Machine learning predicts dangerously low blood pressure during surgery
A machine learning algorithm is able to predict potentially dangerous low blood pressure that can occur during surgery by detecting subtle signs in routinely collected physiological data in surgical patients.
Dangerously low blood pressure—hypotension—can lead to complications such as postoperative heart attack, acute kidney injury and even death. The algorithm was able to accurately predict an intraoperative hypotensive event 15 minutes before it occurred in 84 percent of cases, 10 minutes before in 84 percent of cases, and five minutes before in 87 percent of cases.
Researchers leveraged two datasets to build and validate the predictive algorithm, based on recordings of the increase and decrease of blood pressure in the arteries during a heartbeat—including episodes of hypotension. For each heartbeat, they were able to derive 3,022 individual features from the arterial pressure waveforms, producing more than 2.6 million bits of information used to build the algorithm.
In a new study published this week in the journal Anesthesiology, machine learning was able to identify which of these individual features—when they happen together and at the same time—predict hypotension and could potentially reduce the risk of harm to patients.
“The results demonstrate that a machine learning algorithm can be trained, with large data sets of high-fidelity arterial waveforms, to predict hypotension in surgical patients’ records,” conclude the authors.
“It is the first time machine learning and computer science techniques have been applied to complex physiological signals obtained during surgery,” according to lead researcher Maxime Cannesson, MD, vice chair for perioperative medicine and professor of anesthesiology at UCLA Medical Center. “Physicians haven't had a way to predict hypotension during surgery, so they have to be reactive and treat it immediately without any prior warning. Being able to predict hypotension would allow physicians to be proactive instead of reactive.”
While future studies are needed to evaluate the real-time value of such algorithms in a broader set of clinical conditions and patients, Cannesson contends that the research “opens the door to the application of these techniques to many other physiological signals, such as EKG for cardiac arrhythmia prediction or EEG for brain function” and “could lead to a whole new field of investigation in clinical and physiological sciences and reshape our understanding of human physiology.”
In March, the U.S. Food and Drug Administration granted a De Novo request to Edwards Lifesciences for its Acumen Hypotension Prediction Index software that uses the new algorithm. Cannesson is a consultant to the company.
“The Acumen Hypotension Prediction Index is a next-generation, predictive monitoring software that represents the future of how we can use patient data to improve healthcare,” said Catherine Szyman, corporate vice president of critical care at Edwards Lifesciences. “HPI improves patient care by providing intelligent decision support, moving beyond descriptive monitoring to facilitate predictive monitoring, which provides more clarity on a patient's condition to enable clinicians to make more proactive decisions.”