Multi-Sensor Feature Integration for Assessment of Endotracheal Intubation

Abstract

Traditionally, proficiency in endotracheal intubation (ETI) has been assessed by human supervisors in a subjective manner during training sessions; however, recent advances in sensor and computing technology have made it possible to obtain objective measures to evaluate the practitioner's performance. This study presents an automated and objective ETI assessment system based on multi-sensor integration which aims at discriminating experienced from novice providers accurately. To this end, four different types of sensors were used to collect data, including hand motion of the provider, and tongue force, incisor force and head angle of the training mannequin. Features were extracted from the datasets, and relevant ones were identified by applying feature selection algorithms to create individual and integrated feature sets. An artificial neural network-based classification model was developed for each feature set. The results show that a classifier based on a small number of integrated features achieves the best accuracy (96.4%), significantly higher than the best obtained by any individual feature sets (91.17% by hand motion). This study demonstrated the feasibility of a multi-sensor based ETI assessment system that can provide practitioners with objective and timely feedback about their performance.