Objective: The objective of the study was to identify EEG features associated with cognitive fluctuations in patients with Lewy body dementia (LBD).

Background: Cognitive fluctuations (CF), defined as spontaneous and time-varying periods of impaired attention and reduced arousal, are a characteristic feature of Lewy body dementia (LBD). An easily obtained and widely available biomarker of CF would greatly advance the ability to diagnose LBD earlier so patients receive timely care. Previous studies suggest specific EEG features correlate with CF and therefore may serve as a diagnostic biomarker.

Method: This prospective cross-sectional study enrolled 50 patients with Parkinson’s disease (PD), PD with dementia and dementia with Lewy bodies.  Participants were divided into LBD with CF (LBDwCF, n=29) and Lewy body disease without CF (LBwoCF, n=21). Resting-state EEG data were processed and analyzed to determine the following features in anterior, temporal, and posterior regions: dominant frequency (DF), dominant frequency variability (DFV) and individual alpha peak frequency (IAF). We used univariate statistical tests, correlational analyses to reduce the model space, multivariate logistic regression models, and elastic net penalty to select the final model to predict CF from EEG feature values. Level of significance was set at p<0.05.

Results: Using correlational analyses, we selected DF-posterior, IAF-posterior, mean DFV, and dominant frequency prevalence – delta (DFP-delta) to reduce multicollinearity among the covariates. LBDwCF had significantly reduced median DF-posterior (6.0, interquartile range (IQR): 5.5-6.9 vs. 8.4, IQR: 7.6-9.2)  and reduced median IAF-posterior (8.0, IQR 8.0-8.4 vs 8.7, IQR: 8.2-10.2) compared to LBwoCF. Only DF-posterior was associated with group after adjustment for age, sex, and Montreal Cognitive Assessment score (OR=0.26, 95% CI: 0.06-0.88, p=0.04). Using elastic net regression with all 4 EEG features, k=100 fold cross validation with 80-20 split, the mean accuracy was 0.86 (95% CI: 0.84-0.88) and the mean AUC was 0.84 (0.82-0.87).

Conclusion: LBDwCF have reduced DF-posterior and IAF-posterior compared to LBwoCF. A model including DF-posterior, IAF-posterior, mean DFV and DFP-delta has good accuracy in predicting CF in Lewy body disease. Incorporating EEG-based biomarkers into clinical practice could improve early diagnosis and management of LBD by providing a non-invasive, accessible tool for detecting CF.

« Back to 2025 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/advancing-lbd-diagnosis-eeg-biomarkers-for-cognitive-fluctuations/