Improving Detection of Aortic Stenosis with Machine Learning

Aortic stenosis (AS) is a serious heart condition in which the aortic valve is narrowed, impairing the flow of blood out of the heart. Because symptoms appear slowly and are difficult to detect through routine physical examination, diagnosis frequently occurs late in the disease’s progression, when risks of heart failure and mortality are sharply increased.

To address the need for improved detection of AS, a multidisciplinary team at Tufts has developed an ML–enabled method to diagnose AS using handheld ultrasound imaging. The team, which is co-led by Benjamin Wessler, MD, a Tufts Medical Center cardiologist, and Michael Hughes, PhD, a Tufts University computer scientist, recently received an R01 grant from the National Heart, Lung, and Blood Institute to test the system at Tufts Medicine primary care clinics. “We are trying to improve the detection of this life-threatening condition upstream of traditional echocardiogram laboratories,” says Dr. Wessler. “Our goal is to leverage machine learning to automate the interpretation of hand-held cardiac ultrasound imaging to improve detection and ultimately treatment of this condition.”

 

The project builds on the team’s 2021 Tufts CTSI Pilot Award (now known as the Small Grants to Advance Translational Science, or S-GATS Program) to develop machine learning algorithms for detecting AS using images from limited-view scans, such as can be acquired by nonspecialists using an abbreviated imaging protocol. Although comprehensive echocardiography at an imaging center remains the diagnostic gold standard, the team’s AI models have demonstrated high accuracy on images from commercially available, low-cost handheld machines. These results suggest a viable pathway for integrating reliable screening into routine clinical encounters, potentially enabling the earlier identification of high-risk patients.

 

Tufts CTSI support has been catalytic in ways that precede and extend beyond the pilot funding initially provided. Dr. Wessler’s research career began with a Tufts CTSI Fellowship, while Dr. Hughes participated in our Junior Faculty Research Career Development Forum soon after joining the Tufts faculty.

 

Presently, our Dissemination and Implementation Core is helping guide plans for implementing their technology in typical primary care settings.