Stage-specific neurovascular complexity disruption in Alzheimer’s spectrum: A resting-state fNIRS study using HFD, spectral and wavelet entropy

Early and noninvasive detection of Alzheimer’s disease (AD) and its prodromal stage, mild cognitive impairment (MCI), is critical for timely intervention. This study introduces a stage-specific framework that integrates nonlinear complexity metrics—Higuchi’s fractal dimension, spectral entropy, and wavelet entropy—applied to resting-state functional near-infrared spectroscopy (fNIRS) signals. fNIRS data were recorded from the prefrontal cortex of 83 individuals (AD: 19, MCI: 37, healthy controls (HC): 27), and complexity features were systematically analyzed across hemoglobin species. HFD values declined progressively from HC to MCI to AD, whereas SE and WE increased in AD, reflecting altered signal irregularity. Intersection analysis identified a robust biomarker in the right frontopolar prefrontal cortex (channel 12), consistently discriminative across features and hemoglobin types. Classifiers trained on these core features outperformed full-channel models, achieving an AUC of 0.889 with low variability across folds. Complexity features also correlated strongly with Mini-Mental State Examination scores, underscoring their clinical relevance. These findings demonstrate, for the first time, that nonlinear complexity analyses of resting-state fNIRS can reveal stage-specific neurovascular disruptions across the Alzheimer’s spectrum. By identifying robust, interpretable channel-level biomarkers, our framework advances resting-state fNIRS from a supportive technique to a clinically actionable tool for early diagnosis and staging of AD.