Abstract
The locus coeruleus (LC), the earliest site of tau accumulation in Alzheimer’s disease, directly influences pupillary and vergence oculomotor systems. Whether task-evoked oculomotor dynamics can non-invasively differentiate AT(N) profiles, A+T+ (biological Alzheimer’s disease) and A−T+ (suspected non-Alzheimer’s tauopathy), remains unexamined despite their differential impact on LC-mediated modulation and cortical network integrity. Thirty-eight MCI individuals (12 A−T+, 26 A+T+), classified using validated CSF biomarker thresholds, completed a visual oddball paradigm under binocular eye-tracking. Linear mixed-effects models examined profile × condition interactions across full time series and six trial-level features; participant-level associations were tested using Firth-penalized logistic regression to account for the small and imbalanced sample. Profiles did not differ in oculomotor magnitude but diverged in temporal organization. Profile × condition interactions were significant for vergence global slope, vergence time to peak, and pupillary time to peak (all p-Holm ≤ 0.048). Timing differences were condition-dependent: A−T+ showed later responses during distractor trials, while A+T+ showed the most delayed timing during target trials. Subject-level associations were significant for vergence and pupillary time to peak, with greater condition-dependent modulation predicting higher A+T+ probability (ORs = 0.43 and 0.35; both p < 0.05); vergence global slope delta did not reach significance at the subject level. A−T+ showed superior target detection accuracy (89.7% vs. 82.5%, p = 0.003). Cognitive vergence and pupillary dynamics showed condition-dependent patterns differentiating AT(N) biological profiles in this MCI sample, supporting portable, low-cost eye-tracking as a non-invasive complement to CSF-based characterization, pending confirmation in larger, independent cohorts.