Deep Learning Detection of Pulmonary Hypertension and Low Ejection Fraction Via Digital Stethoscope and 3-Lead ECG

Официальное название

Deep Learning for Detection of Pulmonary Hypertension and Reduced Left Ventricular Ejection Fraction Using a Combined Digital Stethoscope and Three-lead Electrocardiogram

Обзор

This is a prospective, observational study evaluating whether heart sounds (phonocardiograms) and three-lead electrocardiograms (ECGs) recorded using the Eko CORE 500 digital stethoscope can help detect pulmonary hypertension (PH) and low left ventricular ejection fraction (EF ≤ 40%). PH is a condition characterized by high blood pressure in the pulmonary arteries, which can lead to heart failure and carries significant risks if undiagnosed. Low EF, which indicates reduced pumping ability of the heart, is also associated with increased risk of severe cardiac events but can remain undetected because patients often have no symptoms or only nonspecific symptoms. In this study, adults undergoing clinically indicated echocardiograms or right heart catheterization at outpatient sites will be invited to participate. Participants will complete a single study session lasting about 20 minutes, during which heart sounds and a three-lead ECG will be collected using the Eko CORE 500 device. If participants have had a clinical 12-lead ECG within 30 days of their echocardiogram or right heart catheterization, those data may also be used for analysis. A clinically indicated echocardiogram or right heart catheterization (RHC) performed within seven days before or after the Eko CORE 500 recording will serve as the reference standard to confirm the presence or absence of PH and low EF. Up to 3,850 participants may be enrolled across multiple sites to ensure that approximately 3,500 complete the study. The data collected will be used to develop and validate artificial intelligence (AI) algorithms that aim to detect PH and identify low EF, potentially enabling earlier and simpler screening for these conditions in clinical practice.

Подробное описание

Pulmonary hypertension (PH) and low left ventricular ejection fraction (EF) are significant cardiovascular conditions associated with increased morbidity and mortality but often remain underdiagnosed due to the need for specialized imaging such as echocardiography or invasive right heart catheterization. Early detection tools could enable timely intervention and improved patient outcomes.

This prospective, observational study aims to determine whether acoustic heart sounds (phonocardiograms, PCG) and three-lead electrocardiograms (ECG) recorded with the Eko CORE 500 digital stethoscope can identify patients with PH or low EF (defined as EF ≤ 40%) when compared with echocardiographic or right heart catheterization findings as the reference standard. The study will enroll adult patients undergoing clinically indicated transthoracic echocardiography or right heart catheterization at outpatient sites.

Participants will complete a single study visit, lasting approximately 20 minutes, during which heart sounds and three-lead ECG signals will be recorded at four standard auscultation sites (aortic, pulmonic, tricuspid, and mitral) while seated. Each recording lasts approximately 15 seconds. If a participant has undergone a 12-lead ECG within 30 days of their echocardiogram or right heart catheterization, de-identified ECG data will also be included for comparison purposes. Poor-quality recordings will be repeated once before moving to the next auscultation site. No results from the CORE 500 device or developed algorithms will be shared with participants or entered into the medical record.

De-identified demographic data collected will include age, race/ethnicity, and sex. Clinical data will include past medical history, relevant laboratory results (such as BNP or NT-proBNP), electrocardiographic findings, and echocardiographic measurements including tricuspid regurgitant jet velocity, pulmonary artery pressures, chamber size, and left ventricular ejection fraction.

Data will be analyzed by Eko Health, Inc. using machine learning techniques, including transformer-based models implemented in Python with PyTorch. Models will initially be pre-trained on unlabeled data and then fine-tuned on labeled data, optimizing performance using the Adam optimizer and binary cross-entropy loss. Algorithm performance will be evaluated based on sensitivity, specificity, area under the receiver operating characteristic curve (AUC), and other diagnostic metrics. Confidence intervals for sensitivity and specificity will be calculated to assess the statistical reliability of results.

Sample size calculations account for the estimated prevalence of each condition. For PH, an anticipated prevalence of 25-30% and target algorithm sensitivity and specificity exceeding 0.7 drive a required enrollment of approximately 2,400 participants to achieve statistical confidence. For low EF, assuming a prevalence of 10%, a minimum of 2,000 participants is required to adequately power analyses for sensitivity and specificity above 0.7.

The primary endpoint of this study is to evaluate sensitivity and specificity for the PH and low EF detection algorithms, respectively. The secondary endpoint is to measure algorithm accuracy, area under the ROC curve, negative predictive value, and positive predictive value for detecting low EF.

The study plans to enroll up to 3,850 participants across multiple sites to ensure sufficient evaluable data from approximately 3,500 participants. The intended outcome is to develop and validate AI-based tools that may facilitate non-invasive, point-of-care screening for PH and low EF using the Eko CORE 500 digital stethoscope, potentially reducing the burden of undiagnosed cardiovascular disease.

Вмешательства

• Устройство Eko CORE 500 Digital Stethoscope
  The FDA-cleared Eko CORE 500 digital stethoscope is used to collect phonocardiogram (PCG) and three-lead ECG recordings from participants. This observational study uses these recordings to develop and validate artificial intelligence algorithms to detect pulmonary hypertension and low left ventricular ejection fraction. No modifications to the device or device functionality are being tested.

Первичные конечные точки

• Sensitivity and specificity of the deep-learning algorithm for detecting pulmonary hypertension (PH) [Срок оценки: Up to 24 months]

Вторичные конечные точки (1)

• Algorithm Diagnostic Performance for Detection of Low Ejection Fraction [Срок оценки: Through study completion, 2 years]

Критерии участия

Критерии включения

• Adults aged 18 years and older
• Able and willing to provide informed consent
• Completed a clinical echocardiogram or right heart catheterization within 7 days before or after study procedures

Критерии исключения

• Unwilling or unable to provide informed consent
• Patients who are hospitalized
• Patients undergoing echocardiography with a limited echocardiogram (does not apply to patients undergoing right heart catheterization)

Критерии приведены из реестра в оригинале (на английском). Окончательную оценку соответствия проводит исследовательский центр.

Дизайн исследования

Модель наблюдения

Когортное

Центры проведения

США · 4 центра

• Prairie Cardiovascular — O'Fallon
• Prairie Education & Research Cooperative — Springfield
• St Johns Hospital, Springfield — Springfield
• Sentara Norfolk General Hospital — Norfolk

Публикации

• Frost A, Badesch D, Gibbs JSR, Gopalan D, Khanna D, Manes A, Oudiz R, Satoh T, Torres F, Torbicki A. Diagnosis of pulmonary hypertension. Eur Respir J. 2019 Jan 24;53(1):1801904. doi: 10.1183/13993003.01904-2018. Print 2019 Jan. PMID 30545972
• Maron BA, Hess E, Maddox TM, Opotowsky AR, Tedford RJ, Lahm T, Joynt KE, Kass DJ, Stephens T, Stanislawski MA, Swenson ER, Goldstein RH, Leopold JA, Zamanian RT, Elwing JM, Plomondon ME, Grunwald GK, Baron AE, Rumsfeld JS, Choudhary G. Association of Borderline Pulmonary Hypertension With Mortality and Hospitalization in a Large Patient Cohort: Insights From the Veterans Affairs Clinical Assessmen PMID 26873944
• Assad TR, Maron BA, Robbins IM, Xu M, Huang S, Harrell FE, Farber-Eger EH, Wells QS, Choudhary G, Hemnes AR, Brittain EL. Prognostic Effect and Longitudinal Hemodynamic Assessment of Borderline Pulmonary Hypertension. JAMA Cardiol. 2017 Dec 1;2(12):1361-1368. doi: 10.1001/jamacardio.2017.3882. PMID 29071338
• Strange G, Stewart S, Celermajer DS, Prior D, Scalia GM, Marwick TH, Gabbay E, Ilton M, Joseph M, Codde J, Playford D; NEDA Contributing Sites. Threshold of Pulmonary Hypertension Associated With Increased Mortality. J Am Coll Cardiol. 2019 Jun 4;73(21):2660-2672. doi: 10.1016/j.jacc.2019.03.482. PMID 31146810
• Choudhary G, Jankowich M, Wu WC. Elevated pulmonary artery systolic pressure predicts heart failure admissions in African Americans: Jackson Heart Study. Circ Heart Fail. 2014 Jul;7(4):558-64. doi: 10.1161/CIRCHEARTFAILURE.114.001366. Epub 2014 Jun 5. PMID 24902739
• Maron BA, Choudhary G, Khan UA, Jankowich MD, McChesney H, Ferrazzani SJ, Gaddam S, Sharma S, Opotowsky AR, Bhatt DL, Rocco TP, Aragam JR. Clinical profile and underdiagnosis of pulmonary hypertension in US veteran patients. Circ Heart Fail. 2013 Sep 1;6(5):906-12. doi: 10.1161/CIRCHEARTFAILURE.112.000091. Epub 2013 Jun 27. PMID 23811965
• Attia ZI, Kapa S, Lopez-Jimenez F, McKie PM, Ladewig DJ, Satam G, Pellikka PA, Enriquez-Sarano M, Noseworthy PA, Munger TM, Asirvatham SJ, Scott CG, Carter RE, Friedman PA. Screening for cardiac contractile dysfunction using an artificial intelligence-enabled electrocardiogram. Nat Med. 2019 Jan;25(1):70-74. doi: 10.1038/s41591-018-0240-2. Epub 2019 Jan 7. PMID 30617318
• Guo L, Khobragade N, Kieu S, Ilyas S, Nicely PN, Asiedu EK, Lima FV, Currie C, Lastowski E, Choudhary G. Development and Evaluation of a Deep Learning-Based Pulmonary Hypertension Screening Algorithm Using a Digital Stethoscope. J Am Heart Assoc. 2025 Feb 4;14(3):e036882. doi: 10.1161/JAHA.124.036882. Epub 2025 Feb 3. PMID 39895552

Идентификаторы

NCT: NCT07087613 · 2025.2

Первоисточники (государственные реестры)

• ClinicalTrials.gov (NIH, США)

Открыть это исследование на ClinicalTrials.gov ↗