Single Channel EEG Based Score Generation to Monitor the Severity and Progression of Mild Cognitive Impairment


Mild Cognitive Impairment (MCI) is a preliminary stage of Dementia. MCI is determined by behavioral screening measures such as Montreal Cognitive Assessment (MoCA) and Mini-Mental Status Examination (MMSE). Therefore, monitoring the progression of MCI and predicting MoCA scores from objective physiological measures like the EEG is crucial as it will not only help to improve the mental healthcare of the aging population but also to reduce healthcare costs. In this study, we demonstrate a single channel EEG based MoCA score generation method, which is cost-effective and suitable for continuous patient monitoring in the longitudinal study. We collected scalp EEG data while subjects were stimulated with five auditory speech signals. We extracted 590 features from Event-Related Brain Potentials (ERPs), which included time and spectral domain characteristics of the response. The top 11 features, ranked by mutual information, were used for building regression models to generate MoCA scores of the subjects. Robustness of our model was tested using R-squared value, mean square error (MSE), residual's quantile plot, and cook's distance. The analysis shows R-squared=0.78 with MSE=1.63, and residual analysis suggests that the model is acceptable in terms of quantile plot, leverage, and Cook's distance. The outcomes indicate that single-channel based EEG can be used to estimate cognitive scores automatically for severity detection and progression monitoring, which will help us to efficaciously assess the mental health status of elderly people to improve the prognosis and rehabilitation of age-related cognitive impairments.

Publication Title

IEEE International Conference on Electro Information Technology