Artificial intelligence models to optimize decannulation and early detection of complications in patients on ECMO therapy: a scoping review

This project is a scoping review that aims to systematically map the scientific evidence on the use of artificial intelligence (AI) models to support clinical decision-making in patients receiving extracorporeal membrane oxygenation (ECMO) therapy. Specifically, it will identify and describe studies that apply AI techniques (such as machine learning, deep learning, and neural networks) to hemodynamic and related clinical variables on venovenous (VV) or venoarterial (VA) ECMO, with the purpose of optimizing the timing of decannulation and enabling early detection of complications. The review will describe the types of AI models used, the variables incorporated, the ECMO modality and clinical settings in which they are applied, and the main outcomes reported, including predictive performance for mortality, successful weaning/decannulation, and complications. By synthesizing these findings, the study seeks to outline the current state of the art, highlight potential benefits and limitations of AI-based tools in ECMO care, and identify gaps and future directions for research and clinical implementation.