Abstract
Purpose This study investigates the dynamic regulation of cerebral oxygenation in individuals with mild cognitive impairment (MCI) compared to healthy controls, using a novel pseudorandom binary sequence (PRBS) gas challenge. Traditional assessments often overlook frequency-dependent features of cerebrovascular control; this work aims to uncover latent deficits through broadband perturbations and nonparametric dynamic modeling. Methods Seventeen ApoE4-negative participants (9 controls, 8 MCI) underwent a three-session supine protocol involving PRBS-modulated inhalation of hypoxic and hypercapnic gas mixtures. Physiological signals—including near-infrared spectroscopy (TOI), arterial pressure (ABP), and end-tidal CO 2 (PETCO 2 )—were recorded. Laguerre-based Volterra modeling and Principal Dynamic Mode (PDM) decomposition were used to analyze the dynamic response of cerebral oxygenation to ABP and CO 2 stimuli. Group differences were statistically assessed using Welch’s t -tests and repeated measures ANOVA. Results Controls showed significant increases in ventilation and TOI from baseline to stimulation ( $$p p 0.01 ), while MCI did not. Gain of the first CO 2 PDM increased in controls during PRBS ( p = 0.016) but decreased in MCI. A low-frequency ABP-derived PDM ( $$\sim$$ ∼ 0.014 Hz), consistent with endothelial-dependent vasodilation, was elevated only in controls during stimulation. Persistent differences in spectral recovery, kernel profiles, and PDM Gains suggested impaired baroreflex and chemoreflex regulation, and possible disruption of cholinergic-linked neurovascular coupling in MCI. Conclusion PRBS gas modulation combined with dynamic modeling revealed subtle but significant cerebrovascular control impairments in MCI. This methodology enables mechanistic insights into early pathophysiology and may aid future physiomarker development.