EEG-Based Stress Assessment During Robot Assisted Surgery. Comparison of Statistical Methods with Machine Learning

Abstract

This study proposes a adaptive pipeline (feature extraction, feature selection, and classification) for the processing of electroencephalographic (EEG) signals to assess Mental Workload (MWL) during training sessions with a surgical robot. Unlike previous studies in this field, EEG signals were acquired by a dry eight-electrode device to maximize usability and comfort for the users. Fourteen medical students were EEG monitored while performing four different tasks on a simulation platform incorporated into the Da Vinci System. The proposed pipeline was compared with typical machine-learning approaches based on handcraft feature extraction approaches. The results confirmed the effectiveness of the proposed solution as a further contribution to EEG-based monitoring of robotic surgery for adaptive training.