Multimodal neuroimaging and electrophysiological signatures in amnestic mild cognitive impairment with different amyloid-β deposition status: correlations with cognitive function and diagnostic value

Authors:
Open Access
Publisher website
Google Scholar
Abstract
Background: Amnestic mild cognitive impairment (aMCI) is a heterogeneous transitional cognitive state between normal aging and dementia. Aβ positivity elevates dementia conversion risk but does not equate to prodromal Alzheimer’s disease (AD), and not all Aβ-positive aMCI individuals progress to clinical AD dementia. Distinguishing Aβ-positive from Aβ-negative aMCI facilitates pathological subtyping rather than definitive early AD diagnosis. This cross-sectional study aimed to characterize scalp-level resting-state EEG and structural MRI signatures specific to Aβ-positive aMCI, explore their covariate-adjusted correlations with cognitive performance and regional Aβ burden, and develop an internally validated multimodal auxiliary classification model for Aβ stratification. Methods: Fifty enrolled aMCI participants (30 Aβ-positive, 20 Aβ-negative) and 35 cognitively normal controls (NC) completed unified 20-min eyes-closed resting-state scalp EEG, 3.0 T structural MRI, 18 F-AV45 amyloid PET and 18 F-FDG cerebral metabolism PET examinations. Artifact-free scalp EEG spectral power and regional θ/α ratio were calculated via Welch spectral analysis; voxel-based morphometry (VBM) was adopted to quantify whole-brain gray matter volume. Covariate-controlled partial correlation corrected for multiple comparisons was applied to evaluate cross-modal associations among cortical Aβ SUVR load, cerebral glucose metabolism and domain-specific cognitive scores. A binary logistic regression classification model combining screened EEG and MRI markers was established, with DeLong test for ROC curve comparison and 1,000-bootstrap internal validation. Results: (1) Relative to Aβ-negative aMCI and NC groups, the Aβ-positive subgroup presented significantly elevated whole-scalp θ relative power and scalp θ/α ratio, decreased scalp α and β relative power (all q < 0.05, FDR-corrected), with predominant scalp oscillatory alterations distributed over temporal, parietal and occipital scalp electrode clusters. The temporal scalp θ/α ratio yielded optimal classification performance for Aβ stratification (AUC = 0.875). (2) The Aβ-positive subgroup showed significant gray matter atrophy across 13 AD-vulnerable brain regions including bilateral fusiform gyrus, inferior parietal lobule, parahippocampal gyrus, precuneus and right entorhinal cortex (all q < 0.05), with the maximum effect size identified in the right precuneus (partial η 2 = 0.477, AUC = 0.778). (3) Within the Aβ-positive subgroup, the temporal scalp θ/α ratio was positively correlated with medial temporal Aβ SUVR ( r = 0.501, 95% CI [0.146, 0.732], q = 0.030); right precuneus gray matter volume was negatively correlated with multi-region cortical Aβ burden ( q < 0.05). Left parahippocampal gyrus volume was positively correlated with MoCA memory subscores ( r = 0.479, 95% CI [0.118, 0.709], q < 0.05). No significant associations between scalp EEG metrics and FDG glucose metabolism survived FDR correction. (4) The multimodal logistic regression model integrating temporal scalp θ/α ratio and right precuneus gray matter volume achieved favorable classification efficacy for Aβ-positive aMCI differentiation (AUC = 0.852, 95% CI [0.748, 0.955]), with good calibration and clinical net benefit based solely on internal bootstrap validation. Conclusion: The temporal scalp θ/α ratio and right precuneus gray matter volume are promising candidate non-invasive auxiliary biomarkers for Aβ pathological stratification in aMCI, rather than validated clinical diagnostic biomarkers. The multimodal model provides supplementary subtyping value within this single-center cohort, while external multicenter validation is mandatory before clinical application. This cross-sectional dataset cannot support prediction of longitudinal progression to AD dementia.

This publication has 39 references indexed in Scilit: