An Overview on Cognitive Stress Assessment using Deep Learning and EEG: Perspectives on Yogic Interventions

Excessive stress brings serious risks to both psychological and physiological wellbeing and affects individuals across all ages and demographics. Conventional methods of evaluating stress generally include collecting self-reported questionnaires and observing behavior, and thus are subjective and usually not consistent between persons. Hence, this study focuses on the objective physiological methods for stress evaluation related to these limitations. Specifically, we study the incorporation of electroencephalogram (EEG) signals in deep learning for the purpose of identifying and classifying brain activity associated with stress. Mood detection through EEG allows for stress characterization through a variety of analytical frameworks, such as event-related potentials, power asymmetry, time-frequency analysis, and spectral analysis. Recent advances in hybrid deep learning architectures made it possible to combine convolutional neural networks and recurrent neural networks, which further improves the accuracy and robustness of the stress classification model. The present work also presents evident changes in the neurophysiological correlates of improved mental states when the Kaiya yoga is introduced as an adjunctive intervention in reducing stress. It found observable changes in alpha and theta wave activity which corresponded to an ameliorated state of both stress and emotional regulation. Bridging neuroscience with machine learning and yogic practices, intelligent real-time EEG-based stress monitoring and intervention systems shall be developed through this study. Stress-related disorders management and holistic mental healthcare advancement have promising applications through such systems.