Real-World Evaluation of Eko Algorithms in a Point of Care Setting

The purpose of this research is to prospectively test and validate the utility of Eko artificial intelligence (AI) plus Eko Murmur Analysis Software (EMAS) murmur characterization in algorithm in a real world, point-of-care setting.

Study Overview

• Status: Completed
• Conditions: Atrial Fibrillation; Heart Murmurs; Murmur, Heart; Innocent Murmurs; Pathologic Murmur
• Intervention / Treatment: Device: Use of Eko CORE and Eko CORE 500 electronic stethoscope

Detailed Description

Eko has developed a platform to aid in screening for cardiac conditions using a digital stethoscope and machine-learning algorithms to detect the presence or absence of heart conditions such as heart murmurs and atrial fibrillation.

In November 2019, the US Food and Drug Administration (FDA) granted Eko a 510(k) clearance for the marketing of "Eko AI", a set of machine learning algorithms that includes atrial fibrillation (AF) and heart murmur detection. The detection of heart murmurs may aid in detecting occult and dangerous valvular heart disease (VHD). Other Eko AI outputs include bradycardia, tachycardia, noisy signal, QRS duration, and unclassified data. Eko AI has accuracy comparable to physician judgment (atrial fibrillation sensitivity of 98.9% and specificity of 96.9%, murmur sensitivity of 87.6% and specificity of 87.8%). Both AF and VHD can cause significant morbidity and mortality when missed or diagnosed late.

Eko has further developed the murmur detection function of Eko AI to now not only identify whether a murmur is present, but also to inform the clinician of its timing during the cardiac cycle (systole vs diastole), and whether it is innocent or structural. We are calling this product the Eko Murmur Analysis Software (EMAS) and submitted a premarket notification to FDA in December 2021.

This study sets out to understand the utility of the Eko AI plus EMAS murmur characterization algorithm in real world use. Collecting data in a point-of-care setting will demonstrate how accurately the algorithm characterizes murmurs in comparison to an AI-unassisted clinical examination. Algorithm output and clinical determination will be confirmed by echocardiographic ground truth, with the results being blinded until the end of the patient visit.

• Study Type: Observational
• Enrollment (Actual): 371

Contacts and Locations

Study Locations

• United States
  • Massachusetts
    • Haverhill, Massachusetts, United States, 01830
      • Pentucket Medical Associates
    • Lawrence, Massachusetts, United States, 01843
      • Pentucket Medical Associates
  • North Carolina
    • Dunn, North Carolina, United States, 27546
      • Maria Medical Center
    • Lillington, North Carolina, United States, 27546
      • Edgewater Medical Center
    • Lillington, North Carolina, United States, 27546
      • Hometown Medical PLLC

Participation Criteria

Eligibility Criteria

• Ages Eligible for Study: 50 years and older (Older Adult)
• Accepts Healthy Volunteers: Yes

• Sampling Method: Non-Probability Sample

Study Population

Adult primary care clinic

Description

Inclusion Criteria:

• Patient consents to participation
• Willing to have heart sounds recorded with an electronic stethoscope
• Willing to undergo echocardiography
• Willing to undergo a 12-lead electrocardiogram
• Adults aged 65 years and older
• History of at least one of the following: hypertension, BMI ≥ 30, diabetes mellitus, hyperlipidemia, atrial fibrillation, myocardial infarction, stroke/TIA, previous coronary surgery, or previous coronary angiography
• No prior diagnosis of valve disease or heart murmur

Exclusion Criteria:

• Patient is unwilling or unable to give written informed consent
• Patients experiencing a known or suspected acute cardiac event
• Under the age of 65 years old
• Prior diagnosis of valve disease or heart murmurs

Study Plan

How is the study designed?

Design Details

• Observational Models: Cohort
• Time Perspectives: Prospective

Number of groups / cohorts

2

Cohorts and Interventions

Group / Cohort	Intervention / Treatment
Patients with Significantly Higher Risk of Developing VHD based on Medical History; History of at least one of the following: hypertension, BMI ≥ 30, diabetes mellitus, hyperlipidemia, atrial fibrillation, myocardial infarction, stroke/TIA, previous coronary surgery, or previous coronary angiography.	Device: Use of Eko CORE and Eko DUO electronic stethoscope; Auscultation of heart sounds using electronic stethoscopes
Patients without Significantly Higher Risk of Developing VHD based on Medical History; No history of hypertension, BMI ≥ 30, diabetes mellitus, hyperlipidemia, atrial fibrillation, myocardial infarction, stroke/TIA, previous coronary surgery, or previous coronary angiography.	Device: Use of Eko CORE and Eko DUO electronic stethoscope; Auscultation of heart sounds using electronic stethoscopes

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Sensitivity and Specificity of Eko's AI relative to ground truth	Sensitivity and specificity of Eko's murmur detection algorithm relative to ground truth. Ground truth is defined as echocardiography-confirmed clinically-significant (graded "mild-to-moderate" or greater severity) VHD that is associated with a murmur, as confirmed by an expert panel.	02/20/2022 - 05/20/2024
Sensitivity and Specificity of Eko's AI relative to PCP auscultation and ground truth	Sensitivity and specificity of Eko's murmur detection algorithm relative to ground truth and PCP auscultation ground truth. Ground truth is defined as echocardiography-confirmed clinically-significant (graded "mild-to-moderate" or greater severity) VHD that is associated with a murmur, as confirmed by an expert panel.	02/20/2022 - 05/20/2024
Positive and negative predictive values for AI detecting new VHD	Positive and negative predictive values of Eko's algorithms for detecting new clinically significant valvular heart disease	02/20/2022 - 05/20/2024
Positive and negative predictive values for PCP detecting new VHD	Positive and negative predictive values of a PCP's outpatient visit for detecting new clinically significant valvular heart disease	02/20/2022 - 05/20/2024

Secondary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Performance of the machine algorithm vs. the physician	Performance of the machine algorithm vs. the physician	02/20/2022 - 05/20/2024
Number of cardiac tests and consultations ordered	Number of cardiac tests and consultations ordered	02/20/2022 - 05/20/2024
Sensitivity and Specificity of Eko's AI relative to echocardiographic ground truth.	Sensitivity and specificity of Eko's murmur detection algorithm relative to echocardiographic ground truth.	02/20/2022 - 05/20/2024

Collaborators and Investigators

• Sponsor: Eko Devices, Inc.

Study record dates

Study Major Dates

• Study Start (Actual): April 15, 2022
• Primary Completion (Actual): June 30, 2023
• Study Completion (Actual): June 30, 2023

Study Registration Dates

• First Submitted: July 13, 2022
• First Submitted That Met QC Criteria: July 13, 2022
• First Posted (Actual): July 15, 2022

Study Record Updates

• Last Update Posted (Estimated): November 25, 2024
• Last Update Submitted That Met QC Criteria: November 21, 2024
• Last Verified: November 1, 2024

More Information

Terms related to this study

• Keywords: machine learning; artificial intelligence
• Additional Relevant MeSH Terms: Cardiovascular Diseases; Pathologic Processes; Heart Diseases; Arrhythmias, Cardiac; Atrial Fibrillation; Heart Murmurs
• Other Study ID Numbers: 2021.11

Plan for Individual participant data (IPD)

• Plan to Share Individual Participant Data (IPD)?: NO
• IPD Plan Description: All participant data will be de-identified. Data will only be shared with Eko Devices, and not shared with any other researchers.

Drug and device information, study documents

• Studies a U.S. FDA-regulated drug product: No
• Studies a U.S. FDA-regulated device product: No