Abstract
Finding the biomarkers associated with ASD is helpful for understanding theunderlying roots of the disorder and can lead to earlier diagnosis and moretargeted treatment. A promising approach to identify biomarkers is using GraphNeural Networks (GNNs), which can be used to analyze graph structured data,i.e. brain networks constructed by fMRI. One way to interpret importantfeatures is through looking at how the classification probability changes ifthe features are occluded or replaced. The major limitation of this approach isthat replacing values may change the distribution of the data and lead toserious errors. Therefore, we develop a 2-stage pipeline to eliminate the needto replace features for reliable biomarker interpretation. Specifically, wepropose an inductive GNN to embed the graphs containing different properties oftask-fMRI for identifying ASD and then discover the brain regions/sub-graphsused as evidence for the GNN classifier. We first show GNN can achieve highaccuracy in identifying ASD. Next, we calculate the feature importance scoresusing GNN and compare the interpretation ability with Random Forest. Finally,we run with different atlases and parameters, proving the robustness of theproposed method. The detected biomarkers reveal their association with socialbehaviors. We also show the potential of discovering new informativebiomarkers. Our pipeline can be generalized to other graph feature importanceinterpretation problems.1