GAT-LI is a repository of codes and experiment results for Research on graph neural networks Approaches in Resting-state Functional Connectivity Analysis, using Keras (version 2.2.4) with TensorFlow (version 1.12.0) as backend.

Our research collection on artificial intelligence for brain image analysis is available on this link.

See the following publications for examples of this code in use:

• Jinlong Hu, Lijie Cao, Tenghui Li, Shoubin Dong & Ping Li, GAT-LI: a graph attention network based learning and interpreting method for functional brain network classification. BMC Bioinformatics, Volume 22, 379 (2021). paper

gat-nr_exp.py is the code of GAT2 model.
gat-average.py is the code of GAT-average model.
gat-fc.py is the code of GAT-fc model.
hack_model.py is the code for hacking model.
ABIDEParser-ref.py is the reference code for using abide data to establish connectivity matrices.