Electroencephalographic neural correlates and deep learning analysis of a single brief focused mindfulness meditation in young adults: A pilot study.

Abstract

This study builds on brief focused mindfulness meditation (BFMM) to examine its associations with physiological indices and electroencephalographic (EEG) neural features related to stress in young adults. In addition, deep learning models are employed to identify complex, nonlinear patterns in EEG signals during BFMM, aiming to determine the most effective classification model. Twenty-nine participants (n=29) were enrolled in a before-and-after study of the same cohort. Participants underwent a 10-min resting state, then were instructed to perform BFMM for 10 min. Physiological indices were recorded pre- and post-BFMM, while EEG signals were captured during both the resting and BFMM states. Deep learning techniques including multi-layer perceptron (MLP), long short-term memory (LSTM), convolutional neural network (CNN), and ensemble models were subsequently employed to classify EEG signals. Final Twenty-four participants (n=24) were included in the analysis. The differences in both heart rate (t = 4.22, p < 0.001) and respiratory rate (t = 5.05, p < 0.001) were significant between pre- and post-BFMM levels. Results showed significant power spectral density differences between the resting and BFMM states in the theta (Z = 3.17, q = 0.039, beta (Z = 3.17, q = 0.049) bands of the right frontotemporal region (T8) and the theta (t = 3.41, q = 0.039) band of the right frontocentral region (FC6). In addition, the ensemble model (MLP+LSTM+CNN) outperformed other methods, achieving an accuracy of 79.0% in classifying EEG signals. These finding suggested that a single session of BFMM may regulate the autonomic nervous system and modulate neural activity. The proposed ensemble model shows promise in distinguishing BFMM from resting-state EEG, providing a foundation for future EEG-based assessment of mindfulness meditation.