AI-enhanced auscultation for left ventricular dysfunction, validated in prospective Japanese cohorts
An AI model reading digital heart sounds from four chest sites with a single-lead ECG detected an ejection fraction of 40% or less with an AUROC of 0.904 in an external prospective cohort of 964 patients, and was more accurate than a consensus of seven cardiologists in a 100-patient reader substudy (82.0% against 62.0%).
- Model
- Transformer and one-dimensional CNN ensemble over four-site digital auscultation and single-lead ECG
- Field
- Medicine
- Date
- 2026-10-15
- Human collaborators
- Eiichiro Sato, Nobuyuki Kagiyama
What was found
Adults referred for clinically indicated echocardiography at multiple Japanese centres underwent digital auscultation with a simultaneous single-lead ECG, with same-visit echocardiography as the reference. The ensemble was trained on 2,359 patients, 8.9% with reduced ejection fraction, and validated on 964 from a separate multicentre registry, 7.6% with it. External sensitivity, specificity and accuracy were 86.3%, 74.2% and 75.1%. The model outperformed ECG-only models and clinical regression models including B-type natriuretic peptide, and a simplified two-site version kept an AUROC of 0.903.
Novelty check
Europe PMC, 2026-10-29: TITLE_ABS:"auscultation" AND TITLE_ABS:"artificial intelligence" AND TITLE_ABS:"ventricular" returned 10 records, including a 2026 study of a digital stethoscope's single-lead ECG with AI for low ejection fraction and a 2026 study of AI ECG with digital auscultation for peripartum cardiomyopathy. AI analysis of stethoscope recordings for low ejection fraction is established; this adds multi-site heart-sound analysis validated prospectively and against a cardiologist panel.
Caveats and known objections
Funded by AMI Inc., whose staff are among the 17 authors, and by a Japan Society for the Promotion of Science grant; the external registry is a separate multicentre cohort, but the validation is the developer's, not independent in this registry's sense. Prospective in enrolment, but the endpoint is detection against echocardiography in patients already referred for it, so prevalence and case mix differ from a screening population, and the authors call for further validation before screening use. The reader comparison used seven cardiologists on 100 patients. Autonomy is ai-assisted: the output is proposed as risk enrichment before echocardiography, a triage step clinicians act on.
No independent check is recorded on this entry yet.
Reading the primary source closely enough to say whether it supports the claim counts as a check, and you are credited on the entry.
Or on GitHub: submit a check challenge the grade send a correction or send a pull request
Flag this for triage
Signals order the review queue and nothing else. They are never published, and they never move a grade: that takes a citation.
Entry history (1 event)
- AddedEntered the registry graded Peer reviewed and AI-assisted.
Entries are never deleted. A grade that does not hold up is downgraded on the record, with the reason beside it.
Graded peer reviewed for verification and ai-assisted for autonomy. What these mean.
Cite this entry
whataifound.org. (2026). AI-enhanced auscultation for left ventricular dysfunction, validated in prospective Japanese cohorts. whataifound.org: A Registry of AI Scientific and Mathematical Discoveries. https://whataifound.org/finding/2026-10-15-ai-auscultation-lv-dysfunctionBibTeX
@misc{whataifound-juntendouniversity-2026-dysfunction,
title = {AI-enhanced auscultation for left ventricular dysfunction, validated in prospective Japanese cohorts},
author = {{whataifound.org}},
year = {2026},
howpublished = {whataifound.org: A Registry of AI Scientific and Mathematical Discoveries},
note = {Result by Juntendo University (with AMI Inc.). Verification: Peer reviewed. Autonomy: AI-assisted.},
url = {https://whataifound.org/finding/2026-10-15-ai-auscultation-lv-dysfunction}
}