This code provides a python - Tensorflow implementation of graph convolutional networks (GCNs) for semi-supervised disease prediction using population graphs, as described in:
Parisot, S., Ktena, S. I., Ferrante, E., Lee, M., Moreno, R. G., Glocker, B., & Rueckert, D. (2017).
Spectral Graph Convolutions for Population-based Disease Prediction.
arXiv preprint arXiv:1703.03020.
We provide an implementation applied to the ABIDE dataset for diagnosis of Autism Spectrum Disorder (preprocessed-connectomes-project.org/abide).
To run the programme, you will need to install the implementation of graph convolutional networks (GCN) by Kipf et al. This project is only compatible with our forked GCN project.
The root folder in fetch_data.py (line 12) and ABIDEParser.py (line 17) has to be updated to the folder were the data will be stored.
Next, to install, organise and pre-process the ABIDE database: python fetch_data.py
python main_ABIDE.py
python main_ABIDE --help provides a detailed description of parameters
tensorflow (>0.12)
networkx
nilearn
scikit-learn
joblib
Please cite our paper if you use this code in your own work:
@article{parisot2017spectral,
title={Spectral Graph Convolutions on Population Graphs for Disease Prediction},
author={Parisot, Sarah and Ktena, Sofia Ira and Ferrante, Enzo and Lee, Matthew and Moreno, Ricardo Guerrerro and Glocker, Ben and Rueckert, Daniel},
journal={arXiv preprint arXiv:1703.03020},
year={2017}
}