Neurophysiological Effects of Sudarshan Kriya Revealed by Subject-Independent EEG Machine Learning

Abstract

Sudarshan Kriya Yoga (SKY) is a structured breathing-based meditation associated with improvements in stress regulation and mental health, yet its acute neurophysiological effects remain incompletely characterized using multivariate, subject-independent approaches. In this study, we investigated short-term pre-post neural modulation induced by a single long SKY session using electroencephalography (EEG) and traditional machine learning. EEG signals were pre-processed, and multiple feature representations were extracted, including raw EEG voltage statistics (mean, standard deviation), Short-Time Fourier Transform (STFT)-based spectral features, Discrete Wavelet Transform (DWT)-based multi-resolution features, and coherence-based functional connectivity measures. Classification was performed under a leakage-free Leave-One-Subject-Out Cross-Validation (LOSO-CV) framework to assess subject-level generalizability. Across feature types and classifiers, the intervention group consistently showed higher accuracy and more stable performance than the control group, indicating systematic SKY-related neural changes. STFT-based spectral features achieved the highest peak performance, with a Multilayer Perceptron reaching [~]89% accuracy, followed by DWT and coherence-based features, which also exhibited strong discriminative capacity. Raw EEG voltage features yielded lower performance, suggesting limited specificity. Control group classification remained close to the 0.50 baseline expected under random classification across all feature types, supporting the specificity of intervention effects. These results demonstrate that acute SKY practice induces measurable neural modulation most robustly captured in frequency-domain and network-level EEG representations. The combination of subject-independent validation and comparative feature analysis provides a rigorous, data-driven framework for characterizing meditation-related brain dynamics and advancing the objective quantification of breathing-based interventions.