Foundation Model Artificial Intelligence for Electrocardiogram Detection of Rheumatic Mitral Stenosis

Abstract

BackgroundRheumatic mitral stenosis (RMS) disproportionately affects low-resource settings where echocardiography is limited. We evaluated foundation model (FM) feasibility for RMS detection using limited, indirectly labeled ECGs.

MethodsAn ECG-FM was fine-tuned using 804 RMS-positive ECGs per fold and externally validated across clinical and wearable lead configurations.

ResultsInternal AUROCs were 0.89, 0.87, and 0.87, for 12-Lead, 6-Lead, and 1-Lead respectively. External AUROCs were 0.82, 0.80, and 0.77. Performance remained comparable across demographic subgroups.

ConclusionsFM fine-tuning enables RMS detection using limited ICD-derived labels, supporting further evaluation of reduced-lead ECG screening and echocardiography triage in resource-constrained settings.

Artificial Intelligence-Enabled Electrocardiogram Screening for Detection of Low Ejection Fraction in Routine Care: A Systematic Review and Meta-Analysis of Randomized Controlled Trials.
Ali Abdelhaleem Omar Ahmed, Omer Abdalhaleem Omer Ahmed, Rana Abdelhaleem Omar Ahmed +6
Annals of noninvasive electrocardiology : the official journal of the International Society for Holter and Noninvasive Electrocardiology, Inc28 Oct 2026Peer-reviewedCC BYFull textCardiovascular medicineEurope PMCPMC
External validation of artificial intelligence-guided electrocardiogram (AI-ECG) in detecting asymptomatic left ventricular dysfunction: A systematic review and meta-analysis
Jacky Chen, Thomas H Marwick, Cheng Hwee Soh
American journal of preventive cardiology30 May 2026Peer-reviewedCC BY-NC-NDFull textGeneral medicineEurope PMC
Hypertrophic cardiomyopathy phenotype is associated with postmyectomy change in diastolic function measured by artificial intelligence-enabled electrocardiogram.
Julia G Debertin, Hartzell V Schaff, Masoomeh Aslahishahri +9
JTCVS open2 Jul 2026Peer-reviewedCC BY-NC-NDFull textCardiovascular medicineEurope PMCPMC
Artificial Intelligence-Driven Nanobody Engineering for Pathogen Recognition: From Functional Epitope Prioritization to Multimodal Foundation Models
Rui-Si Hu, Muhammad Ehsan, Wei Luo +5
Biomolecules1 Oct 2026Peer-reviewedCC BY 4.0Full textMedical imaging & radiologyCrossrefEurope PMC
Advances in Multimodal Artificial Intelligence in Radiology: Data Integration, Foundation Models, and Clinical Applications-A Narrative Review.
Ghada Alfattni
Healthcare (Basel, Switzerland)29 Sep 2026Peer-reviewedCC BYFull textEpidemiology & public healthEurope PMC

Cite

BibTeX RIS

BibTeX

@misc{sriram2026foundation,
  title = {{Foundation Model Artificial Intelligence for Electrocardiogram Detection of Rheumatic Mitral Stenosis}},
  author = {R. Sriram and I. Nenadic and Z. Loring and M. Fudim and A. Mazella and A. K. Gehi and V. L. Murthy},
  howpublished = {medRxiv},
  year = {2026},
  month = oct,
  doi = {10.64898/2026.10.12.26365300},
  url = {https://www.medrxiv.org/content/10.64898/2026.10.12.26365300v1},
  note = {CC BY},
}

RIS

TY  - UNPB
TI  - Foundation Model Artificial Intelligence for Electrocardiogram Detection of Rheumatic Mitral Stenosis
AU  - R. Sriram
AU  - I. Nenadic
AU  - Z. Loring
AU  - M. Fudim
AU  - A. Mazella
AU  - A. K. Gehi
AU  - V. L. Murthy
PY  - 2026
DA  - 2026/10/12
PB  - medRxiv
DO  - 10.64898/2026.10.12.26365300
UR  - https://www.medrxiv.org/content/10.64898/2026.10.12.26365300v1
AB  - BackgroundRheumatic mitral stenosis (RMS) disproportionately affects low-resource settings where echocardiography is limited. We evaluated foundation model (FM) feasibility for RMS detection using limited, indirectly labeled ECGs.  MethodsAn ECG-FM was fine-tuned using 804 RMS-positive ECGs per fold and externally validated across clinical and wearable lead configurations.  ResultsInternal AUROCs were 0.89, 0.87, and 0.87, for 12-Lead, 6-Lead, and 1-Lead respectively. External AUROCs were 0.82, 0.80, and 0.77. Performance remained comparable across demographic subgroups.  ConclusionsFM fine-tuning enables RMS detection using limited ICD-derived labels, supporting further evaluation of reduced-lead ECG screening and echocardiography triage in resource-constrained settings.
ER  -

Abstract

Related

Cite