EPIC-Survival

This repository provides training and testing scripts for the article EPIC-Survival: End-to-end Part Inferred Clustering for Survival Analysis with Prognostic Stratification Boosting by Muhammad and Xie et al. 2021.

How to Use

The main training script is train.py. Please use python train.py --help to see complete set of training parameters and their descriptions.

The tile library (csv) should have the following format:

SlideID	x	y	Duration	Event	Split
1	24521	23426	16.2	0	'train'
1	6732	3323	16.2	0	'train'
1	1232	5551	16.2	0	'train'
...	...	...	...	...
324	34265	122	3.0	1	'val'
...	...	...	...	...
556	2264	2436	174.0	1	'test'

In short, this tile library should be a record of all tile coordinates with associated slide level information (duration, event, training split, slide name).

The dataloader will load all .svs images located at args.slide_path during initiation, and pull tiles on-the-fly using the (x,y) coordinates during training.

The following will be generated in the output folder:

• convergence.csv
  • a file containing training loss, training concordance index, and validation condorance index over training epochs
• /clustering_grid_top
  • a folder where a clustering visualization for top 20 tiles of each cluster is displayed and saved as a .png

Python Dependencies

• torch 1.8.1
  • torchvision 0.9.1
• lifelines 0.23.8
• openslide 1.1.1
  • Note: We recommend modifying openslide to correct for memory leak issue. Please see openslide/openslide-python#24 for more information.

License

This project is under the CC-BY-NC 4.0 license. See LICENSE for details. (c) MSK

Cite

If you find our work useful, please consider citing our EPIC-Survival Paper:

@inproceedings{muhammad2021epic,
  title={EPIC-Survival: End-to-end Part Inferred Clustering for Survival Analysis, with Prognostic Stratification Boosting},
  author={Muhammad, Hassan and Xie, Chensu and Sigel, Carlie S and Doukas, Michael and Alpert, Lindsay and Simpson, Amber Lea and Fuchs, Thomas J},
  booktitle={Medical Imaging with Deep Learning},
  year={2021}
}