Common spatial pattern for classification of loving kindness meditation EEG for single and multiple sessions

Area of Science:

• Neuroscience
• Cognitive Science
• Biomedical Engineering

Background:

• Limited research exists on classifying multi-session electroencephalography (EEG) data for loving kindness meditation (LKM).
• Developing algorithms to support meditation practices requires robust classification of meditation states from EEG data.

Purpose of the Study:

• To classify single and multiple session EEG data for LKM and non-meditation states.
• To explore the utility of Common Spatial Patterns (CSP) for feature extraction in meditation EEG analysis.
• To compare classification accuracies across different meditation and resting states.

Main Methods:

• Collected EEG data from 32 participants across four conditions: Pre-Resting, Post-Resting, LKM-Self, and LKM-Others.
• Applied Common Spatial Patterns (CSP) for feature extraction.
• Utilized Linear Discriminant Analysis (LDA) for classification of meditation/non-meditation instances.

Main Results:

• Achieved 99.5% accuracy in classifying single-session meditation/Pre-Resting EEG data.
• Obtained 83.6% accuracy for multi-session (five sessions) meditation/Pre-Resting EEG data.
• Pre-Resting data showed distinct features, leading to higher classification accuracy compared to other mind tasks.

Conclusions:

• Demonstrated the feasibility of classifying meditation states from EEG data, even across multiple sessions.
• The application of CSP in meditation EEG analysis opens new avenues for research.
• Findings support the development of objective measures for meditation practice and its effects.