Prediction of hypnotic trance with brain-evoked responses to an auditory oddball using magnetoencephalography.

Hypnosis alters consciousness and is used for pain management, but assessing hypnotic trance relies on subjective signs. Using magnetoencephalography (MEG) and an auditory oddball paradigm, brain signals were recorded from 20 healthy subjects during critical consciousness, hypnotic trance, and distraction. Feature extraction and classification models were tested; EEGNet performed best, achieving 70% and 84% ROC-AUC in distinguishing hypnotic trance from critical consciousness and distraction, respectively, with predictions every 4 seconds after a 19-minute training session. This method offers real-time, objective assessment of hypnotic trance and could be adapted for clinical use with electroencephalography (EEG).

Study at a glance

Abstract

Hypnosis is widely used in pain management. This technique involves the induction of a hypnotic trance, which is a state of consciousness distinct from normal waking state (critical consciousness). In this altered state, individuals become more receptive to the hypnotherapist's suggestions, which can influence their perception, especially of pain. However, hypnotic trance is a fluctuating state that is diagnosed by signs that are not pathognomonic. To address the issue of subjective assessment, we propose a method to accurately and in real time predict an individual's state of consciousness using brain signals.The auditory oddball paradigm elicits brain responses, which have been shown to be modulated by attention and under hypnosis. To investigate the influence of state of consciousness on these responses, we used an auditory oddball paradigm and magnetoencephalography (MEG) to record signals from 20 healthy subjects in three conditions: critical consciousness (CC), hypnotic trance (HYP) and distraction (DIS). We then performed feature extraction using several models (xDAWN, CSP and PLS) followed by classification using common methods reported in the literature: EEGNet, DCPM, LDA, sKLDA, QDA and GBoost.The classifier that performed best with this dataset was EEGNet, with a training session of 19 minutes including data from critical consciousness, distracted state and hypnotic trance. Considering our auditory stimulation rate, the model allowed a prediction every 4 seconds with a performance of 70 and 84, respectively, in separating hypnotic trance from critical consciousness and distraction (metric: ROC-AUC).This method shows promising potential for the rapid assessment of hypnotic trance by measuring and processing brain signals with MEG. Further improvements to models and training paradigms could make this method suitable for clinical practice using electroencephalography (EEG).