AI-generated summary
The authors studied resting-state eyes-closed EEG from 88 participants (36 with Alzheimer's disease, 23 with frontotemporal dementia, 29 healthy controls) to classify groups using spectral and directed connectivity features. The spectral-only model performed best for three-class classification, and the combined model distinguished dementia from controls, but Alzheimer's versus frontotemporal dementia remained difficult. The authors describe the photobiomodulation analysis as exploratory, based on a single case.
- Participants
- 88
- Who or what was studied
- Adults with Alzheimer's disease (36), frontotemporal dementia (23), and healthy controls (29), plus one exploratory single-case photobiomodulation participant
- Limits stated in the abstract
- Modest sample size, cross-validation results with limited balanced accuracy, group differences not surviving FDR correction, reduced hyperparameter grid in permutation testing, and an exploratory single-case photobiomodulation analysis.
Written by an AI model from this paper’s abstract only and checked automatically against it, but it can still be wrong. It is not medical advice. Read the full paper. See the editorial policy.
Abstract
Alzheimer's disease (AD) and frontotemporal dementia (FTD) are neurodegenerative disorders with partially overlapping clinical manifestations, making early and differential diagnosis challenging. This study investigated whether electroencephalography (EEG)-derived spectral features and Granger-causality (GC)-based directed functional connectivity features can characterize and classify AD, FTD, and healthy control (HC) subjects. Resting-state eyes-closed EEG recordings from 88 participants were analyzed, including 36 AD, 23 FTD, and 29 HC subjects. Spectral features included absolute and relative band power and spectral ratios, while directed connectivity features were extracted from broadband and frequency-specific GC matrices. Statistical analyses identified theta/alpha ratio (TAR) as the dominant spectral marker, with the strongest three-group differences observed in frontal and global TAR features. GC analysis revealed group-related alterations mainly in alpha-band regional directed connectivity, although three-group GC features did not survive false discovery rate (FDR) correction at q < 0.05. In the main nested cross-validation analysis, the spectral-only model achieved the best three-class performance, with balanced accuracy of 0.572 and macro-F1 of 0.557. For dementia group (DEM) vs. HC classification, the combined GC + spectral feature (GC + SPEC) set achieved balanced accuracy of 0.710 and macro-F1 of 0.665. For AD vs. FTD classification, the combined GC + SPEC feature set achieved the highest numerical performance in the main nested cross-validation (CV) comparison, with balanced accuracy of 0.584 and macro-F1 of 0.559. In the separate long permutation-testing analysis, which used a reduced hyperparameter grid for computational feasibility, above-chance performance was confirmed for the three-class spectral model and the DEM vs. HC GC + SPEC model (p < 0.001), but not for AD vs. FTD (p = 0.270). An exploratory photobiomodulation (PBM) single-case analysis showed longitudinal EEG reorganization, including increased alpha power, reduced delta/alpha ratio (DAR) and beta/alpha ratio (BAR), mixed TAR changes, and HC-like GC/GC + SPEC centroid projections. Overall, the results support the value of spectral and directed connectivity EEG markers for dementia-related EEG characterization, while highlighting the persistent difficulty of AD vs. FTD differentiation.
Abstract reproduced under the CC BY license. Source: PubMed, a service of the National Library of Medicine.
How to read a single study: what to check before relying on this kind of research.
Terms in this paper: Electroencephalography, Dementia, Photobiomodulation. Each links to its definition in the glossary.
A summary of published research, not medical advice. One study rarely settles a question. See the disclaimer and how to read research.