Abstract
BACKGROUND: Oculomotor abnormalities are recognized as potential biomarkers in Parkinson’s disease (PD). However, their broader clinical adoption can be hindered by the accessibility and practicality of assessment tools. OBJECTIVE: This study evaluates a virtual reality (VR)-integrated eye-tracking system for characterizing oculomotor function in PD. We aimed to comprehensively assess oculomotor performance in PD patients compared to healthy controls (HC), investigate its relationship with clinical measures of disease severity, and develop a diagnostic model based on multi-paradigm eye movement parameters to distinguish PD from HC. METHODS: Eye movements were recorded in 44 PD patients and 34 HC participants during horizontal and vertical saccade, anti-saccade, and smooth pursuit tasks. Parameters were compared between groups and directions. Correlations with duration, disease severity, motor symptoms, medication, and specific eye movement metrics were analyzed. Logistic regression identified key diagnostic predictors, and their performance was evaluated using ROC analysis. RESULTS: PD patients demonstrated significant oculomotor impairments across multiple tasks, with notably worse performance in the vertical direction compared to the horizontal. Specific parameters, particularly in anti-saccade and smooth pursuit, were significantly correlated with clinical disease severity scores. A diagnostic model combining three key eye movement parameters achieved an AUC of 0.932, showing significantly better discriminatory power than any single parameter alone for distinguishing PD from HC. CONCLUSIONS: This study validates that a multi-parameter oculomotor model can effectively differentiate PD patients from healthy controls. The findings support the value of quantitative eye movement analysis, as a promising auxiliary tool for PD assessment.