DOI

https://doi.org/10.25772/VXWG-Q330

Defense Date

2019

Document Type

Thesis

Degree Name

Master of Science

Department

Biomedical Engineering

First Advisor

Paul A. Wetzel, PhD

Second Advisor

Mark Baron, MD

Abstract

Purpose: Due to the neurological aspects of Parkinson’s Disease (PD) and the sensitivity of eye movements to neurological issues, eye tracking has the potential to be an objective biomarker with higher accuracy in diagnosis than current clinical standards. Currently when PD is diagnosed clinically, there is an accuracy of 74% when diagnosed by a general practitioner and 82% when diagnosed by a movement disorder specialist. This study was designed to: 1. Assess eye movements as a potential biomarker for Parkinson’s Disease. 2. Determine if eye movements can distinguish between Parkinson’s Disease and commonly confounded movement disorders with parkinsonian symptoms. 3. Determine if the eye movements of Rapid Eye Movement Behavior Disorder (RBD) patients who will likely convert to PD are distinguishable from healthy controls and if RBD patients have eye movements with similar features to PD.

Methods: The eye movements of 160 subjects (43 healthy controls, 63 PD, 31 REM Behavior Disorder, and 22 Other Parkinsonisms) were recorded at 500 Hz and analyzed. Each subject performed five eye tracking tasks that included reflexive saccades, inhibition of reflexive saccades, predictive saccades, and reading. Based on an analysis of selected eye movement measurement parameters, a multivariable logistic regression model was developed that compared: PD vs. Control, PD vs. “Other”, PD vs RBD, and Control vs RBD. The resulting predictive model was then assessed for accuracy, sensitivity, and specificity.

Results: After screening, the most statistically significant predictors that were included in the final multivariate model were: Site, Sex, Age, Age squared, UPDRS Score, mean absolute fixation velocity (Horizontal Step Task), saccadic duration, average saccadic velocity, and mean fixation velocity (Predictive Task). The model predicted with an accuracy of: 92% for Controls, 88% for PD, 86% for RBD, and 68% for Other Parkinsonisms. The model was best at distinguishing between PD and Other Parkinsomisms with an accuracy of 89% and RBD and Controls with an accuracy of 88%.

Conclusion: This research found that specific combinations of eye tracking parameters from simple tasks can be used to distinguish between PD and commonly confounded movement disorders with parkinsonism symptoms. The model’s ability to distinguish between groups indicates that in a confirmatory study we should have relatively high accuracy in discriminating between groups. This model is able to accurately distinguish Controls from RBDs, however due to an insufficient number of follow-up visits to date, the current study is unable to confirm if the RBDs tested will convert to PD. With such high error rates in diagnosing PD clinically, this model is a potentially beneficial and could serve as an easy screening tool to add to the suite of diagnostic tests and improve clinician’s ability to diagnose accurately.

Rights

© The Author

Is Part Of

VCU University Archives

Is Part Of

VCU Theses and Dissertations

Date of Submission

12-9-2019

Download

Share

COinS