Rationale:
Despite rapid gains in the fields of cognitive neuroscience, data science, and technology, assessment in clinical neuropsychology has remained stagnant. Current assessments are insensitive to early, subtle deficits and coarsely approximate real-world episodic memory function. Eye movements have recently been shown to track distinct memory processes and are associated with hippocampal function. Thus, we wondered if eye-tracking could provide sensitive and robust measure of memory and could be diagnostic of memory deficits observed in epilepsy patients.
Methods:
Twenty-two (22) patients with temporal lobe epilepsy (TLE) and twelve (12) healthy control (HC) subjects participated in a computer-based inferential learning task. Eye-tracking was measured while participants performed the task, which assessed memory for paired visual objects (direct association) as well as inferentially linked items (items linked by a common item in training pairs). Eye tracking metrics (e.g., number of fixations, duration of eye movements, dwell-time on specific items, and eye movement entropy across the visual scene) were acquired during the test phase (recognition) and generalized linear mixed-effects (GLME) models were used to test the association between eye movement behaviors and subsequent memory (e.g., successful vs. failed trials). In addition, GLMEs were used to test whether eye movements could be diagnostic for TLE in this cohort of subjects.
Results:
Behaviorally, while TLE patients exhibited greater variability in explicit recognition memory performance, we did not find a significant difference between TLE patients and HCs in their recognition accuracy for either direct (p=0.15, Wilcoxon’s ranksum test) or indirect association memory (p=0.1). Across both groups of participants, we found that eye tracking metrics such as number of fixations, eye-movement entropy, and dwell time were predictive of successful versus failed recognition in both the direct and indirect associations. Dwell time on the target (e.g., percentage of time spent looking at the target object) showed the greatest predictive power for dissociating between correctly associated and failed trials. When examining eye movement behaviors between groups, eye movement entropy predicted group assignment (TLE vs. HC) in both task conditions (direct & indirect associations), indicating that entropy captured the subclinical memory deficits in TLE patients that were not observed during explicit behavior.
Conclusions:
While explicit measures of direct and indirect association memory failed to distinguish between HCs and patients with TLE, basic measures of eye movements, such as fixations, and looking behaviors, such as dwell time, were predictive of successful vs. failed memory. In addition, more complex eye movement behaviors such as entropy were diagnostic of subclinical memory deficits. These findings provide exciting new evidence for the use of novel, non-invasive eye-tracking in the detection and diagnosis of memory deficits.
Funding:
Finding a Cure for Epilepsy and Seizures (FACES), NYU Comprehensive Epilepsy Center