EEG-based investigation of effects of mindfulness meditation training on state and trait by deep learning and traditional machine learning
Frontiers in Human Neuroscience August 31, 2023 Baoxiang Shang, Feiyan Duan, Ruiqi Fu et al. 28 citations
Short-term mindfulness-based stress reduction (MBSR) training produces electroencephalogram (EEG)-detectable state and trait effects. Using convolutional neural networks (deep learning) and support vector machines (SVM) with features from common spatial patterns, classifiers were trained on EEG from 11 novice MBSR practitioners during rest and meditation at early and late training stages. Shallow ConvNet classifiers achieved mix-subject and intra-subject accuracies superior to prior studies for both novice and expert meditators across meditation types. FBCSP+SVM classifiers gave inter-subject accuracies of 68.50% for state effect recognition, 85.00% for trait effect recognition using meditation EEG, and 78.96% using resting EEG. Deep learning outperforms for state effects in novices and is comparable for experts.