This project is a collaboration between myself, the division of Perioperative medicine and intensive care (PMI) at Karolinska University Hospital and Getinge. The aim is to use machine learning (ML) techniques to predict the onset of hypotension during surgery.
Approximately 800,000 patients are operated in Sweden each year. According to large international cohort studies, the risk for postoperative complications is high, which affects both short and long-term outcomes. Intraoperative low blood pressure (hypotension) has been associated with adverse outcomes, mainly renal and myocardial injury and even death. Hypotension is common during large surgery. By identifying and treating hypotensive periods early, preferably even before onset, the amount of hypotension and thus the risk of damage to the heart muscle and kidney are reduced. However, it is very difficult today to predict hypotension a head of time.
Data-driven healthcare, in the form of continuous monitoring in the perioperative period with both invasive and non-invasive sensors, advanced signal processing, and machine learning, has proven to be able to predict hypotension. This project is focusing on predicting near-term onset of hypotension using artificial intelligence (AI) techniques based on deep learning on waveform data of physiological signals.
- Martin Jacobsson, Navid Zandpour, Thorir Sigmundsson, Håkan Björne, Caroline Hällsjö Sander, "Deep Learning-Based Early Prediction of Intraoperative Hypotension", Poster presentation in the 43rd Annual International Conference of the IEEE Engineering in Medicine and Biology Society (IEEE EMBC'21), November 1-5, 2021.
- Navid Zandpour, "Deep Learning for Prediction of Falling Blood Pressure During Surgery: Prediction of Falling Blood Pressure", Master Thesis, KTH, 2022.
- Clara Escorihuela Altaba, "Machine Learning personalizationfor hypotension prediction", Master Thesis, KTH, 2022.
This work is partly funded by the Swedish Governmental Agency for Innovation Systems and the Swedish Engergy Agency.