Multi-dimensional conditional mutual information with application on the EEG signal analysis for spatial cognitive...

Area of Science:

• Neuroscience
• Cognitive Science
• Signal Processing

Background:

• Evaluating spatial cognitive ability is crucial for understanding brain function.
• Traditional methods may lack objectivity and quantitative accuracy.
• Virtual reality (VR) environments offer immersive platforms for cognitive assessment.

Purpose of the Study:

• To develop an effective method for evaluating spatial cognitive ability using EEG signals.
• To extract and classify EEG signal features within a VR environment.
• To quantitatively assess spatial cognition training effects for improved accuracy.

Main Methods:

• Proposed a multi-dimensional conditional mutual information (MCMI) method to calculate channel coupling strength.
• Transformed multi-frequency band combinations into multi-spectral images.
• Classified image data using a convolutional neural networks (CNN) model.

Main Results:

• MCMI-based multi-spectral image features outperformed other methods in classification.
• Achieved a peak classification accuracy of 98.3% for the Beta1-Beta2-Gamma frequency band combination.
• Identified MCMI characteristics on the Beta1-Beta2-Gamma band as a potential biological marker.

Conclusions:

• The proposed MCMI feature extraction method offers a novel approach for spatial cognitive ability assessment.
• MCMI provides a quantitative and objective means to evaluate cognitive training effects.
• EEG signal analysis using MCMI holds promise for advancing cognitive neuroscience research.