Cardiology · AI-assisted cardiac auscultation (murmur detection and characterisation)
Eko Murmur Analysis Software (EMAS)
Eko Health (Eko Devices)
A murmur caught at the bedside is cardiology's cheapest early warning, and EMAS turns any Eko stethoscope into a second set of ears for it. FDA-cleared for adults and children, validated in JAHA, and — unusually for this catalogue — tested independently against echocardiography, which found real limits alongside real skill.
Performance Metrics
Clinical Evidence
The evidence base has three layers, and the honest read requires all of them. The foundation is Chorba and colleagues (Journal of the American Heart Association, 2021), a multisite study that recruited 962 patients and recorded at the four standard auscultation positions. The deep-learning murmur detector reached a sensitivity of 76.3 percent and specificity of 91.4 percent; excluding the softest grade-1 murmurs, sensitivity rose to 90.0 percent. At the appropriate listening post the algorithm detected moderate-to-severe or greater aortic stenosis with 93.2 percent sensitivity and 86.0 percent specificity, and moderate-to-severe or greater mitral regurgitation at 66.2 and 94.6 percent. The pivotal validation of the cleared EMAS suite is Prince and colleagues (Journal of the American Heart Association, 2023): algorithms trained on more than 15,000 heart-sound recordings, validated on 2,375 recordings from 615 subjects. Structural murmurs were detected with 85.6 percent sensitivity and 84.4 percent specificity, and the algorithm's 84.7 percent accuracy beat the 77.9 percent averaged by clinicians reading the same recordings. Both studies carry manufacturer-affiliated authors and should be read as such. The third layer is the one most catalogue entries lack: an external evaluation independent of Eko. Roquemen-Echeverri and colleagues (International Journal of Cardiology, 2025) recorded phonocardiograms from 1,029 patients immediately before transthoracic echocardiography at a single academic centre; only 79 percent of the 4,081 recordings were analysable by the software. Against the echocardiographic reference standard, EMAS reached an overall sensitivity of 39.3 percent (95 percent confidence interval 37.2 to 41.3) and specificity of 82.3 percent (80.0 to 84.5) for valvular heart disease. Per lesion, sensitivity was 88.9 percent for moderate-severe and severe aortic stenosis, 75.0 percent for aortic regurgitation, 62.5 percent for mitral stenosis and 63.3 percent for mitral regurgitation; the authors concluded overall sensitivity was too limited for use as a general screening tool while supporting detection of specific severe lesions. The study is independent of the manufacturer, but not disinterested: three authors declare a financial interest in Tellunostics LLC, and one is its employee. A published exchange followed — a manufacturer-authored critique by an Eko Health author holding company equity (International Journal of Cardiology 2025, PMID 40233833) and the study authors' reply (PMID 40294801) — which is noted here without adjudication. The consistent picture is a competent murmur flag whose strength is concentrated where murmurs are loudest and most consequential.
| Study | Design | n | Sensitivity | Specificity | AUC | Published |
|---|---|---|---|---|---|---|
| Chorba JS, et al. (multisite; manufacturer-affiliated co-authors) | ProspectiveProspective | 962 | 76.3% (murmur detection; 90.0% excluding grade-1 murmurs) | 91.4% | — | J Am Heart Assoc 2021;10(9):e019905 (PMID 33899504). 962 patients recorded at the four standard auscultation positions; AS (moderate-severe+) detected at 93.2%/86.0% sensitivity/specificity, MR (moderate-severe+) at 66.2%/94.6%. |
| Prince J, Maidens J, Kieu S, Currie C, Barbosa D, et al. (manufacturer-affiliated) | RetrospectiveRetrospective | 615 | 85.6% (structural murmurs) | 84.4% | — | J Am Heart Assoc 2023;12(20):e030377 (PMID 37830333; PMC10757522). Pivotal validation of the FDA-cleared EMAS suite on 2,375 recordings from 615 subjects; algorithm accuracy 84.7% vs 77.9% average clinician accuracy. |
| Roquemen-Echeverri V, Jacobs PG, Shalen EF, Schulman PM, Heitner SB (independent of Eko; authors declare interests in Tellunostics LLC) | ProspectiveProspective | 1,029 | 39.3% overall VHD (95% CI 37.2–41.3); per lesion AS 88.9%, AR 75.0%, MS 62.5%, MR 63.3% (moderate-severe+) | 82.3% (95% CI 80.0–84.5) | — | Int J Cardiol 2025;419:132653 (PMID 39433158). External evaluation at a single academic centre, independent of the manufacturer: phonocardiograms recorded with EkoDUO/EkoCORE stethoscopes immediately before transthoracic echocardiography in 1,029 patients; 79% of 4,081 recordings were analysable. Conflicts declared: Heitner, Schulman and Jacobs hold a financial interest in Tellunostics LLC, of which one author is an employee. A published exchange followed — a manufacturer-authored critique by McDonough RV of Eko Health, who holds company equity (Int J Cardiol 2025;432:133260, PMID 40233833), and the study authors' reply (Int J Cardiol 2025, PMID 40294801). |
Inside the algorithm
Editorial featureHow EMAS reaches a verdict.
Five stages — from raw input to verdict — drawn from manufacturer documentation and the public regulatory record.
- INGEST
- NORMALISE
- DETECT
- LOCALISE
- VERDICT
Stage 01 · INGEST
A recording from the stethoscope already in hand.
EMAS begins where every physical exam already does. The clinician auscultates with an Eko electronic stethoscope — CORE or DUO class hardware — which captures a phonocardiogram of the heart sounds and, on devices that provide one, a single-lead ECG alongside. The recording uploads to Eko's cloud, where the analysis runs. No new imaging, no new appointment: the input is the routine exam itself.
Input
Heart-sound recording, per auscultation position
Inference
Cloud-based
Inside the Auris+ Listing
Five more sections complete this device’s Auris+ Listing.
Decision Ledger
ProPro unlocks a private, cross-vendor log of every case you read on this device — what the AI called, what you concluded, and a one-line reason if you overrode it. Ready for the EU AI Act's deployer logging obligations when they land in 2028.
Clinical Evidence Deep Dive
ProPro unlocks the structured clinical-evidence summary — study count, target patient population, and a tabular accuracy-metrics view drawn from peer-reviewed sources.
Peer-Reviewed Publications
ProPro unlocks the curated peer-reviewed publication list with PubMed cross-links — the citation backbone of every editorial verdict.
Post-Market & Regulatory Conditions
ProPro unlocks the post-market surveillance summary, recall record, and the conditions of approval that bound real-world use.
AI Algorithm Version History
ProPro unlocks the chronological record of algorithm version changes — what changed when, drawn from manufacturer changelogs and regulatory filings.
Regulatory Approvals
K213794
Class II
Safety Record
No recall, FDA safety communication or indexed adverse-event report naming EMAS was located as of September 2026. An openFDA query returns no recall for Eko Devices and no MAUDE adverse-event report for EMAS; the MAUDE records returned for "eko" concern EKOS Corporation catheters, a name collision unrelated to this manufacturer. The cleared design bounds the risk: the software flags and characterises murmurs for a clinician who retains the diagnosis and the referral decision.
Intended Use & Indications
EMAS is prescription software for environments where healthcare is provided by clinicians. It analyses phonocardiogram recordings captured with Eko smart stethoscopes, together with single-lead ECG when the recording device provides one, and reports the presence or absence of a heart murmur, murmur timing, and murmur character — including S1, S2, innocent murmurs and structural murmurs — in adults and paediatric patients. The interpretation is advisory: it supports, and does not replace, the clinician's own auscultation and referral judgement.