ETD

• Artificial Intelligence
• Biological Signals
• Linear Mixed Models
• Meditation
• Signal Processing
• Statistical Modeling
• XGBoost

In this thesis work, we investigate how different styles of Tibetan Buddhist meditation shape the integrated dynamics of body and brain. Specifically, we study concentrative, loving-kindness, and emptiness practices in experienced monks using a multimodal psychophysiological approach. From cardiovascular, respiratory, electrodermal, and EEG recordings, we extract features describing autonomic parasympathetic/sympathetic (PNS/SNS) dynamics and central brain activity and connectivity and analyze them with linear mixed-effects models and explainable machine-learning classifiers.
Across all three practices, we observe a consistent pattern that can be described as “relaxed vigilance”: decreased heart rate, increased heart rate variability, slower and deeper breathing, and EEG signatures of internally oriented attention, indicating a calm but alert state. Emptiness meditation is associated with particularly strong vagal engagement, suggesting that even an analytically demanding practice can coexist with deep physiological relaxation in trained practitioners. At the cortical level, all meditations share elevated alpha–theta activity, while loving-kindness and emptiness additionally show enhanced gamma-band power compared to pure concentration, pointing to greater high-level cognitive-affective engagement. Large-scale functional connectivity remains broadly similar to eyes-closed rest, consistent with a refined, trained resting-state rather than a radically distinct network configuration.
Finally, we assess the feasibility of AI-based detection of meditation-related states. Using features summarizing autonomic regulation and EEG activity/connectivity, explainable boosted-tree models achieve above-chance discrimination between pre- vs post-meditation segments and between the three meditation types. For state detection, both autonomic and EEG features contribute, whereas meditation-type classification is driven mainly by frontal and central EEG patterns. Overall, this thesis provides an integrated framework that combines psychophysiology, brain connectivity, and explainable AI to characterize the signatures of advanced contemplative practice.