Clinical applications of artificial intelligence and machine learning in the modern cardiac intensive care unit

Abstract

The depth and breadth of data produced in the modern cardiac intensive care unit (CICU) poses challenges to clinicians and researchers. Artificial intelligence (AI) and machine learning (ML) methodologies have been increasingly used to provide insights into this complex patient population. Major analytical tasks where ML methodology can be applied in the CICU and other critical care settings include mortality risk stratification, prognostication, non-fatal event prediction, diagnosis, phenotyping, identification of occult heart disease from the electrocardiogram and interpretation of echocardiographic images. In this review, we will discuss existing and future applications of different ML methods for CICU and other critical care populations, including penalized regression, standard ML methods (e.g., tree-based and other non-linear approaches) and advanced ML methods (e.g., deep learning and neural networks). While comparatively few published studies have applied ML methods in CICU populations, a more robust literature including patients with acute cardiovascular disease and non-cardiovascular critical illness can provide insights into CICU care. The CICU of the future is likely to utilize a sophisticated array of ML algorithms to streamline patient care by facilitating early recognition, diagnosis, phenotyping, and intervention for critically ill or deteriorating patients to improve providers’ cognitive load.