Hypnosis Measuring: The Pilot Study of Machine Learning Approach for Instrumental Control of the Trance Dynamics in Patients with Anxiety and Depressive Disorders
The therapeutic effects of hypnosis in some cases seem to be most marked when the patient has achieved sufficient hypnotic depth. It could be possible to monitor the deepening process using electrophysiological data to obtain information on depth changes throughout the session. However, although hypnosis is characterized by some common EEG patterns, significant differences between subjects are also observed. Therefore, an individualized approach is required to quantify the depth continuously during a session. To achieve this, we proposed the machine learning approach, using an EEG-based Brain-Computer interface, and tested it on video-EEG recordings of 8 outpatients. Based on the data from the first sessions, we trained the classification models to discriminate between conditions of wakefulness and deep hypnosis. Then, we applied them to subsequent sessions to predict the probability of deep hypnosis, i.e., to continuously measure depth level in real time. The models trained using frequency ranges of 1.5-14 and 4-15 Hz provided high accuracy. The applications and perspectives are discussed.