Real-Time Cognitive Workload States Recognition from Ultra Short-Term ECG Signals on Trainee Surgeons Using 1D Convolutional Neural Networks

Abstract

Surgery is a mentally demanding task that is focused on patient safety and requires the precise execution of motor control and decision making in a timely manner. Episodes of high Cognitive Workload (CWL) induced by stressors or distractions have been shown to lead to inferior performance potentially compromising patient safety [1]. We have proposed a promising CWL assess- ment platform utilising a wide range of physiological sensors [2]. However, there are some disadvantages associated with a complex multimodal sensing design, including high device cost, long set up time and the dis- comfort caused by wearing multiple wearable sensors for long periods during surgery. To address this problem, the proposed one-dimensional convolutional neural network (1D-CNN) model discussed here, offers an alternative solution to recognising CWL states, achieving satisfac- tory performance (91.3% accuracy) with the use of a wireless ECG sensor alone, showing great potential for widespread deployment in the operating room (OR). MATERIALS AND METHODS