Deep Learning Tool May Help Detect Pediatric Rheumatic Heart Disease

Researchers from Children’s National Hospital have developed a deep learning (DL) to detect latent rheumatic heart disease (RHD) in children, which may improve case identification and treatment in low-resource areas, according to a study published this week in the Journal of the American Heart Association.

The World Health Organization (WHO) indicates that RHD is the most common heart disease acquired by people under the age of 25. The disease primarily occurs in children and is caused by heart valve damage that results from one or more bouts of A streptococci bacterial infection. RHD causes nearly 300,000 deaths annually, the majority of which occur in low- and middle-income countries.

If caught early, the condition can be treated before it results in permanent heart damage.

However, detecting RHD requires a cardiologist to read an echocardiogram. While echocardiograms are non-invasive and more widely available than some other imaging technologies, specialized care like cardiology is often unavailable or limited outside of high-income countries.

To bridge this gap, the research team sought to develop an artificial intelligence (AI) approach that could be used to detect RHD with an echocardiogram, but without a cardiologist.

“There is a true art to interpreting this kind of information, but we now know how to teach a machine to learn faster and possibly better than the human eye and brain,” explained Marius George Linguraru, DPhil, MA, MSc, the Connor Family Professor in Research and Innovation and principal investigator in the Sheikh Zayed Institute for Pediatric Surgical Innovation at Children’s National, in a news release.

“Although we have been using this diagnostic and treatment approach since World War II, we haven’t been able to share this competency globally with low- and middle-income countries, where there are far fewer cardiologists. With the power of AI, we expect that we can, which will improve equity in medicine around the world.”

To develop the model, the researchers combined machine learning and DL modalities to build an ultrasound interpretation algorithm. The algorithm was then incorporated into a system of ultrasound probes and portable electronic devices. Using data collected with a handheld ultrasound probe, a tablet, and a laptop, the algorithm can accurately identify features of RHD – such as heart size – from images of the heart.

The system can also flag features that cannot be detected with the naked eye, such as variations in a child’s heart size. The research team noted that current RHD diagnostic criteria use two weight categories – below or above 66 pounds – as a proxy for pediatric heart size. However, heart size can vary significantly across both categories.

“Our algorithm can see and make adjustments for the heart’s size as a continuously fluid variable,” said co-lead author and staff scientist Pooneh Roshanitabrizi, PhD. “In the hands of healthcare workers, we expect the technology to amplify human capabilities to make calculations far more quickly and precisely than the human eye and brain, saving countless lives.”

The researchers posited that the system could significantly improve RHD detection and treatment worldwide.

“One of the most effective ways to prevent rheumatic heart disease is to find the patients that are affected in the very early stages, give them monthly penicillin for pennies a day and prevent them from becoming one of the 400,000 people a year who die from this disease,” stated Craig Sable, MD, interim division chief of Cardiology at Children’s National. “Once this technology is built and distributed at a scale to address the need, we are optimistic that it holds great promise to bring highly accurate care to economically disadvantaged countries and help eradicate RHD around the world.”

Others are also investigating how AI can improve the identification of other cardiovascular problems.

Last week, researchers from the Icahn School of Medicine at Mount Sinai revealed that they have developed an AI model capable of accurately predicting which patients are at risk for poor right ventricular function.

The right ventricle of the heart plays a critical role in pumping blood to the lungs, and issues with the heart’s right side can lead to a host of adverse outcomes if not flagged early on.

The challenge is that conditions like right ventricular ejection fraction (RVEF) and end‐diastolic volume (RVEDV) are not easily assessed via traditional methods. The AI model seeks to bridge this gap by predicting right ventricular dilation, dysfunction, and numerical RVEDV and RVEF based on electrocardiogram and magnetic resonance imaging (MRI) data.