A startup called Eko and a handful of cardiologists are using artificial intelligence in an effort to improve the 200-year-old tool

The stethoscope is a ubiquitous screening tool and a symbol of the medical profession. It may also need an update.

The effectiveness of the 200-year-old device depends on the skill, experience and hearing abilities of the doctor using it. Without years of cardiac training and experience with patients, it’s difficult to recognize, much less diagnose, heart and lung abnormalities. Interpreting the sounds is sometimes more art than science, doctors say.

“It’s a relic. It’s rubber tubes,” said Eric Topol, a cardiologist and founder and director of Scripps Research Translational Institute. “It has very little utility compared to how it continues to be so highly revered.”

Eko, a Berkeley, Calif.-based startup, is one of a handful of companies working to bring the simple stethoscope into the 21st century. Connor Landgraf, then a student in biomedical engineering at the University of California, Berkeley, got the idea to build a better version in 2013 after medical residents told him their stethoscopes were unhelpful for gathering patient information—they wore them mostly for show. Mr. Landgraf and co-founders Jason Bellet and Tyler Crouch launched the company in May 2014 and said they have raised $9 million.

Eko’s latest stethoscope is the Duo, a gadget that looks like a small television remote and also takes an electrocardiogram, or ECG. The test, which measures the heart’s electrical activity, is typically done by sticking electrodes on a patient’s chest and limbs for less than a minute. With the Duo, a doctor holds the device on the patient’s chest, it records heart sounds and electrical signals, and then it sends the data to a mobile app that displays the recordings in real-time. A doctor can save, replay and share the information—a function that’s increasingly important for medical training and for treating patients remotely. The Duo costs $349.

Digital stethoscopes and mobile ECG monitors have become more common in the last decade. Thinklabs, Littmann and HD Medical Group sell digital stethoscopes for roughly $200 to $600 apiece. (A classic stethoscope runs $5 to $150, depending on the type and quality.) AliveCor sells mobile ECG devices aimed at consumers that use algorithms to screen for heart conditions. The latest version of the Apple Watch, which the technology giant unveiled in September, includes an ECG feature that can identify an irregular heart rhythm, though doctors have raised concerns about potential false positives and other risks of relying on the app.

Mr. Landgraf hopes to improve the Duo’s functionality with artificial intelligence—to make it a device that not only records ECGs for a doctor to interpret but also performs diagnoses of heart conditions. By comparing data sets of normal heart sounds and electric patterns to the sounds and patterns of specific abnormalities, the Eko team and their partners are developing algorithms that they hope will identify specific heart-related issues as effectively as an experienced cardiologist. The machine would essentially “hear” an irregularity at a more granular level than a doctor’s ear and then flash a “yes” or “no” diagnosis on the accompanying app.

The Duo received FDA clearance for use as a stethoscope and ECG device in 2017. Now, the company is seeking clearance to incorporate two diagnostic algorithms.

The first is to screen for asymptomatic left ventricular dysfunction, a heart rhythm disorder that’s often a precursor to heart failure yet rarely shows symptoms. Last year, the Mayo Clinic trained an algorithm to identify ALVD by feeding it ECG data from more than 45,000 patient records. The team then used the algorithm on ECGs from 52,870 other patients and found that it was accurate 93% of the time, compared to results from a gold-standard test known as an echocardiogram. In instances where the algorithm detected ALVD but the echocardiogram did not show it, the team later found the patients were more likely to develop the condition than those with a negative diagnosis. “The signals are hidden in plain sight,” said Paul Friedman, chair of cardiovascular medicine at the Mayo Clinic and senior researcher on the study. “It’s like looking into the future.”

The results were published in January in Nature Medicine. The Mayo Clinic has since licensed the algorithm to Eko and is currently testing it on patients using the Duo.

Eko has also developed an algorithm to identify heart murmurs, which the company tested on a data set of pediatric heart sounds. They found that the algorithm was more effective at picking up the murmurs than five pediatric cardiologists who listened to the sounds, when compared against echocardiogram results, according to an abstract of a study presented at an American Heart Association conference in 2018. In March, the company started a clinical trial with Northwestern Medicine Bluhm Cardiovascular Institute to test the heart-murmur algorithm on 800 patients, who will also be screened with an echocardiogram.

“It’s one thing to listen with a digital stethoscope. Then you need a whole lot of separate training anyway to see what those squiggles read,” said Patrick McCarthy, executive director at the institute.

Many medical professionals have yet to adopt digital stethoscopes, cardiologists say. The ability to share, replay and amplify heart sounds is welcome, but it may not justify the additional cost, particularly for non-specialists. Some doctors say the medical profession is slow to embrace new technology.

Eko is betting that an AI-enabled diagnostic function will push more medical professionals to make the switch. The company said it has sold over 20,000 Duos and an earlier digital stethoscope to more than 2,000 hospitals, clinics and private practices.

Johns Hopkins University recently developed a digital stethoscope that uses an algorithm to screen for lung sounds that may indicate pneumonia—an often fatal disease for children in low-resource countries, said Mounya Elhilali, an engineer who designed the tool. The algorithm, trained on data from 1,500 patients across Africa and Asia, can distinguish patients with pneumonia 87% of the time, according to a study in the journal Transactions on Biomedical Engineering. A Baltimore-based startup, Sonavi Labs, sprung from the project, and the stethoscope, called the Feelix, is being tested in four countries. Sonavi CEO Ellington West said accuracy has since improved to 91%.

Some doctors want to abandon stethoscopes in favor of portable ultrasound devices, which create an image of the heart or lungs and are undergoing a similar digital transformation. Though even these pocket-size ultrasound machines—which cost a few thousand dollars —require additional time and training, proponents say they provide more information and help doctors make more accurate diagnoses. “Instead of just hearing, you can see,” said Davinder Ramsingh, an associate professor of anesthesiology at Loma Linda University School of Medicine in California.

Even the traditional stethoscope didn’t become popular until years after its invention, according to Dr. Topol. Doctors of the day were reluctant to learn how to interpret heart sounds and worried the tool would interfere with examinations, he said. “There was a war about the stethoscope,” he said.

 


 

Recent Case Studies

 

Request a demo to learn more about our cardiac monitoring platform.