Detection of Autism Spectrum Disorder Using Graph Representation Learning Algorithms and Deep Neural Network, Based on fMRI Signals

Abstract

In this paper, we are going to apply graph representation learning algorithms to identify autism spectrum disorder (ASD) patients within a large brain imaging dataset. Since ASD is characterized by social deficits and repetitive behavioral symptoms, it is mainly identified by brain functional connectivity patterns. Attempts to unveil the neural patterns that emerged from ASD are the essence of ASD classification. We claim that considering the connectivity patterns of the brain can be appropriately executed by graph representation learning methods. These methods can capture the whole structure of the brain, both local and global properties. The investigation is done for the brain imaging worldwide multi-site database known as ABIDE (Autism Brain Imaging Data Exchange). The classifier adapted to the features embedded in graphs is a LeNet deep neural network. Among different graph representation techniques, we used AWE, Node2vec, Struct2vec, multi node2vec, and Graph2Img. The best approach was Graph2Img, in which after extracting the feature vectors representative of the brain nodes, the PCA algorithm is applied to the matrix of feature vectors. Although we could not outperform the previous 70% accuracy of 10-fold cross-validation in the identification of ASD versus control patients in the dataset, for leave-one-site-out cross-validation, we could obtain better results (our accuracy: 80%). It is evident that the effect of graph embedding methods is making the connectivity matrix more suitable for applying to a deep network.