Using Machine Learning and Lung Ultrasound to Predict Postoperative Pulmonary Complications in Adults Undergoing Elective Surgery with General Anesthesia

What this Trial Is About

Postoperative pulmonary complications POPC are common after general anesthesia and cause increased illness and death in surgical patients. This research aims to improve early detection of POPC by developing a machine learning algorithm that uses standard clinical data and lung ultrasound images collected after surgery. The study will include 512 adults undergoing planned surgery with general anesthesia to better predict who is at risk of developing these complications before symptoms appear. All patients will have their routine clinical data collected during surgery and will receive a standardized lung ultrasound in the recovery room. Follow-up clinical exams will occur on postoperative days 1, 3, and 7 to check for signs of POPC using standardized criteria. The collected data, including ultrasound images, will be used to train and evaluate a neural network-based machine learning model to predict postoperative lung complications. Participants will be monitored through clinical exams and lung sonography after surgery, with data stored securely for analysis. Researchers will measure the number of patients developing POPC by day 7 or discharge. The models predictive accuracy will be compared to existing risk scores. This study may help improve patient outcomes by enabling earlier detection and targeted prevention of lung complications after surgery.

PrEventing PostoPERative Pulmonary Complications by Establishing a MachINe-learning assisTed Approach