Bipolar EEG Power Spectral Density Analysis of Resting-State for Multi-Level Cognitive Impairment Evaluation

Abstract

The prevention of dementia is a significant concern nowadays, with particular clinical focus on evaluating mild cognitive impairment (MCI) as a precursor to dementia. Plenty of indicators of cognitive decline require comprehensive investigation. This study utilized bipolar EEG to examine regional brain waves and power spectral density (PSD) properties to explore the distinctions across different levels of cognitive impairment. This study compared PSD-related features: absolute power, average power, and individualized alpha peak frequency of different brain areas among three distinct level cognitive impairment groups, comprising 50 healthy controls, 70 MCI patients, and 70 dementia patients. The results showed that the dementia group had considerably higher delta and theta band power and lower peak frequencies than the other two groups across the brain. Moreover, we localized significant differences in EEG features between the healthy control and MCI groups to low-frequency power in the frontal (Fp1-F3), temporal (F7-T3), and parietal (C3-P3) areas. Furthermore, we propose preliminary dementia and MCI models using these bipolar EEG features to evaluate different cognitive levels, achieving high performance as measured by ROC AUC over 81.7% and 63.79%, respectively. These results indicated that bipolar EEG-derived PSD features can effectively differentiate cognitive impairment levels, suggesting the potential for cognitive impairment evaluation for clinical applications.Clinical Relevance- Bipolar EEG-derived power spectral density features provide a non-invasive, cost-effective approach for early detection and evaluation of cognitive impairment progression.