Development of Classification Models for Assessment of Endotracheal Intubation Training by a Cyber-Physical System

Abstract Endotracheal intubation (ETI) is one of the most staple skills in prehospital medicine that is performed to prevent suffocation of an unconscious person. To evaluate ETI skills of medical practitioners, an effective and reliable assessment system is required; however, the current assessment method relies on subjective evaluation by supervisors during training sessions that may be inaccurate and biased. To provide objective and immediate feedback to trainees, this paper proposes a cyber physical system (CPS)-based ETI assessment system. The CPS is composed of wearable gloves with embedded sensors to capture hand motion data and software to discriminate between two groups: experienced and novice. To this end, we collected hand motion data from the two groups and extracted 18 features. Furthermore, we identified reduced sets of 8 and 10 based on their statistical significance. To discriminate the two groups, artificial neural network-based classification models were developed with the three feature sets. Experimental results show that the classifiers with 18, 8, and 10 features achieved an accuracy of 90.94%, 87.87% and 89.60%, respectively. This work corroborates that wearable gloves with embedded motion sensors can be effective in assisting self-training of ETI on a CPS.