Monitoring Disease Severity of Mild Cognitive Impairment from Single-Channel EEG Data Using Regression Analysis

Area of Science:

• Neuroscience
• Biomedical Engineering
• Medical Informatics

Background:

• Mild cognitive impairment (MCI) is a decline in cognitive health, necessitating early monitoring to prevent progression to dementia, Alzheimer's disease (AD), and Parkinson's disease (PD).
• Current MCI severity assessment relies on manual scoring using the Montreal Cognitive Assessment (MoCA), which can be time-consuming and subjective.

Purpose of the Study:

• To develop and evaluate a novel algorithm for automated MCI severity scoring using single-channel electro-encephalography (EEG) data.
• To generate automated severity scores equivalent to MoCA scores, enabling more efficient and objective monitoring of cognitive health.

Main Methods:

• Feature extraction from event-related potential (ERP) points and time-domain characteristics of multi-trial EEG data.
• Application of lasso regression for feature selection, followed by multivariate regression (MR), ensemble regression (ER), support vector regression (SVR), and ridge regression (RR).
• Development of a convolutional deep neural network (CNN) model for single-trial analysis using time-frequency plot images.

Main Results:

• Ensemble regression (ER) achieved the best performance in multi-trial analysis with a Root Mean Square Error (RMSE) of 1.6.
• The single-trial CNN model demonstrated an RMSE of 2.76.
• This research represents the first attempt to automate MoCA-equivalent MCI severity scoring from single-channel EEG data using both multi-trial and single-trial analyses.

Conclusions:

• The proposed automated algorithm effectively estimates MCI severity using single-channel EEG data.
• The multi-trial analysis with ensemble regression shows promising results for objective and efficient MCI monitoring.
• This approach offers a potential advancement in the early detection and management of cognitive impairments.