ORCID Identifier(s)

0009-0008-3747-3115

Graduation Semester and Year

2023

Language

English

Document Type

Thesis

Degree Name

Master of Science in Computer Science

Department

Computer Science and Engineering

First Advisor

Manfred Huber

Abstract

Autism Spectrum Disorder (ASD) affects the patient’s cognitive development which leads to difficulties in social functioning, daily tasks, and independent living. This necessitates intervention at an early age to take preventive measures and provide vital care. Manual diagnosis methods like Autism Diagnostic Observation Schedule (ADOS) assessment adopts symptom-based criteria which typically manifest at a later age. To automate this process, correlations computed from BOLD (Blood Oxygen-level dependent) signals obtained through resting state functional magnetic resonance imaging (rs-fMRI) data of patients across sparse brain regions has been used recently as a measure of functional connectivity. The goal of this study is to investigate the effect of temporal patterns in rs-fMRI time-series data through functional connectivity for automated identification of ASD using a worldwide multisite dataset called Autism Brain Imaging Data Exchange (ABIDE). Our suggested 2-stage network consisting of i) Ensemble Convolutional Neural Network (CNN) for feature extraction from correlation matrices of multiple shorter windows of rs-fMRI time-series and ii) Temporal Convolutional Network (TCN) for classification on the same data after integrating the temporal dimension, has shown improvements in identification of ASD versus typical controls. Examining the rs-fMRI time-series functional connectivity in segments has shown higher gain in classification suggesting presence of relevant features in smaller segments. Due to the limited availability of functional neuro-imaging data for examination using a deep learning architecture, our study also demonstrates tackling the overfitting problem using noise injection and data augmentation.

Keywords

Autism spectrum disorder, Deep learning

Disciplines

Computer Sciences | Physical Sciences and Mathematics

Comments

Degree granted by The University of Texas at Arlington

31970-2.zip (6770 kB)

Share

COinS
 
 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.