Search for a command to run...
Early detection of Alzheimer’s disease (AD) requires biomarkers sensitive to pre-neurodegenerative dysfunction. Task-evoked ocular responses index arousal‑based gain and the stability of executive timing, but their relationship to brain structure, especially within AD‑signature cortex; regions that thin early in AD, remains poorly characterized. We tested whether ocular metrics map onto cortical thickness and subcortical volumes, and whether brain–ocular coupling differs between cognitively normal (CN) adults and those with mild cognitive impairment (MCI). Participants with MCI (n = 212) and CN controls (n = 516) completed an interleaved prosaccade–antisaccade task; binocular pupil diameter and eye movements were time‑locked to cue and target onsets to derive pupil amplitude/variability and saccade metrics. Region‑wise linear mixed‑effects models quantified brain–ocular coupling to cortical thickness and subcortical volumes and tested group (CN vs. MCI) differences in coupling slopes. Diagnosis‑independent analyses showed that saccade latency variability (Latency SD) and pupil amplitude/variability exhibited robust region‑dependent coupling across cortical and subcortical volumes, whereas mean saccadic latency and pupil timing measures displayed no reliable spatial pattern. Diagnostic effects were modest and spatially selective but directionally opposite across modalities. For saccades, CN displayed positive thickness–variability slopes in frontal, cingulate, and insular cortices whereas MCI showed negative or near‑zero slopes. In the AD‑signature cortex, this inversionn CN(+)/MCI(−/0) was localized specifically to the supramarginal gyrus (Δβ≈–0.026). Subcortical volumes showed no significant diagnostic differences. For pupils, MCI showed more positive pupil–thickness coupling than CN within global cortex, most prominently in temporal, parietal, and insular regions, and within AD‑signature cortex this selective increase localized to medial temporal cortex (Δβ = 0.042) and to supramarginal gyrus (Δβ = 0.030); subcortical diagnostic differences were not significant after correction. Task‑evoked ocular signals yield two complementary, region‑specific readouts of early Alzheimer‑relevant dysfunction. Pupil amplitude and variability reflect the strength of phasic, arousal‑linked responses arising from coordinated brainstem control hubs and show selective increases in MCI within medial temporal and temporoparietal regions that are vulnerable early in Alzheimer’s disease. Saccadic latency variability indexes loss of timing stability, revealing a thickness–stability inversion—positive in controls, absent or negative in MCI—that localizes to supramarginal gyrus within attention and executive‑control networks. Effects are modest and spatially circumscribed, positioning ocular measures as scalable, adjunct biomarkers of prodromal disease.