Multi‐task analysis of oculographic biomarkers to evaluate motoric and cognitive patterns in Alzheimer’s Disease

Abstract

AbstractBackgroundAlzheimer’s disease (AD) is characterized by progressive loss or damage to the nervous system which can cause impairment in movement, coordination, and cognition. Due to the complexity of clinical features at different stages of the disease, it may require months for an accurate diagnosis. Oculomotor researchers have demonstrated the utility of examining eye movements for differentiating various neurodegenerative disorders including AD, Parkinson’s disease (PD), and other PD mimics (PDM). In this study, we propose an automatic pipeline to identify saccade, fixation, and blinks and extract response‐to‐stimuli‐derived interpretable biomarkers that could inform clinicians in making accurate assessment.MethodEye tracking data in smooth pursuit, pro‐saccade, and anti‐saccade tasks were recorded from 146 participants ‐ 14 with AD or Mild Cognitive Impairment (MCI), 46 with PD, 16 with PDM, and 70 age‐matched normal controls (CTL). Biomarkers based on the identified eye movement and interaction with the stimuli were extracted with our algorithms. Then, pairwise Kruskal‐Wallis H tests were conducted to determine if there were statistically significant differences between groups. The Benjamini‐Hochberg correction was performed to control the false discovery rate.ResultThe results show that, in the smooth pursuit task, the AD/MCI group had a significantly higher (p < 0.05) saccade count than CTL and PD groups. The AD/MCI group also exhibited significantly longer (p < 0.05) average saccade duration than CTL and PDM groups. The mean and standard deviation for the number of saccadic adjustments to reach and stay on the target significantly differed (p < 0.05) between AD/MCI and CTL and between PD and CTL in trials of the pro‐saccade task. The magnitude of overshoot or undershoot was significantly greater (p < 0.05) for AD/MCI, PD, and PDM compared to CTL. Moreover, the number of saccades in the wrong direction was significantly higher (p < 0.05) for AD/MCI and PD compared to CTL in the anti‐saccade task.ConclusionOculographic biomarkers can be leveraged within a machine learning framework and may be useful in distinguishing different types of neurodegenerative diseases in their early stages yielding more objective and precise protocols to monitor disease progression than current methods.