FDA clears first AI-powered mobile app to catch heart murmurs

Following years in development, the FDA has cleared a mobile-based solution that uses artificial intelligence to detect and classify the sounds of potentially dangerous heart murmurs using a digital stethoscope.

While physicians have been manually listening to chest and heart sounds to screen for disease for centuries, the platform developed by eMurmur allows doctors, nurses and other practitioners to employ their smartphones—allowing for greater accuracy anywhere without the need of a trained specialist.

Using a third-party digital stethoscope, the recording is uploaded via the company’s mobile app to a HIPAA-compliant server, where it’s analyzed against a large database of heart sounds and cardiologist annotations derived from patients of all ages.

But the time needed to develop a full-fledged medical device, as well as to secure the FDA’s blessing, can be a little longer than a lot of people would like—so what can you do while you wait?

“So we're sitting on his large database, and, you know, clinical clearances take a while. And then we talked to doctors, and they said it would be really good to have a more modern tool to teach murmur education as well,” eMurmur CEO Andreas Schriefl told FierceMedTech.

Traditionally, medical schools have used lectures to large groups of students with recordings played over speakers to train prospective doctors. Additionally, students may spend time in clinics, waiting for relevant cases that may or may not come along, Schriefl said.

It’s also not an easy task: Listening for signs of congenital heart disease is common in newborn screenings—where rather noncompliant patients scream, cry, move and have very rapid heart rates that can make subtle murmurs difficult to hear.

So though development began with the now-cleared AI program, eMurmur has been able to offer a handful of training courses and continuing medical education products in the meantime, featuring at-home practice and group testing modules.

“From an IT perspective, the educational apps are an easy product to do,” Schriefl said. “You have got to build an app and link it to the database, which we did while wrapping up the quality and risk management systems for the clinical product.”

Now, the goal is driving adoption. While the accuracy of eMurmur’s detection app lies between 85% and 90%, that’s still far higher than historical frontline screening rates of 25% to 30%. Still, many providers may see a high-tech solution as unnecessary when a trusty stethoscope will do.

“I think that's the challenge generally in medical devices—you have lots of data saying there is a problem, but then with everyone you ask, they say ‘Oh, I know how to do it,’” he said. “But it's not about us against the doctor, it's just us giving them a tool that helps them identify which patients should come in earlier.”

And through its educational programs, Schriefl and eMurmur have run into many practitioners that admit they don’t always know exactly what they’re hearing, and out of an abundance of caution they will send their patient along to a specialist.

Getting a quick answer would not only be welcomed in the hospital or the clinic but also in school offices, sports screenings and corporate health exams.  

“It took us five years to develop, and it takes two seconds to determine the result,” Schriefl said. “That's the frustrating part, from an engineer's perspective.”

In the future, eMurmur sees itself expanding into other cardiovascular disease indicators, such as the detection of third heart sounds, or S3, which can be related to heart failure, as well as lung issues. The company also hopes to secure a CE mark for its latest app in the coming weeks.