Dr. Partho Sengupta, WVU Heart and Vascular Institute Cardiology division chief and director of cardiac imaging, and an international team of researchers have implemented machine learning as an effective screening tool for early diagnosis of left ventricular diastolic dysfunction, a common form of heart failure.
Machine learning uses algorithms that automatically improve and become more accurate over time through the continued input of data.
“This research can be used by internists and physicians for early identification of heart dysfunction and to predict the progression of heart disease in order to help patients take action to improve their cardiac health,” Sengupta said. “We applied machine learning to the results of an electrocardiogram, a common, cost-effective test used by physicians. We can use that data to extract information that would otherwise come from more expensive imaging tests, using the heart’s electrical signals to provide information about the speed with which the heart is squeezing and relaxing.”
The researchers used electrocardiogram data gathered from a diverse patient population at risk of heart failure at four centers in North America to develop a predictive model of patient risk. Recent studies suggest that heart failure develops in 50% of patients due to the heart’s inability to relax. Slower relaxation requires higher pressures to fill the cardiac chamber, despite normal left ventricular ejection fraction, the heart’s ability to pump blood out of the left side.
“By taking data from a large sample of patients in varying locations, we are able to gather information that can be applied to people from different locations and backgrounds,” Sengupta said. “We know that there are areas that are more likely to have higher incidences of heart disease due to socioeconomic factors that make it difficult to maintain a healthy lifestyle. This research will allow us to develop cost-effective screening tools that can be used in community to identify patients at risk even before they develop any symptoms.”
This model can be used to determine the patient’s current stage of disease so that early interventions can be taken in order to halt the progress or reduce the likelihood of heart failure. It takes into account the patient’s age and gender, which can inform individualized risk estimation and interventions for men and women.
Sengupta’s work is supported in part by funds from the National Science Foundation. The work is posted online on the website of the Journal of American College of Cardiology and features Naveena Yanamala, Ph.D., principal data scientist for the WVU Heart and Vascular Institute Innovation Center and Dr. Grace Casaclang-Verzosa, M.B.A., WVU Heart and Vascular Institute administrative director of non-invasive cardiology, among other distinguished international collaborators.