This repository contains the code for the paper ProMIL: Probabilistic Multiple Instance Learning for Medical Imaging. The paper was accepted at the conference ECAI 2023.
Before using the code prepare the data following the instructions from https://github.com/apardyl/ProtoMIL
We use the following packages:
- python=3.10
- pytorch 2.0
Run specific main file responsible for a given dataset. The model assumes that the representations for patches from big histology datasets (e.g. TCGA NSCLC) are already extracted using pretrained network.
If you find this work useful, please cite our paper:
@article{struski2023promil,
title={ProMIL: Probabilistic Multiple Instance Learning for Medical Imaging},
author={Struski, {\L}ukasz and Rymarczyk, Dawid and Lewicki, Arkadiusz and Sabiniewicz, Robert and Tabor, Jacek and Zieli{\'n}ski, Bartosz},
journal={arXiv preprint arXiv:2306.10535},
year={2023}
}
This research was partially funded by the National Science Centre, Poland, grants no. 2020/39/D/ST6/01332 (work by Łukasz Struski), 2022/47/B/ST6/03397 (work by Bartosz Zieli ́nski), and 2022/45/N/ST6/04147 (work by Dawid Rymarczyk). Moreover, Dawid Rymarczyk received an incentive scholarship from the funds of the program Excellence Initiative – Research University at the Jagiellonian University in Kraków. Works of Jacek Tabor, Arkadiusz Lewicki, and Robert Sabiniewicz were supported by company the UES Ltd. Some experiments were performed on servers purchased with funds from a grant from the Priority Research Area (Artificial Intelligence Computing Center Core Facility) under the Strategic Programme Excellence Initiative at Jagiellonian University.