Decoding Depth of Meditation: Neurophysiological Insights from Expert Vipassana Practitioners
Poster No:
1640
Submission Type:
Abstract Submission
Authors:
Nicco Reggente1, Tracy Brandmeyer1, Ninette Simonian1, Christian Kothe2
Institutions:
1Institute for Advanced Consciousness Studies, Santa Monica, CA, 2Intheon, Inc., San Diego, CA
First Author:
Co-Author(s):
Introduction:
This project's overarching objective is to develop a personalized, multivariate neurofeedback tool to assist meditation, addressing the limitations of generic models in the face of neurodiversity and compensatory mechanisms (Brandmeyer & Reggente, 2024). In pursuit of this goal, we first conducted an EEG study with expert Vipassana practitioners (n=40), leveraging the standardized nature of Vipassana and the experts' ability to discern subtle gradations of meditative depth. Recognizing the challenges posed by self-reports (e.g., the "observer phenomenon" where reporting can disrupt the meditative state), we used these expert-defined self-reports as a gold standard. Across two sessions (held 2 weeks apart), we leveraged machine learning (deep neural nets) to predict meditative depth across sessions, aiming to identify implicit neural markers that track meditative depth. We also included a passive auditory oddball task to identify if oddball ERPs varied as a function of depth. The goal of this current project is to refine these implicit, predictive markers, facilitating non-intrusive tracking of meditation in future studies without the need for explicit self-reports.
Methods:
Participants
Expert Vipassana practitioners were recruited based on the following criteria: a minimum of 5 years of consistent practice, participation in at least one 3-day silent retreat, and a report of meditating at least 20 minutes for a minimum of 5 days per week.
Procedure
Upon arrival, participants completed pre-experiment questionnaires. They were then fitted with a 64-channel EEG and peripheral monitoring devices to record heart rate, respiration rate, EMG, and GSR. Participants underwent two 35-minute meditation blocks (additional blocks and tasks beyond the scope of this report were also conducted). Post-meditation, participants filled out subsequent questionnaires. The entire process was repeated in a follow-up session 2 weeks later. During the four meditation blocks, participants were either exposed to a passive auditory oddball tone sequence they were instructed to ignore. Meditation blocks were characterized by one of two types:
Probed Trials: Participants were instructed to report their meditative depth on a scale of 1-5 using a finger-attached presentation clicker in response to an auditory probe presented approximately every 4 minutes.
Emergence Trials: Participants reported their meditative depth organically, only when they perceived a change from a deeper state since their last report.
Data Analysis
Event-related potential (ERP) analyses were conducted to ascertain changes in response to the oddball tone contingent on meditative depth. Deep neural networks employing FIR-type filter kernels (Bouallegue et al., 2020) were applied to the EEG time series between reporting periods across all channels. Two cross-validation methods were adopted:
Eight-fold Cross-validation: This was executed across participants, amalgamating data from all sessions and participants.
Leave-one-session-out Cross-validation: This was conducted within individual participants.
Predictive models aimed to discern meditation depth at either all 5 levels or as a binary classification (shallow [1,2] vs. deep [4,5]). Performance metrics included mean square error for the 1-5 classification and area under the curve (AUC) for the binary classification.
Expert Vipassana practitioners were recruited based on the following criteria: a minimum of 5 years of consistent practice, participation in at least one 3-day silent retreat, and a report of meditating at least 20 minutes for a minimum of 5 days per week.
Procedure
Upon arrival, participants completed pre-experiment questionnaires. They were then fitted with a 64-channel EEG and peripheral monitoring devices to record heart rate, respiration rate, EMG, and GSR. Participants underwent two 35-minute meditation blocks (additional blocks and tasks beyond the scope of this report were also conducted). Post-meditation, participants filled out subsequent questionnaires. The entire process was repeated in a follow-up session 2 weeks later. During the four meditation blocks, participants were either exposed to a passive auditory oddball tone sequence they were instructed to ignore. Meditation blocks were characterized by one of two types:
Probed Trials: Participants were instructed to report their meditative depth on a scale of 1-5 using a finger-attached presentation clicker in response to an auditory probe presented approximately every 4 minutes.
Emergence Trials: Participants reported their meditative depth organically, only when they perceived a change from a deeper state since their last report.
Data Analysis
Event-related potential (ERP) analyses were conducted to ascertain changes in response to the oddball tone contingent on meditative depth. Deep neural networks employing FIR-type filter kernels (Bouallegue et al., 2020) were applied to the EEG time series between reporting periods across all channels. Two cross-validation methods were adopted:
Eight-fold Cross-validation: This was executed across participants, amalgamating data from all sessions and participants.
Leave-one-session-out Cross-validation: This was conducted within individual participants.
Predictive models aimed to discern meditation depth at either all 5 levels or as a binary classification (shallow [1,2] vs. deep [4,5]). Performance metrics included mean square error for the 1-5 classification and area under the curve (AUC) for the binary classification.
·Figure 1. Study Design
Results:
Significant ERP differences were noted by depth, with increased frontal midline theta and parietal beta values. ERP scaled linearly with depth, predicting it in meditation with 1.2 MAE (5-way) and .81 AUC (deep vs. shallow). Cross-validation showed similar results (1.1 MAE, .76 AUC). A real-time metric, ranging 0-1 at 33Hz, measured the qualia distance to an ideal meditation state.
·Figure 2. Oddball ERP as a function of reported depth (rows) across the frequency spectra (columns).
Conclusions:
Expert meditators show amplified ERPs to oddballs as a function of meditative depth. Meditatiive depth can also be implicitly decoded in ways that promote personalized neurofeedback systems.
Modeling and Analysis Methods:
Classification and Predictive Modeling
EEG/MEG Modeling and Analysis 1
Novel Imaging Acquisition Methods:
EEG
Perception, Attention and Motor Behavior:
Attention: Auditory/Tactile/Motor 2
Keywords:
Consciousness
Electroencephaolography (EEG)
Other - meditation, attention, ERP, neurofeedback
1|2Indicates the priority used for review
Provide references using author date format
Bouallegue, G., Djemal, R., Alshebeili, S. A., and Aldhalaan, H. (2020). A Dynamic Filtering DF-RNN Deep-Learning-Based Approach for EEG-Based Neurological Disorders Diagnosis. IEEE Access 8, 206992–207007. doi: 10.1109/ACCESS.2020.3037995.
Debener, S., Kranczioch, C., Herrmann, C. S., and Engel, A. K. (2002). Auditory novelty oddball allows reliable distinction of top–down and bottom–up processes of attention. International Journal of Psychophysiology 46, 77–84. doi: 10.1016/S0167-8760(02)00072-7.