TE-HI-GCN

The implementation of TE-HI-GCN in paper:

Lanting Li et.al "TE-HI-GCN: An Ensemble of Transfer Hierachical Graph Convolutional Networks for Disorder Diagnosis."

code by

Lanting Li

https://doi.org/10.1007/s12021-021-09548-1

Require

Python 3.6

Reproducing Results

For ABIDE Datasets：

mkdir model

cd model

mkdir <fGCN/hiGCN/ehiGCN> (choose a floder name that you need)

cd <fGCN/hiGCN/ehiGCN>

mkdir

1、f-GCN: python train-fGCN.py

2、HI-GCN: python train-hiGCN.py

3、E-HI-GCN: python train-ehiGCN.py

4、TE-HI-GCN:

Train the model of Transfer learning part that you need.

Load the Transfer learning part into the E-HI-GCN model.

Then, python train-ehiGCN.

For ADNI Datasets：

Using the code in the folder: ADNI.

Other operations are consistent with the ABIDE datasets.

Change "name" in these codes when you need to product results of different atlas.

Change "thr" in train-fGCN.py when you need to product results with different thresholds.