Graph Theory Based Classification of Brain Connectivity Network for Autism Spectrum Disorder

Abstract

Connections in the human brain can be examined efficiently using brain imaging techniques such as Diffusion Tensor Imaging (DTI), Resting-State fMRI. Brain connectivity networks are constructed by using image processing and statistical methods, these networks explain how brain regions interact with each other. Brain networks can be used to train machine learning models that can help the diagnosis of neurological disorders. In this study, two types (DTI, fMRI) of brain connectivity networks are examined to retrieve graph theory based knowledge and feature vectors of samples. The classification model is developed by integrating three machine learning algorithms with a naïve voting scheme. The evaluation of the proposed model is performed on the brain connectivity samples of patients with Autism Spectrum Disorder. When the classification model is compared with another state-of-the-art study, it is seen that the proposed method outperforms the other one. Thus, graph-based measures computed on brain connectivity networks might help to improve diagnostic capability of in-silico methods. This study introduces a graph theory based classification model for diagnostic purposes that can be easily adapted for different neurological diseases.