Predicting Spontaneous Breathing Trial Outcomes via Deep Learning

Mechanical ventilation is a life-saving intervention in critical care, but prolonged use can lead to complications. Successful liberation from mechanical ventilation, often assessed through a Spontaneous Breathing Trial (SBT), is a crucial step in patient recovery. The SBT aims to evaluate a patient's readiness to breathe independently; however, predicting its outcome remains a clinical challenge, often leading to delayed extubation or failed trials with associated risks. This research investigates the application of Artificial Intelligence (AI), specifically deep learning methodologies, to predict the outcomes of SBTs. Leveraging a unique dataset derived from electronic health records and physiological monitoring of patients undergoing mechanical ventilation at the University Hospital of Padua, we developed and evaluated deep learning models to forecast SBT success or failure. This work, a product of a research training collaboration with clinicians at the Padua University Hospital, explores the potential of AI to provide clinicians with a valuable tool for timely and informed decision-making in the mechanical weaning process, ultimately aiming to improve patient outcomes and resource utilization in the intensive care setting.