Go to OpenReview Public Article ORCID homepageMEG Spectral Biomarkers of Alzheimer's Disease: Integrating MEG and MRI Features Using the BioFIND Dataset
Abstract: Background: Alzheimer's disease (AD), the leading cause of dementia, disrupts brain communication through amyloid-beta and tau protein accumulation, leading to memory and cognitive impairments. Magnetoencephalography (MEG) and magnetic resonance imaging (MRI) offer non-invasive approaches to investigate these changes. Objective: To assess the value of combining MEG power spectral density with MRI-derived brain volumetrics and to evaluate a novel classification approach using sign-constrained logistic regression with L1 regularization (GLMNET). Methods: The BioFIND dataset, including MEG from 324 participants (<tex>$\mathbf{1 5 8 ~ M C I}$</tex>, <tex>$\mathbf{1 6 6 ~ H C}$</tex>) and MRI for most subjects, was analysed. MEG source localization was performed using linearly constrained minimum variance (LCMV) beamforming and exact low-resolution electromagnetic tomography (eLORETA), which were applied separately for MEG's magnetometer (MAG) and gradiometer (GRAD) signals. MRI regional volumes were extracted with Freesurfer. Correlation-based feature preselection for different thresholds <tex>$(0: 0.05: 0.25)$</tex> was applied, and classification was conducted using Monte Carlo of a replicated 10fold nested cross-validation. Results: Highest performance was obtained by combining LCMV MAG and GRAD with MRI features at 0 -correlation threshold, with an accuracy of 77.9 % and F1 score of 75.7 %. Conclusions: This study demonstrates the effectiveness of MEG MAG GARD integrated with MRI in distinguishing healthy ageing from cognitive decline. By comparing classification performance for different combinations, selected through multiple source localization methods and varying correlation thresholds, their potential importance was assessed more robustly.
External IDs:doi:10.1109/icbip66535.2025.11284312
Loading