A new study by a team of researchers, including Partho Sengupta, MD, chief of cardiology at the WVU Heart and Vascular Institute, utilized machine learning to predict the clinical presentations and treatment outcomes of patients with aortic stenosis.
Aortic stenosis, or a narrowing of blood vessels, can restrict blood flow to the heart, leading to lasting damage. This condition occurs in nearly two percent of people over the age of 65 and affects more men than women. It often appears with other comorbidities, such as hypertension and diabetes.
“There is always a clinical struggle about how to identify these patients because it can be difficult to correlate patient symptoms with aortic stenosis in the presence of several common comorbidities,” Dr. Sengupta said. “We used topological data analysis – a technique that extracts meaning from patient data automatically – to develop a new model for the diagnosis and progression of the disease.”
This study used data from 246 patients to develop a timeline of disease risk and progression both with and without medical intervention. This data was compared to a model developed using a mouse study in which aortic stenosis progresses rapidly and showed similar results.
“As researchers, we may not have the follow-up data from patients across their lifetime to monitor the progression of their disease, so it is important we have methods to quickly predict the disease progression from their short-term data,” Sengupta said. “Combining the knowledge gained from patient data and the validity of the experimental mouse study, we were able to confirm precisely how the model predicts what is likely to happen and the future risks for new patients who come to us.”
This model showed that modern computational techniques using machine learning can help doctors with new insights to understand how aortic stenosis presents and progresses in a predictable way. Medical interventions can place the patient at an earlier stage on the timeline, though long-term damage has already been done.
“This information opens up new ways to look at patients with aortic stenosis and can help us develop new therapies for the treatment of aortic stenosis by looking at individualized risks,” Sengupta said. “Often, clinical trials give the results of the average patient, but there is no average patient. By analyzing data, we are able to target individualized therapies and identify individualized risks to provider better interventions for patients.”
The study, titled “Network Tomography for Understanding Phenotypic Presentations in Aortic Stenosis,” was published today (Feb. 5) in The Journal of the American College of Cardiology: Cardiovascular Imaging.