Answers are grounded in this paper and cite their sources.
Background Oculomotor dysfunction is increasingly recognized in Alzheimer’s disease (AD), but conventional eye-tracking approaches have limited clinical scalability. Methods We investigated whether oculomotor features acquired using a tablet-based eye-tracking platform could provide an objective signature of AD and clinical disease severity. Seventy-five participants with AD and 75 cognitively and neurologically healthy controls (HC) completed five oculomotor tasks using an iPad-based eye-tracking platform. A broad set of oculomotor features was used to develop 13 higher-order composite features summarizing recurrent task- and signal-level patterns. AD–HC classification was evaluated across complementary model families using repeated nested cross-validation, and associations between classifier output, composite measures, cognition, and clinical severity were subsequently examined. Results The best model achieved an AUC of 0.906, balanced accuracy of 0.860, sensitivity of 85.3%, and specificity of 86.7%, with broadly similar performance across regularized linear and nonlinear models. Discrimination performance increased with disease severity, with sensitivity rising from 73.0% in participants with CDR 0.5 to 96.8% with CDR 1.0, and misclassified participants tended to show oculomotor profiles resembling the opposite diagnostic group. Within AD, higher classifier-derived probability was associated with poorer MoCA and MMSE performance and greater CDR-SB severity. Conclusion These findings support the presence of a compact, multidomain oculomotor signature of AD that reflects both diagnostic status and clinical disease severity. Although independent validation of the predefined composites and classifier is needed to establish their generalizability, the results highlight the potential of tablet-based eye tracking as a scalable and objective approach for capturing these signals.
No takes yet. Share an insight, caveat, or question.
Voss et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: