This repository holds code to generate counterfactual images of MR brain images for people with (and without) MS using a structural causal model (SCM) [1] built in Pyro.
This code was used to generate the counterfactual images in our paper "A Structural Causal Model of MR Images of Multiple Sclerosis".
Our work builds on the work of Pawlowski, Castro, and Glocker [2]. The code in this repository is a fork of their code which can be found here.
The code for the segmentation experiment in our paper can be found here; it depends on msseg. The exact hyperparameters and information about the data used for the segmentation experiment are in the docstring of that script.
This package was developed by Jacob Reinhold of the Image Analysis and Communication Lab (IACL).
From inside this directory, run:
python setup.py install
or (if you'd like to make updates to the package)
python setup.py develop
This repository contains code and assets structured as follows:
counterfactualms/: contains the code used for running the experimentsarch/: model architectures used in experimentsdatasets/: script for dataset generation and data loading used in experimentsdistributions/: implementations of useful distributions or transformationsexperiments/: implementation of experiments
assets/: contains hyperparameters for the experiments listed in the paper
- Pearl, Judea. Causality. Cambridge university press, 2009.
- Pawlowski, Nick, Daniel Coelho de Castro, and Ben Glocker. "Deep Structural Causal Models for Tractable Counterfactual Inference." Advances in Neural Information Processing Systems 33 (2020).
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent