Skip to content

Latest commit

 

History

History
104 lines (75 loc) · 4.14 KB

README.md

File metadata and controls

104 lines (75 loc) · 4.14 KB
Raw

An explainability framework for cortical surface-based deep learning

This repository contains all source code necessary to replicate our recent work entitled "An explainability framework for cortical surface-based deep learning" available on arXiv. Note that, this repo is a modified version of deepRetinotopy.

You can also check out our notebook available on Open In Colab, in which you can test our perturbation-based approach and visualize some of the figures in our manuscript.

Table of Contents

Installation and requirements

Models were generated using Pytorch Geometric. Since this package is under constant updates, we highly recommend that you follow the following steps to run our models locally:

  • Create a conda environment (or docker container)
  • Install torch first:
	conda install pytorch==1.6.0 torchvision==0.7.0 cudatoolkit=10.1 -c pytorch
  • Install torch-scatter, torch-sparse, torch-cluster, torch-spline-conv and torch-geometric:
    pip install --no-index torch-scatter -f https://pytorch-geometric.com/whl/torch-1.6.0+cu101.html
    pip install --no-index torch-sparse -f https://pytorch-geometric.com/whl/torch-1.6.0+cu101.html
    pip install --no-index torch-cluster -f https://pytorch-geometric.com/whl/torch-1.6.0+cu101.html
    pip install --no-index torch-spline-conv -f https://pytorch-geometric.com/whl/torch-1.6.0+cu101.html
    pip install torch-geometric==1.6.3

Note, there are installations for different CUDA versions. For more: PyTorch Geometric Installation

  • Install the remaining required packages that are available at requirements.txt:
    pip install -r requirements.txt
  • Install the following git repository for plots:
    pip install git+https://github.com/felenitaribeiro/nilearn.git

Explainability

This folder contains functions for the occlusion of input features within a target vertex's neighborhood.

Manuscript

This folder contains all source code necessary to reproduce all figures and summary statistics in our manuscript.

Models

This folder contains all source code necessary to train a new model and to generate predictions on the test dataset using our pre-trained models. Note, models were updated for PyTorch 1.6.0.

Retinotopy

This folder contains all source code necessary to replicate datasets generation, in addition to functions and labels used for figures and models' evaluation.

Citation

Please cite our papers if you used our model or if it was somewhat helpful for you 😉

@article{Ribeiro2022,
	author = {Ribeiro, Fernanda L and Bollmann, Steffen and Cunnington, Ross and Puckett, Alexander M},
	arxivId = {2203.08312},
	journal = {arXiv},
	keywords = {Geometric deep learning, high-resolution fMRI, vision, retinotopy, explainable AI},
	title = {{An explainability framework for cortical surface-based deep learning}},
	url = {https://arxiv.org/abs/2203.08312},
	year = {2022}
}


@article{Ribeiro2021,
	author = {Ribeiro, Fernanda L and Bollmann, Steffen and Puckett, Alexander M},
	doi = {https://doi.org/10.1016/j.neuroimage.2021.118624},
	issn = {1053-8119},
	journal = {NeuroImage},
	keywords = {cortical surface, high-resolution fMRI, machine learning, manifold, visual hierarchy,Vision},
	pages = {118624},
	title = {{Predicting the retinotopic organization of human visual cortex from anatomy using geometric deep learning}},
	url = {https://www.sciencedirect.com/science/article/pii/S1053811921008971},
	year = {2021}
}

Contact

Fernanda Ribeiro <fernanda.ribeiro@uq.edu.au>