Automatic Extraction of Oculographic Signals as Digital Biomarkers for Alzheimer's Disease

Abstract

BackgroundSimilar to other neurological diseases, subtle early neurological changes that occur in Alzheimer’s Disease (AD) are difficult to quantify and track objectively. One relevant bio‐signal is eye movement, as it presents a unique window into cognitive processes as a direct measure of real‐time inputs to the brain. Eye‐tracking allows us to estimate timestamped activity of neural function and begin to infer and quantify attention, reprocessing (eg. word revisits), and many other aspects relevant to assessing AD. The goal of this study was to investigate the saccadic movements and trial progress during administration of the Stroop test in AD compared to normal controls.Method5 AD subjects and 12 age‐matched cognitively normal controls were recruited from the Johns Hopkins Memory and Alzheimer’s Treatment Center and the Movement Disorders Clinic at the Johns Hopkins University School of Medicine. An Eyelink Portable Duo in head‐free‐to‐move mode was used to track eye movements. Movement of the eyes were analyzed by automatically segmenting saccade movements throughout a trial. We then analyzed the eye movements in the context of cognitive performance, by identifying when subjects progress to each word in the Stroop test. We compared the performance of each group using a Wilcoxon RankSum test with the SciPy package using Python.ResultPreliminary results show that 60% (3/5) of AD subjects were unable to complete the Stroop task in 80 seconds. Individuals with AD had more saccades (p=0.105) with greater saccadic overshoot (p=0.092) and a greater maximum saccade velocity (p=0.035) on average compared to cognitively normal controls. AD subjects also took much longer to complete the first (p=0.114) and second lines (p=0.092) compared to healthy controls, suggesting initial confusion and hesitation.ConclusionAD subjects change their gaze more often with greater speed and less precision than age‐matched controls. These data demonstrate feasibility of using automatic quantitative evaluation of eye movements in AD. In the future, we will determine the utility of using this tool for early detection of AD and tracking its progression, along with combining other measures such as voice and handwriting (multimodal) to provide insights into early physiological changes in AD.