Machine learning to predict benefits from cardiac resynchronization therapy

Dr. Matthew Kalscheur
Dr. Matthew Kalscheur

Cardiac resynchronization therapy (CRT) involves using a special pacemaker to regulate the rhythm of both lower chambers (ventricles) of the heart in people with heart failure whose left and right ventricles are not pumping in sync. While CRT can be helpful, about 30 percent of patients who meet the criteria for CRT do not end up benefitting from it, and at present it's unclear why.

To better predict which patients will be helped by CRT, researchers at UW-Madison used an artificial intelligence approach called machine learning, in which a computer program analyzes a training dataset to "teach" itself how to be more accurate with future predictions.   

Matthew Kalscheur, MD, assistant professor (CHS), Cardiovascular Medicine, and eight other UW-Madison colleagues published about their work, which was also highlighted in Cardiology Today. The team's machine-learning algorithm was developed to predict CRT clinical outcomes in people with heart failure based on data from patients enrolled in the COMPANION trial. 

Once "trained", the algorithm produced a model that predicted clinical outcomes after CRT significantly better than previous methods. The researchers hope that this will help doctors and patients make better decisions together about whether CRT is in the patient's best interest.  

“Machine learning is a powerful, computational method that could allow for improved description of phenotypes and development of decision support tools to predict clinical outcomes and better inform shared decision-making with patients,” wrote Dr. Kalscheur and colleagues.

“As clinical data sets expand, application of machine-learning algorithms will lead to further improvements in precision cardiovascular medicine.” 

 

Resources:

Photo caption: Dr. Matthew Kalscheur (foreground) analyzes data in this 2015 file photo. Credit: Clint Thayer/Department of Medicine