Development of an automated detection algorithm to improve efficiency of Operating Room Black Box analyses

点击翻译按钮获取中文摘要

Operating rooms present one of the most complex, high-risk situations in clinical settings, and the Operating Room Black Boxsystem seeks to enable the analysis of surgical procedures to improve patient safety. The Black Box is a device that recordsaudio, video, environmental, and biometric data within the operating room during a surgical case. This information can be used forsurgical safety improvement. For example, retrospectively, analyzing this data can enable researchers to understand the origin ofan adverse event. Prospectively, it can allow for the identification of intraoperative factors that potentially risk or enhance patientsafety. Currently, deriving such insight from the Black Box data is labour intensive, requiring significant amount of time and manualeffort in reviewing recordings. Thus, we aim to improve the efficiency and effectiveness of Black Box analysis by automating thedetection of interested segments within Black Box recordings, specifically episodic uncertainty. Episodic uncertainty is a stateexperienced by individuals or teams during abnormal situations that may lead to adverse events if not managed correctly.Research will first be conducted to uncover the association between recorded data and observed instances of episodicuncertainties. Machine learning techniques informed by the correlation will be developed to process Black Box data and identifyinstances of episodic uncertainty. Such detection capability in Black Box will decrease the resource demands associated with dataanalysis, expand its availability to more hospitals, expedite the identification of uncertainty during patient care, and enable rapidimprovements and higher quality management in patient safety.