Concept and potential metabolic benefits of machine learning on extracorporeal technologies

Machine learning has experienced a revolutionary decade with advances across many disciplines. There has been enormous interest in applying machine learning and artificial intelligence to healthcare and, in particular, to data-rich environments like the intensive care unit [1]. In this context, we present our theoretical concept of a system based on artificial intelligence and on management algorithms designed to support operator choices during metabolic management in cardiopulmonary bypass. Intelligent systems and algorithmic supports are widely established and considered indispensable in the safety systems integrated into the extracorporeal consoles that regulate cardiopulmonary bypass and extracorporeal membrane oxygenation. They have prevented adverse events during the procedures performed in the last decade [2]. Although management of metabolism during cardiopulmonary bypass remains crucial, no systems and algorithms exist that are integrated to support the prevention of acute kidney injury or anaerobic metabolism. This situation results in wide variability of the management procedures carried out by the operators. For example, one can achieve 280 ml/min/m2 of indexed oxygen delivery in several ways: through an increase in the cardiac index or transfusion of blood products or ultrafiltration or diuresis [3]. These different pathways can have different clinical impacts. Integrating management algorithms in a metabolic monitoring system in relation to the analysis of the monitored variables could provide the most appropriate predictive path. The concept that we favour and propose should be able to provide a colour code to indicate the parameter to be improved (red), the secondary parameters to be improved (orange), and the final corrective solution (green).