Using Brain Topological Features Extracted from Resting State fMRI to Classify Autism Spectrum Disorder
Date
2019-07-03
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Autism Spectrum Disorder is a neurodevelopmental disease manifesting in early childhood and hindering the social and behavioral outlooks of individuals suffering from it. Early identification of this disorder leads to better patient outcome. Many imaging studies have been conducted in order to gather insight into the inner workings of this disorder with some using machine learning in autism diagnosis. The success of this approach is heavily dependent on the features that are used for the classification task. Graph theoretical measures, extracted from resting state functional MRI, have already proven useful in classifying other neurological disorders. I hypothesized that by using these features for Autism Spectrum Disorder classification, the model performance (accuracy) will improve over previously reported imaging-based methods. Furthermore, this allowed me to identify possible biomarkers for the disorder based on the importance of features selected. This thesis shows that graph theoretical features may help improve classification accuracies and extracting biomarkers relevant to ASD.
Description
Keywords
Autism Spectrum Disorder, fMRI, Resting State, Machine Learning, SVM, Neural Network, Graph Theory
Citation
Kazeminejad, A. (2019). Using Brain Topological Features Extracted from Resting State fMRI to Classify Autism Spectrum Disorder (Master's thesis, University of Calgary, Calgary, Canada). Retrieved from https://prism.ucalgary.ca.