American Society of Echocardiography – Handheld Wireless Digital Phonocardiography for Machine Learning-Based Detection of Aortic Stenosis

July 25, 2019

Handheld Wireless Digital Phonocardiography for Machine Learning-Based Detection of Aortic Stenosis

Brent E. White MD, Jason Paek BA, Patrick M. McCarthy MD, and James D. Thomas MD, Bluhm Cardiovascular Institute at Northwestern Memorial Hospital. Chicago, IL

Steve Pham MD and John Maidens PhD, Eko Devices, Inc. Berkeley, California

This research was presented by Dr. Bent White at the 2019 American Society of Echocardiography Scientific Sessions, where he received the 2019 ASE Foundation Top Investigator Award.


  • Aortic stenosis (AS) is a common disease which can be detected as a murmur on auscultation, but studies show the majority of new primary care physicians do not detect AS murmurs (citations 1, 2).
  • Although transthoracic echocardiography remains the gold standard for diagnosis of AS (citation 3), this typically requires a referral from a provider who has recognized an abnormality on auscultation.
  • The FDA-approved Eko CORE device is a digital stethoscope wirelessly paired with the Eko Mobile application (figure 1) to allow recording and analysis of phonocardiograms (PCG) (figure 2).


  • Our objective is to use these PCG data to drive a machine learning-based detection algorithm to automatically identify clinically significant AS, validated by TTE, as part of the ongoing Phono- and Electrocardiogram Assisted Detection of Valvular Disease (PEA-Valve) Study.
Figure 1: Eko CORE Digital Stethoscope, Figure 2: PCG recordings from a patient with significant AS


  • Consecutive patients undergoing TTE at Northwestern Medicine were prospectively enrolled to undergo PCG recording by the Eko CORE device.
  • Recordings 15 seconds long were obtained at the four standard auscultation positions
  • A TensorFlow-based machine learning algorithm assessed the presence or absence of murmur with dominant localization to the right upper sternal border indicating clinically significant AS, defined as moderate or greater on TTE (figure 3).


  • 161 patients were enrolled at the time of data analysis, yielding 639 recordings.
  • 14 of these patients (8.7%) were found to have significant AS on TTE.
  • The receiver-operating characteristic curve had an area of 0.964, yielding a sensitivity of 97.2% (95% CI, 84.7-99.5%) and a specificity of 86.4% (95% CI, 84.0-88.7%) for the detection of clinically significant AS (figure 3).
Figure 3: Development and testing of TensorFlow-based machine learning algorithm to detect AS

Figure 4: Receiver-operating characteristic curve for detection of AS


  • PCG assessment using the Eko CORE Digital Stethoscope and machine learning interpretation is a fast and effective method to screen for significant AS and should be validated in a primary care setting, which may lead to more appropriate referrals for an echocardiogram.


  1. Mangione S, Nieman LZ. Cardiac auscultatory skills of internal medicine and family practice trainees. A comparison of diagnostic proficiency. JAMA. 1997 Sep 3;278(9):717-22.
  2. Mangione S. Cardiac auscultatory skills of physicians-in-training: a comparison of three English-speaking countries. Am J Med. 2001;110(3):210-216.
  3. Baumgartner H, Hung J, Bermejo J, et al. Recommendations on the Echocardiographic Assessment of Aortic Valve Stenosis: A Focused Update from the European Association of Cardiovascular Imaging and the American Society of Echocardiography. J Am Soc Echocardiogr. 2017 Apr;30(4):372-392.

Visit the American Society of Echocardiography

More from our blog