Date of Award

Spring 4-13-2023

Embargo Period

4-16-2023

Document Type

Thesis

Degree Name

Master of Biomedical Science

Department

Neuroscience

College

College of Graduate Studies

First Advisor

Jens Jensen

Second Advisor

Jane E. Joseph

Third Advisor

Andreana Benitez

Fourth Advisor

Stefano Berto

Abstract

Major types of psychotic disorders include schizophrenia (SCZ), bipolar disorder (BP) and schizoaffective disorder (SZA). These disorders have profound and overlapping symptoms with marked cognitive deficits, and their diagnosis relies on symptom clusters. The treatments for psychosis are usually focused on positive symptoms such as delusions and hallucinations. Although cognitive impairments underlie both positive and negative symptoms, functional brain imaging biomarkers that can reliably predict a patient's cognitive deficits are still lacking. Therefore, this project used functional MRI to explore the feasibility of using functional connectivity (FC) to predict cognitive performance.

A total of 207 subjects (BP: 79, SZ/SZA: 48, and HC: 80) with high functional MRI image (fMRI) quality (SNR> 100, motion < 0.3) were selected from the McLean MATRICS dataset. Subjects were divided into a discovery cohort (n=104) and an age, gender, and head motion matched validation cohort (n=103). The hypothesis was that FC could predict cognitive performance in the discovery cohort and that the prediction models could be generalized to the validation cohort. The connectomes for each subject were obtained by calculating the whole- brain connectivity using networks from the individualized functional parcellation as region of interests (ROIs). Models were trained to predict the 8 cognitive scores in the discovery cohort, respectively. The generalizability of these models was tested by applying these models to the validation cohort.

The trained models were able to predict 6 out of 8 cognitive scores using a LOOCV procedure. Models for working memory, composite score and attention score could be generalized to the validation cohort. A total of 35 FC features were identified as important for predicting performance in these cognitive domains. Significant differences between patients and controls were found for 13 of these features when considered individually.

In summary, this project has established a framework for biomarker discovery that may have clinical relevance for the diagnosis of psychosis early in the disease process by providing possible FC features that can be detected using fMRI and may help guide therapeutic interventions. The identified biomarkers also provide convergent evidence for network dysfunction in psychosis and suggest personalized treatment targets.

Rights

Copyright is held by the author. All rights reserved.

Download

Share

COinS