Inferring amyloid pathologies among patients with mild cognitive impairment using phase‒amplitude coupling of electroencephalography: A case‒control study

Early detection of Alzheimer's disease (AD) is crucial in the era of disease-modifying therapies (DMTs). While amyloid positron emission tomography (PET) is an important approach for detecting amyloid pathologies, the high cost and limited availability of this method have led to challenges in community-wide screening. Electroencephalography (EEG) is a low-cost, noninvasive method that may help to bridge this gap. Previous studies have suggested that certain EEG features, including phaseamplitude coupling (PAC), could reflect AD-related network abnormalities; however, it remains unclear whether PAC can serve as a reliable predictor of amyloid pathology. We analyzed resting-state EEGs from 61 participants with mild cognitive impairment (MCI) who underwent amyloid PET scans: 28 were PET-positive, and 33 were PET-negative. We examined PAC (synchronization indices) between low-frequency (delta, theta, alpha, beta) and high-gamma (70-100 Hz) activities. A naïve Bayes classifier with a stratified 10-fold cross-validation was used to predictor amyloid PET status. Theta-high gamma PAC was lower in the PET-positive group, particularly at the posterior (P3) and occipital (O2) electrodes (p < 0.05). The classifier trained on the theta-high gamma PAC achieved a mean balanced accuracy of 64.6 % (95 % confidence interval = 0.599-0.693), which was significantly greater than chance, whereas the other bands were not predictive. Theta-high gamma PAC in resting-state EEG can distinguish amyloid PET-positive patients with MCI from PET-negative patients, thereby providing a feasible, noninvasive biomarker for AD pathology. Integrating PAC analysis into routine EEG could improve the early identification of individuals who qualify for DMTs, thereby facilitating timely intervention.