Effect of Heartfulness Meditation Among Long-Term, Short-Term and Non-meditators on Prefrontal Cortex Activity of Brain Using Machine Learning Classification: A Cross-Sectional Study

A feasibility study examined whether electroencephalogram (EEG) connectivity patterns can distinguish long-term Heartfulness meditation practitioners, short-term practitioners, and non-meditators. EEG data from 34 participants were analyzed using functional connectivity parameters as features for machine learning classifiers. When classifying long-term meditators versus non-meditators, model accuracy ranged from 84% to 100%; for short-term meditators versus non-meditators, accuracy ranged from 80% to 93%. Decision trees, support vector machines, k-nearest neighbors, and ensemble classifiers outperformed linear discriminant analysis and logistic regression. The results suggest that machine learning applied to EEG functional connectivity may serve as a marker for meditation proficiency.

Study at a glance

Abstract

Background Meditation is a mental practice with health benefits and may increase activity in the prefrontal cortex of the brain. Heartfulness meditation (HM) is a modified form of rajyoga meditation supported by a unique feature called "yogic transmission." This feasibility study aimed to explore the effect of HM on electroencephalogram (EEG) connectivity parameters of long-term meditators (LTM), short-term meditators (STM), and non-meditators (NM) with an application of machine learning models and determining classifier methods that can effectively discriminate between the groups. Materials and methods EEG data were collected from 34 participants. The functional connectivity parameters, correlation coefficient, clustering coefficient, shortest path, and phase locking value were utilized as a feature vector for classification. To evaluate the various states of HM practice, the categorization was done between (LTM, NM) and (STM, NM) using a multitude of machine learning classifiers. Results The classifier's performances were evaluated based on accuracy using 10-fold cross-validation. The results showed that the accuracy of machine learning models ranges from 84% to 100% while classifying LTM and NM, and accuracy from 80% to 93% while classifying STM and NM. It was found that decision trees, support vector machines, k-nearest neighbors, and ensemble classifiers performed better than linear discriminant analysis and logistic regression. Conclusion This is the first study to our knowledge employing machine learning for the classification among HM meditators and NM The results indicated that machine learning classifiers with EEG functional connectivity as a feature vector could be a viable marker for accessing meditation ability.