4338 ehjdh European Heart Journal. Digital Health Eur Heart J Digit Health Oxford University Press PMC12873438 12873438 12873438 41660663 10.1093/ehjdh/ztag003 Artificial-intelligence-enabled digital stethoscope improves point-of-care screening for moderate-to-severe valvular heart disease Rancier Moshe Writing - review & editing 1 Israel Igor Writing - review & editing 2 Monickam Vimalson Writing - review & editing 3 Currie Caroline Writing - review & editing 4 Verschoore Ben Writing - review & editing 5 Lastowski Emileigh Writing - review & editing 6 Van Pelt Douglas W Writing - review & editing 7 Prince John Writing - review & editing 8 McDonough Rosalie V Writing - review & editing 9 ✉ 2 1 Adult Primary Care, Mass General Brigham Community Physicians, Lawrence, MA, USA 2 Medicine, Parker Jewish Institute for Health Care and Rehabilitation, New Hyde Park, NY, USA 3 Medicine, Parker Jewish Institute for Health Care and Rehabilitation, New Hyde Park, NY, USA 4 Eko Health, Inc. 2100 Powell Street, Ste. 300, Emeryville, CA 94608, USA 5 Eko Health, Inc. 2100 Powell Street, Ste. 300, Emeryville, CA 94608, USA 6 Eko Health, Inc. 2100 Powell Street, Ste. 300, Emeryville, CA 94608, USA 7 Eko Health, Inc. 2100 Powell Street, Ste. 300, Emeryville, CA 94608, USA 8 Eko Health, Inc. 2100 Powell Street, Ste. 300, Emeryville, CA 94608, USA 9 Eko Health, Inc. 2100 Powell Street, Ste. 300, Emeryville, CA 94608, USA ✉ Corresponding author. Tel: +1 707 400 4327, Email: rosalie.mcdonough@ekohealth.com 2 Conflict of interest: C.C., B.V., E.L., D.W.V.P., J.P., and R.V.M.: current or former employees of Eko Health, Inc. Remaining authors have no relevant disclosures. 5 2 2026 7 2 ztag003 ztag003 6 2 2026 © The Author(s) 2026. Published by Oxford University Press on behalf of the European Society of Cardiology. This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial License ( https://creativecommons.org/licenses/by-nc/4.0/ ), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited. For commercial re-use, please contact reprints@oup.com for reprints and translation rights for reprints. All other permissions can be obtained through our RightsLink service via the Permissions link on the article page on our site—for further information please contact journals.permissions@oup.com. Abstract Aims Valvular heart disease (VHD) can affect more than one in two adults over 65 yet remains underdiagnosed due to the limitations of traditional auscultation. Earlier detection is critical for improving outcomes, but many cases go unrecognized. This study evaluates whether an artificial intelligence (AI)-enabled digital stethoscope can augment primary care providers’ (PCPs) ability to detect clinically significant VHD compared with analogue auscultation alone. Methods and results In this prospective study, 357 patients aged ≥50 years at risk for heart disease underwent both analogue cardiac auscultation by PCPs (standard of care, SOC) and digital cardiac auscultation by study coordinators followed by AI analysis using an electronic stethoscope (AI-augmented). Echocardiography and audible murmur annotation served as the reference standard. Sensitivity and specificity of AI vs. SOC were compared using Fisher’s exact test. The AI-augmented system demonstrated significantly higher sensitivity (92.3% vs. 46.2%, P = 0.01) but lower specificity (86.9% vs. 95.6%, P < 0.001) compared with SOC. Artificial intelligence detected 12 cases of previously undiagnosed mod+ VHD, while routine auscultation identified 6. Conclusion Artificial intelligence–enabled digital stethoscopes significantly improve point-of-care VHD detection, offering a promising tool for earlier diagnosis and intervention in primary care settings. Keywords: Valvular heart disease, Structural murmur, Artificial intelligence, Digital stethoscope, Primary care, Early detection Graphical Abstract Graphical Abstract status released display-pdf yes is-olf no is-manuscript no is-preprint no is-journal-matter no is-scanned no is-retracted no Received 2025 May 12; Revised 2025 Sep 12; Accepted 2025 Sep 22; Collection date 2026 Mar. Undiagnosed valvular heart disease (VHD) of any kind can be present in over 50% of individuals 65 and over, with moderate or worse severity (mod+) affecting 6.4% of that population. 1 It is associated with reduced functional capacity, arrhythmia, heart failure, increased hospitalization, and death. 1 , 2 Unfortunately, symptoms of VHD are frequently non-specific, and more than half of the patients with moderate-to-severe disease are asymptomatic. 3 This often leads to delayed diagnosis and treatment, which, in addition to reduced quality of life and increased mortality, poses a substantial financial burden to healthcare systems. 1 Murmurs are often an early clinical indicator of underlying VHD. However, cardiac auscultation, the current point-of-care clinical standard for VHD screening, has shown a 44% sensitivity when performed by experienced general practitioners for detecting significant VHD in asymptomatic patients, leaving many undiagnosed. 4 This limitation in traditional cardiac auscultation contributes to delayed or missed diagnoses of clinically significant VHD, which can result in rapid disease progression, heart failure, and increased mortality, even in initially asymptomatic patients. 2 , 5 More accurate diagnostic methods are therefore of critical importance. This research letter describes the real-world performance of a Food and Drug Administration–cleared suite of convolutional neural network-based artificial intelligence (AI) algorithms that detect and classify cardiac murmurs using a digital stethoscope. 6 From June 2021 to May 2023, this single-arm, single-blinded, prospective study consecutively enrolled 357 patients aged 50 and older at risk for, or with known, heart disease without prior VHD diagnosis or history of murmur from three primary care clinics. Eligible participants had one or more of the following: hypertension, body mass index ≥30, diabetes mellitus, hyperlipidaemia, atrial fibrillation, myocardial infarction, stroke/transient ischaemic attack, previous coronary surgery, or previous coronary angiography. Following collection of demographic and clinical data, a four-point cardiac auscultation was performed on each patient by (i) primary care practitioners (PCPs) using their personal, non-digital stethoscopes according to the standard of care (SOC) and (ii) trained study coordinators equipped with digital stethoscopes to collect phonocardiogram (PCG) data for subsequent AI analysis (AI-augmented). Five general practice clinicians (four MDs and one nurse practitioner) formed the group of PCPs. All procedures and protocols were approved by the Salus Institutional Review Board, and the study is registered at clinicaltrials.gov ( NCT05459545 ). An examination was labelled ‘positive’ if a murmur was detected at any auscultation point. Each patient underwent an echocardiogram by a certified technician to confirm whether mod+ VHD was present, and their PCG recordings were independently reviewed by an outside expert panel to confirm whether audible murmurs were present. Primary care providers and an expert panel were blinded to AI and echocardiogram results. Ground truth was defined as ‘audible VHD’, which was the combination of confirmed, mod+ VHD per echocardiogram results and confirmed, audible murmur per expert auscultation panel. Fisher’s exact test compared the sensitivity and specificity of the AI-augmented vs. SOC auscultation. A separate analysis against a ground truth of echocardiographic mod+ VHD (i.e. comprising both audible and inaudible murmurs) was also performed. Demographic information was collected for 354 patients (99.2%). The median age of the enrolled population was 70 years (interquartile range: 66–77), with 219 (61.9%) being female. Two hundred and forty-eight (70%) were White, followed by 29 (8.2%) Black or African American, 8 (2.3%) Asian, 1 (0.3%) American Indian or Alaska Native, and 3 (0.8%) other. Sixty-five (18.4%) identified as Hispanic or Latino. The most common VHD risk factors in the study population were hypertension (81.8%, 292), hyperlipidaemia (70.3%, 251), and diabetes mellitus (39.5%, 141). Figure 1 demonstrates the performance of the AI and the PCPs at detecting patients with audible VHD (Panels A and B) and echocardiographically confirmed VHD (Panels C and D), respectively. Figure 1 Performance of SOC auscultation vs. AI-augmented auscultation for detection of structural heart disease. (A, B) Performance of SOC auscultation (A) and AI-augmented auscultation (B) for detection of audible structural murmurs, using ground truth defined by echocardiographic evidence of structural disease and audible murmur adjudication by an expert panel. (C, D) Performance of SOC auscultation (C) and AI-augmented auscultation (D) for detection of moderate-to-severe valvular heart disease, using echocardiography as the ground truth reference standard. AI, artificial intelligence; ASM, audible structural murmur; SOC, standard of care. Figure 1. Performance of standard of care (SOC) auscultation versus AI-augmented auscultation for detection of structural heart disease. A-B: Performance of SOC auscultation (A) and AI-augmented auscultation (B) for detection of audible structural murmurs, using ground truth defined by echocardiographic evidence of structural disease and audible murmur adjudication by an expert panel. C-D: Performance of SOC auscultation (C) and AI-augmented auscultation (D) for detection of moderate-to-severe valvular heart disease, using echocardiography as the ground truth reference standard. ASM, audible structural murmur; SOC, standard of care. With AI augmentation, sensitivity for detecting audible VHD more than doubled (46.2–92.3%, P = 0.01). For echocardiographically confirmed VHD, it nearly tripled (13.8–39.7%, P = 0.01), with only modest reductions in specificity (95.6–86.9 and 94.7–88.6%, respectively, P < 0.001). Further, AI-augmented auscultation identified 12 patients with previously undiagnosed clinically significant VHD compared with 6 identified by the SOC. Our study demonstrates that an AI-enabled digital stethoscope significantly outperforms SOC auscultation in the overall detection of VHD. The findings are consistent with previous implementations on independent datasets, further validating its ability to generalize across a wide range of clinical environments. 6 These results suggest that the implementation of digital stethoscopes with structural murmur detection AI into primary care settings can facilitate earlier detection of previously unrecognized VHD, an essential first step to timely referral, monitoring, and potential intervention. While earlier detection does not inherently lead to better outcomes, it may help prevent irreversible cardiac damage and reduce long-term morbidity when followed by appropriate patient management. 5 , 7–9 A notable trade-off is the reduction in specificity, which may increase false positives and subsequent echocardiography referrals. While this could contribute to short-term resource use, it must be weighed against the value of earlier detection, particularly in asymptomatic patients who may otherwise progress to advanced disease undiagnosed. We acknowledge this trade-off and emphasize the need for further studies to assess the cost-effectiveness of AI-augmented auscultation in routine care. Limitations must be acknowledged. The study's sample size was relatively small, which limited the ability to perform more detailed subgroup analyses. As a result, the findings may not capture the full range of outcomes in all patient subgroups. Additionally, the study was conducted in three primary care clinics within the same geographic region, which may limit generalizability. Symptom status at enrolment was not systematically captured, which may affect interpretation. Finally, demographic data were unavailable for some participants. Despite this, our study provides valuable insights into the immediate clinical impact of a commercially available, novel, and easy-to-use point-of-care solution in a real-world setting. Contributor Information Moshe Rancier, Adult Primary Care, Mass General Brigham Community Physicians, Lawrence, MA, USA. Igor Israel, Medicine, Parker Jewish Institute for Health Care and Rehabilitation, New Hyde Park, NY, USA. Vimalson Monickam, Medicine, Parker Jewish Institute for Health Care and Rehabilitation, New Hyde Park, NY, USA. Caroline Currie, Eko Health, Inc. 2100 Powell Street, Ste. 300, Emeryville, CA 94608, USA. Ben Verschoore, Eko Health, Inc. 2100 Powell Street, Ste. 300, Emeryville, CA 94608, USA. Emileigh Lastowski, Eko Health, Inc. 2100 Powell Street, Ste. 300, Emeryville, CA 94608, USA. Douglas W Van Pelt, Eko Health, Inc. 2100 Powell Street, Ste. 300, Emeryville, CA 94608, USA. John Prince, Eko Health, Inc. 2100 Powell Street, Ste. 300, Emeryville, CA 94608, USA. Rosalie V McDonough, Eko Health, Inc. 2100 Powell Street, Ste. 300, Emeryville, CA 94608, USA. Author contributions Moshe Rancier (Conceptualization, Formal analysis, Investigation [lead]), Igor Israel (Investigation, Writing—review & editing [supporting]), Caroline Currie (Conceptualization, Project administration, Writing—review & editing [supporting]), Ben Verschoore (Project administration, Writing—review & editing [supporting]), Emileigh Lastowski (Project administration, Writing—review & editing [supporting]), Douglas W. Van Pelt (Writing—review & editing [supporting]), John Prince (Formal analysis, Writing—review & editing [supporting]), Rosalie V. McDonough (Data curation, Formal analysis, Writing—original draft, Visualization [lead]), and Vimalson Monickam (Writing—review & editing [supporting]) Funding This material is based upon work supported by the National Science Foundation under Grant No. 2R44HL144297-02. Data availability Data that support the findings of this study are available from the corresponding author upon reasonable request. References 1. d'Arcy JL, Coffey S, Loudon MA, Kennedy A, Pearson-Stuttard J, Birks J, et al. Large-scale community echocardiographic screening reveals a major burden of undiagnosed valvular heart disease in older people: the OxVALVE Population Cohort Study. Eur Heart J 2016;37:3515–3522. 2. Gada H, Vora A, Ramlawi B, Sotelo M, Wagner L, Rogers C, et al. Increased risk of clinical outcomes in moderate aortic stenosis patients. J Am Coll Cardiol 2021;77:971–971. 3. Généreux P, Stone GW, O'Gara PT, Marquis-Gravel G, Redfors B, Giustino G, et al. Natural history, diagnostic approaches, and therapeutic strategies for patients with asymptomatic severe aortic stenosis. J Am Coll Cardiol 2016;67:2263–2288. 4. Gardezi SKM, Myerson SG, Chambers J, Coffey S, d’Arcy J, Hobbs FDR, et al. Cardiac auscultation poorly predicts the presence of valvular heart disease in asymptomatic primary care patients. Heart. 2018;104:1832–1835 5. Nishimura RA, Ommen SR, Dearani JA, Schaff HV. Valvular heart disease—a new evolving paradigm. Mayo Clin Proc 2025;100:358–379. 6. Prince J, Maidens J, Kieu S, Currie C, Barbosa D, Hitchcock C, et al. Deep learning algorithms to detect murmurs associated with structural heart disease. J Am Heart Assoc 2023;12:e030377. 7. Nishimura R, O’Gara P, Bavaria J, Brindis RG, Carroll JD, Kavinsky CJ, et al. 2019 AATS/ACC/ASE/SCAI/STS Expert Consensus Systems of Care Document: a proposal to optimize care for patients with valvular heart disease: a joint report of the American Association for Thoracic Surgery, American College of Cardiology, American Society of Echocardiography, Society for Cardiovascular Angiography and Interventions, and Society of Thoracic Surgeons. J Am Coll Cardiol 2019;73:2609–2635. 8. Ran Z, Wang Y, Cao T, He S, Li X, Guo Y. Benefits of heart valve clinics for patients: a systematic review. BMJ Open 2025;15:e096538. 9. Jneid H, Chikwe J, Arnold SV, Bonow RO, Bradley SM, Chen EP, et al. 2024 ACC/AHA clinical performance and quality measures for adults with valvular and structural heart disease: a report of the American Heart Association/American College of Cardiology joint committee on performance measures. J Am Coll Cardiol 2024;83:1579–1613. Associated Data Data Availability Statement Data that support the findings of this study are available from the corresponding author upon reasonable request.