AI Stethoscope Doubles Detection of Valvular Heart Disease

A digital stethoscope enhanced by artificial intelligence (AI) has shown to more than double the sensitivity of a traditional stethoscope when detecting moderate or severe valvular heart disease. These findings were published today in the European Heart Journal – Digital Health.

“Valvular heart disease is unfortunately very common among older adults, yet it often goes undetected until symptoms become advanced,” said Rosalie V. McDonough, MD, senior manager of medical affairs at Eko Health. “This means that patients can experience complications and worsening health which could have been prevented with earlier diagnosis.” 

More than 13% of people aged 75 or older are estimated to have moderate to severe valvular heart disease, but the condition often goes undiagnosed because symptoms are absent in about 50% of patients and often vague in many of the remaining individuals. Delayed diagnosis and treatment can significantly reduce quality of life, pose a life-threatening risk, and poses a financial burden to healthcare institutions in the long term. 

Heart murmurs are an early clinical indicator of valvular heart disease that can be detected through auscultation with a stethoscope, flagging patients for an echocardiogram to confirm the diagnosis. However, this method currently fails to identify about 56% of positive cases. 

The digital stethoscope used in the study is powered by an AI algorithm that recently received FDA clearance to detect a range of heart conditions based on heart sound recordings. While traditional methods rely on the physician’s expertise and hearing, which may be affected by external factors such as background noise, the machine learning model relies on millions of clinical data points to detect unique patterns associated with valvular heart disease. 

“The use of artificial intelligence provides an additional analytical layer, highlighting abnormalities that may be difficult to consistently detect by ear alone,” said McDonough. “But technology is not taking over; use of this device requires doctors to use their own clinical judgement.”  

The study enrolled a total of 357 patients 50 years old or over, who were examined both with a traditional stethoscope in a primary care setting and with the AI-powered digital stethoscope by trained users. Each patient then underwent an echocardiogram to confirm whether they had moderate to severe valvular heart disease. 

Out of 13 patients later confirmed to be positive, the AI-augmented system detected 12, while the traditional stethoscope only flagged six of them, resulting in a sensitivity of 92% and 46%, respectively. 

However, results also revealed a trade-off between sensitivity and specificity, with the AI-system incorrectly flagging 45 false positives compared to 15 with the stethoscope. The study authors argued that the potential rise in healthcare costs associated with a higher number of false positives would be balanced in the longer term by the added value of early detection, especially in patients who would otherwise progress to advanced disease stages before being diagnosed. 

“We have shown that​ an AI​-enabled stethoscope is much better at spotting which patients have moderate to severe valvular disease than a traditional stethoscope in real-world clinical settings,” said McDonough. “We hope this technology will allow patients to get faster access to an echocardiogram to formally diagnose their condition and then access treatment more quickly. At a population level, this technology could reduce hospital admissions and the ​overall ​cost of healthcare.”  

“An additional benefit we observed during the study was that patients assessed with the AI-enabled digital stethoscope seemed more engaged during their appointment,” McDonough added. “We think this was because they could see and hear what the clinician was responding to—which may increase trust and engagement with​ ​follow-up treatment.”