' file).
The default path points to the model that was trained and evaluated as descrived
in the published paper.
--output_file A file path for saving the outut results. If no file is given, the results will
be printed in the terminal, but won't be saved to a file.
Run with default parameters:
$ python3 predict_on_folder.py data_examples
Saving results to an output file:
$ python3 predict_on_folder.py data_examples --output_file results.txt
Predict on a different directory:
$ python3 predict_on_folder.py <path to a directory> --output_file results.txt
Python 3.7 and above.
opencv_python==4.5.5.64
torch==1.10.2+cu113
torchvision==0.11.3+cu113
tqdm==4.64.0
To install the requirements, use:
$ pip3 install -r requirements.txt
Or:
$ bash install_pytorch_env.sh