Biosignal Data Collection Platform Design and Deep Learning-based Data Analytics for Intensive Care Units

Abstract

Patients in intensive care unit (ICU) have high disease severity and mortality. In addition, the demand for ICUs is expected to increase significantly in an aging society. Therefore, there is an urgent need for research to improve the quality of care for ICU patients. However, owing to the lack of medical personnel in ICUs and financial problems in operation, there is a lack of a system for the real time monitoring of critically ill patients as well as an emergency situation prediction system: accordingly, digital solutions are required to provide care to critically ill patients. In this study, we built a platform that can automatically collect the biosignal data in near real time from multiple patient monitors in the ICUs of the Ajou University Hospital. Data collection methods of patient monitors were classified into three types (Nihon Kohden, Philips, and GE) according to the data export protocols. Therefore, three biosignal collection systems were developed and a unified database was constructed. We also developed a high accuracy ECG (electrocardiogram) delineator (QT detector) through deep learning for single-lead ECG data collected using this platform and verified the utility of this platform by applying it to an index for the weaning outcome. A platform that can automatically collect the biosignal data from approximately 200 patient monitors in near real time was established, and a database of approximately 30,000 critically ill patients was constructed. The U-Net-based ECG delineator developed in this study exhibited high performance by recording a score of 11.6, and detrended fluctuation analysis revealed that the QTc time series of the weaning failure group demonstrated higher complexity. The biosignal data collection platform provides an important data source for monitoring critically ill patients and for performing clinical research to improve the prognosis of ICU patients: it is expected that the platform can be used as a tool to discover medical knowledge related to biosignals that has not yet been revealed.