The EEG Signal Analysis for Spatial Cognitive Ability Evaluation Based on Multivariate Permutation Conditional Mutual...

Area of Science:

• Neuroscience
• Cognitive Science
• Biomedical Engineering

Background:

• Evaluating cognitive training efficacy, particularly spatial memory, is crucial for understanding brain function.
• Virtual reality (VR) offers immersive environments for cognitive training, but objective assessment methods are needed.
• Electroencephalography (EEG) provides non-invasive brain activity data, requiring advanced analysis for complex cognitive tasks.

Purpose of the Study:

• To develop and validate an effective EEG signal analysis method for assessing spatial memory cognitive training in a VR environment.
• To classify EEG signals from early and late stages of spatial cognitive training to objectively measure training efficacy.
• To identify reliable EEG biomarkers indicative of successful spatial memory enhancement.

Main Methods:

• A novel EEG signal analysis method, Multivariate Permutation Conditional Mutual Information-Multi-Spectral Image (MPCMIMSI), was proposed.
• MPCMIMSI considers coupled features between EEG channel pairs, transforming them into multi-spectral images.
• A convolutional neural network (CNN) model was employed to classify these images, differentiating training stages.

Main Results:

• The MPCMIMSI method demonstrated superior classification performance compared to Granger Causality Analysis (GCA) and Permutation Conditional Mutual Information (PCMI).
• Classification accuracy reached up to 95%, with the Theta-Beta2-Gamma band combination showing the highest efficacy.
• The MPCMIMSI feature in the Theta-Beta2-Gamma band was identified as a potential biomarker for spatial memory training assessment.

Conclusions:

• The MPCMIMSI method provides an effective and objective approach to evaluate spatial memory training efficacy using non-invasive EEG.
• This novel EEG feature extraction technique offers new insights into characterizing spatial information from brain recordings.
• The proposed method holds potential for assessing other cognitive functions beyond spatial memory.