Specific feature selection in wearable EEG-based transducers for monitoring high cognitive load in neurosurgeons

Abstract

The electroencephalographic (EEG) features for discriminating high and low cognitive load associated with fine motor activity in neurosurgeons were identified by combining wearable transducers and Machine Learning (ML). To date, in the literature, the specific impact of fine-motor tasks on surgeons’ cognitive load is poorly investigated and studies rely on the EEG features selected for cognitive load induced by other types of tasks (driving and flight contexts). In this study, the specific EEG features for detecting cognitive load associated with fine motor activity in neurosurgeons are investigated. Six neurosurgeons were EEG monitored by means of an eight-dry-channel EEG transducer during the execution of a standardized test of fine motricity assessment. The most informative EEG features of the cognitive load induced by fine motor activity were identified by exploiting the algorithm Sequential Feature Selector. In particular, five ML classifiers maximized their classification accuracy having as input the relative alpha power in Fz, O1, and O2, computed on 2-s epochs with an overlap of 50 %. These results demonstrate the feasibility of ML-supported wearable EEG solutions for monitoring persistent high cognitive load over time and alerting healthcare management.