Predicting Dementia With Prefrontal Electroencephalography and Event-Related Potential

Dieu Ni Thi Doan; Boncho Ku; Jungmi Choi; Miae Oh; Kahye Kim; Wonseok Cha; Jaeuk U Kim

doi:10.3389/fnagi.2021.659817

Predicting Dementia With Prefrontal Electroencephalography and Event-Related Potential

Front Aging Neurosci. 2021 Apr 13:13:659817. doi: 10.3389/fnagi.2021.659817. eCollection 2021.

Authors

Dieu Ni Thi Doan^{1

2}, Boncho Ku¹, Jungmi Choi³, Miae Oh⁴, Kahye Kim¹, Wonseok Cha³, Jaeuk U Kim^{1

2}

Affiliations

¹ Korea Institute of Oriental Medicine, Daejeon, South Korea.
² Korean Convergence Medicine, University of Science and Technology, Daejeon, South Korea.
³ Human Anti-Aging Standards Research Institute, Uiryeong-gun, South Korea.
⁴ Korea Institute for Health and Social Affairs, Sejong, South Korea.

Abstract

Objective: To examine whether prefrontal electroencephalography (EEG) can be used for screening dementia. Methods: We estimated the global cognitive decline using the results of Mini-Mental Status Examination (MMSE), measurements of brain activity from resting-state EEG, responses elicited by auditory stimulation [sensory event-related potential (ERP)], and selective attention tasks (selective-attention ERP) from 122 elderly participants (dementia, 35; control, 87). We investigated that the association between MMSE and each EEG/ERP variable by using Pearson's correlation coefficient and performing univariate linear regression analysis. Kernel density estimation was used to examine the distribution of each EEG/ERP variable in the dementia and non-dementia groups. Both Univariate and multiple logistic regression analyses with the estimated odds ratios were conducted to assess the associations between the EEG/ERP variables and dementia prevalence. To develop the predictive models, five-fold cross-validation was applied to multiple classification algorithms. Results: Most prefrontal EEG/ERP variables, previously known to be associated with cognitive decline, show correlations with the MMSE score (strongest correlation has |r| = 0.68). Although variables such as the frontal asymmetry of the resting-state EEG are not well correlated with the MMSE score, they indicate risk factors for dementia. The selective-attention ERP and resting-state EEG variables outperform the MMSE scores in dementia prediction (areas under the receiver operating characteristic curve of 0.891, 0.824, and 0.803, respectively). In addition, combining EEG/ERP variables and MMSE scores improves the model predictive performance, whereas adding demographic risk factors do not improve the prediction accuracy. Conclusion: Prefrontal EEG markers outperform MMSE scores in predicting dementia, and additional prediction accuracy is expected when combining them with MMSE scores. Significance: Prefrontal EEG is effective for screening dementia when used independently or in combination with MMSE.

Keywords: Alzheimer’s disease; Mini-Mental Status Examination; dementia; electroencephalography; electrophysiology; event-related potential.