The work builds upon research published last year showing that an AI program can detect disease in the heart’s mitral valve by analyzing ultrasound images of the heart. For this new study, published in JAMA Cardiology, investigators applied AI to identify tricuspid regurgitation, a condition in which the heart’s tricuspid valve doesn’t close fully when the heart contracts, causing blood to flow backward, which can result in heart failure.  

“This AI program can augment cardiologists’ evaluation of echocardiograms, images from a screening and diagnostic test that many patients with heart disease symptoms would already be getting,” said David Ouyang, MD, a research scientist in the Smidt Heart Institute, an investigator in the Division of Artificial Intelligence in Medicine and senior author of the study. “By applying AI to echocardiograms, we can help clinicians more easily detect the signs of heart valve disease so that patients get the care they need as soon as possible.”

Investigators trained a deep-learning program to flag patterns of tricuspid regurgitation in 47,312 echocardiograms done at Cedars-Sinai between 2011 and 2021.

The program detected tricuspid regurgitation in patients and categorized cases as mild, moderate or severe. They then tested the program on echocardiograms that the AI program never saw before from additional patients who underwent echocardiography at Cedars-Sinai in 2022 and patients from Stanford Healthcare. The program predicted severity of tricuspid regurgitation with similar accuracy as cardiologists who evaluated echocardiograms and when compared with results from MRI images.

“Future studies will focus on obtaining even more specific information about valve disease, such as the volume of blood flowing backward through a valve, and predicting outcomes if patients undergo treatment for heart valve disease,” said first author Amey Vrudhula, MD, a research fellow at Cedars-Sinai.

Investigators in the Smidt Heart Institute are applying AI to a variety of cardiac imaging tests. 

“A major advantage of AI algorithms is that they never get fatigued and have the capacity to identify valve abnormalities from large populations of patients, taking personalized cardiology to a whole different level,” said Sumeet Chugh, MD, director of the Division of Artificial Intelligence in Medicine and the Pauline and Harold Price Chair in Cardiac Electrophysiology Research. 

Other Cedars-Sinai authors involved in the study include Amey Vrudhula, MD; Milos Vukadinovic, BS; Alan C. Kwan, MD; Daniel Berman, MD; Robert Siegel, MD; Susan Cheng, MD, MMSc, MPH.

Other authors include Christiane Haeffele, MD, and David Liang, MD, PhD.

The work was supported by the Sarnoff Cardiovascular Research Award, research grants R00 HL157421 and R01HL173526 and support from AstraZeneca Alexion, as well as consulting from EchoIQ, Ultromics, Pfizer and InVision.

Cedars-Sinai Health Sciences University is advancing groundbreaking research and educating future leaders in medicine, biomedical sciences and allied health sciences. Learn more about the university.

The program’s findings, published in Circulation, may help clinicians identify patients whose mitral valve regurgitation is manageable with medication as well as patients with more severe cases who would benefit from a minimally invasive valve repair procedure or surgery.

“Mitral regurgitation is a common but often missed valvular heart disease. It can be challenging to precisely assess the disease severity, which is critical to know which patients can take a watch-and-wait approach and which should proceed to an intervention,” said David Ouyang, MD, a cardiologist in the Department of Cardiology in the Smidt Heart Institute, an investigator in the Division of Artificial Intelligence in Medicine, and corresponding author of the study. “The program we developed may one day be used by doctors when considering the best treatment approach for individual patients.”

The AI program further cements the Smidt Heart Institute’s longstanding leadership in heart valve care. Interventionalists with the Smidt Heart Institute are believed to have performed more mitral valve repairs than any other center in the U.S., with outcomes that place Cedars-Sinai among the top-performing programs nationally. The Smidt Heart Institute team has also completed more than 1,500 robotic mitral valve repairs with a near 100% success rates.

“This could improve how we identify patients with mitral regurgitation, which is becoming more prevalent in our aging population, and to personalize treatment even more so than we already do,” said Raj Makkar, MD, associate director of the Smidt Heart Institute, vice president of Cardiovascular Innovation and Intervention for Cedars-Sinai and a leader in treating mitral valve disease.

The heart has four valves that open and close to move blood throughout the body. In some people the mitral valve, located on the left side of the heart, does not close properly, which allows blood to flow backwards, a condition called mitral valve regurgitation. The condition prevents enough blood from circulating throughout the body and, over time, can lead to shortness of breath, arrhythmia and heart failure.

“At Cedars-Sinai we are pursuing the use of AI as a complementary tool in diagnosing and treating conditions such as mitral valve regurgitation,” said Sumeet Chugh, MD, director of the Division of Artificial Intelligence in Medicine and the Pauline and Harold Price Chair in Cardiac Electrophysiology Research.

In developing the new program, investigators used more than 58,000 transthoracic echocardiograms from Cedars-Sinai. Echocardiograms are video images of patients’ hearts taken by ultrasound and is the most common way to assess mitral regurgitation. The investigators tested the program on echocardiograms from 1,800 patients at Cedars-Sinai as well on echocardiograms from 915 patients from Stanford Healthcare in Northern California.

The model was able to automatically identify moderate and severe mitral valve regurgitation with high precision.

“Our deep learning model analyzed videos from more than 50,000 echocardiogram studies and can pinpoint the most relevant and important videos to assess mitral regurgitation severity,” said first author Amey Vrudhula, a fellow at Cedars-Sinai.

To treat severe mitral valve regurgitation, experts at the Smidt Heart Institute at Cedars-Sinai rely on either the minimally invasive TEER procedure or minimally invasive surgery. All patients meet with an interventional cardiologist as well as a cardiac surgeon before making their treatment decision.

Other Cedars-Sinai authors involved in the study include Grant Duffy B.S., Milos Vukadinovic B.S^., Susan Cheng, M.D., M.M.Sc., M.P.H.

This work was supported in part by the Sarnoff Cardiovascular Research Award, research grants NIH R01-HL131532, NIH R01-HL142983, R00 HL157421, and R01HL173526.

Read more from the Cedars-Sinai Blog: Treatment Options for Heart Valve Disease

.

“While there is a tremendous amount of excitement around artificial intelligence, particularly as it pertains to healthcare, there remains a gap between research findings on limited historical datasets and bringing unproven algorithms to the clinic and patients,” said Ouyang, a faculty member in the Department of Cardiology in the Smidt Heart Institute at Cedars-Sinai and in the Division of Artificial Intelligence in Medicine.

The recently launched journal will spotlight clinical trials and practice-changing innovations, along with reviews, policy perspectives, and educational materials designed specifically for practicing physicians and clinician leaders interested in leveraging AI.

“There’s a keen interest in understanding and showcasing the results of artificial intelligence in clinical practice and deployment, particularly with an eye on patient impact,” said Ouyang.

As a deputy editor, Ouyang is responsible for the curation, screening and review of many of the submitted studies and editorials about cardiology. He is the only cardiologist among the NEJM AI deputy editors.

At Cedars-Sinai, Ouyang’s research focus is on cardiac imaging as well as application of artificial intelligence and data science to healthcare. A statistician by training, Ouyang said he has always been fascinated by understanding, visualizing, and interpreting data. As a cardiologist, he is interested in the generation and assessment of clinical evidence and data.

“The Smidt Heart Institute and our broader academic enterprise have benefited greatly from David’s vast knowledge and expertise in large language models, artificial intelligence, imaging and cardiology,” said Christine M. Albert, MD, MPH, chair of the Department of Cardiology in the Smidt Heart Institute and the Lee and Harold Kapelovitz Distinguished Chair in Cardiology. “As an innovator in the field, we are particularly honored to have David represent Cedars-Sinai, personifying our commitment to lead in AI medical applications.”

Among his research achievements, Ouyang led the first-of-its-kind, blinded, randomized clinical trial of AI in cardiology. The study found AI was superior in assessing cardiac function in echocardiograms when compared with assessments made by sonographers. The findings were published in late 2023 in the peer-reviewed journal Nature.

Ouyang has published nearly 100 academic research studies in medical journals such as Journal of the American College of Cardiology (JACC), European Heart Journal, and Circulation.

Read more from the Cedars-Sinai Blog: Reducing Bias in AI Models