Machine Learning based Classification of Meditators using Functional Connectivity over Resting State Networks

Abstract: Meditation has several health benefits and is also used as a complementary treatment for various ailments. Neuroimaging studies have shed light on the effects of meditation, especially on the brain. Functional Magnetic Resonance Imaging, a powerful non-invasive imaging technique is used in this study to determine the functional connectivity in meditator’s brain. In this study, long-term effects of Rajayoga Meditation practice were considered where the difference in functional connectivity between two groups of subjects one with long duration and the other with short duration of Rajayoga meditation practice was found. Two groups of subjects with long-term and short-term practice of Rajayoga meditation were recruited. Task-based fMRI was acquired as the subjects performed a Neurocognitive task. Functional connectivity among the regions of Resting-State Networks was performed and four functional connectivity metrics were derived. Machine learning algorithms were used to classify these two groups based on functional connectivity metrics used as features. The ensemble learning algorithms Random Forest and Gradient Boosted Tree could differentiate Long-term and Short-term Rajayoga Practitioners with an accuracy of 62% when all four Functional Connectivity metrics were used as features.