SurvivalHypergraph

The codes of "Stage II Colorectal Cancer Survival Prediction Through Exploring High-Order Information on Whole-Slide Histopathological Images".

Pre-requisites:

• Linux (tested on Ubuntu 20.04)
• python (3.8.3), pytorch (1.10.0), torchvision (0.11.1), numpy (1.21.2), opencv-python (4.5.4), openslide-python (1.1.2), pandas (1.3.4), scikit-learn (1.0), scipy (1.7.1), lifelines (0.26.3), and tqdm (4.62.3)

Installation Guide

1. Install anaconda3 in the machine according to https://www.anaconda.com/

2. Install openslide

   sudo apt-get install openslide-tools

3. Create a conda environment based on the environment configuration file

   conda env create -n survivalhypergraph -f enviroment.yaml

4. Activate the environment

   conda activate survivalhypergraph
   

Preprocess Guide

1. Store whole slide image data in the following format

   SurvivalHypergraph/
   		|-- svs_directory/
   				|-- wsi_1.svs
   				|-- wsi_2.svs
   				|-- ...

2. Run the preprocess code

   python preprocess/process_4_ft.py
   

Run Guide

1. Store patient information in file SurvivalHypergraph/data/opti_survival.json in the following format

   {
   		"patient_1_name":{
   				"status": 0/1,
   				"survival_time": xx (days),
   				"images": [wsi_1, wsi_2, ...]
   		},
   		"patient_2_name":{
   				"status": 0/1,
   				"survival_time": xx (days),
   				"images": [wsi_1, wsi_2, ...]
   		},
   		...
   }

2. Set hyperparameters in train_config.py

3. Run the trainning code

   python train.py