Combining spatial and spectral alignment of EEG signals improves the classification of meditation versus rest states in new subjects without retraining. The Riemannian Space Data Alignment (RSDA) method adjusts brain activity patterns across electrodes, while Convolutional Monge Mapping Normalization (CMMN) aligns brain rhythms across frequencies. Together, they raised leave-one-subject-out classification accuracy to 66.6%, compared to 55.7% for non-aligned data and 59.6% for z-score normalization alone. Theta, Alpha, and Beta frequency bands contributed consistently, and Frontopolar and Temporal brain regions were key for distinguishing mental states. This approach offers a practical step toward calibration-free neurofeedback systems.
Decoding meditation states from EEG signals across different people remains challenging due to individual brain differences. This work compares feature engineering methods with a deep learning approach (EEGNet) for cross-subject mindfulness meditation decoding, using EEG recordings from novice and expert meditators. Riemannian Space Data Alignment (RSDA) was applied session-wise and per subject to reduce variability. EEGNet applied to RSDA-aligned signals achieved the highest decoding performance with the shortest time segments, effectively extracting relevant features for meditation state classification. These findings suggest that EEGNet can enable more effective, calibration-free neurofeedback applications for meditation.
Alpha power can be trained up or down using neurofeedback combined with focused-attention meditation, but up-regulation is harder to achieve than down-regulation in a single session. In a within-subject experiment with 31 novice practitioners (25 women, mean age 23.16), participants attempted to increase or decrease global alpha power while focusing attention above the crown of the head and receiving auditory feedback. Alpha power was overall higher during up-regulation than down-regulation trials, but this difference came mainly from successful reduction of alpha during down-regulation; up-regulation did not significantly increase alpha. Training effects did not persist during a post-training resting-state recording, suggesting that more sessions are needed for lasting change.
Expert meditators show a shift from reactive to proactive control over the connection between brain and body compared to novices. During breath-focused meditation, both groups slowed their breathing, but experts had the lowest rates and higher parasympathetic (rest-and-digest) tone. Information-flow analysis showed that novices' heart and breathing signals more strongly influenced brain alpha activity, whereas experts showed stronger top-down control from the brain to breathing, especially in frontal brain regions. Experts also showed a reduced ratio between alpha brain waves and heart rate during meditation. These results suggest that meditation training changes how the brain and body interact, consistent with predictive processing theories of interoception.