Phenotypic Selectivity of Artificial Intelligence–Enhanced Electrocardiography in Cardiovascular Diagnosis and Risk Prediction
Authors' conclusion
Despite being developed to detect specific cardiovascular conditions, AI-ECG models detect the presence and predict the future development of a broad range of cardiovascular diseases with similar propensity.
Study context
- Population
- Individuals with an ECG from EHR and prospective cohort studies
- Participants
- 233,689 individuals4 study populations
- Outcome
- Selectivity of phenotypic associations of AI-ECG models (condition-specific classifiers vs broader CV risk markers)
Methodological details4 study populations (EHR and prospective cohorts), cross-sectional and prospective analyses, external data sets, 6 image-based AI-ECG models: LV systolic dysfunction, AS, mitral regurgitation, LV hypertrophy, structural heart disease composite, and biological sex (negative control); 6 experimental models for noncardiovascular conditions, Logistic regression for cross-sectional phenotypes; Cox regression for new-onset CVD
- 4 study populations (EHR and prospective cohorts)
- Cross-sectional and prospective analyses
- External data sets
- 6 image-based AI-ECG models: LV systolic dysfunction, AS, mitral regurgitation, LV hypertrophy, structural heart disease composite, and biological sex (negative control); 6 experimental models for noncardiovascular conditions
- Logistic regression for cross-sectional phenotypes; Cox regression for new-onset CVD
Main finding
As reported in the abstract
Association of AI-ECG models with phenotype groups
5 structural/functional AI-ECG models more likely associated with CV phenotypes than other phenotype groups (OR range)
2.16 to 4.41, P<10⁻⁶
Correlation of phenotype association patterns between models
0.67-0.96
Observational data: associations, not effects.
Original publication
Phenotypic Selectivity of Artificial Intelligence–Enhanced Electrocardiography in Cardiovascular Diagnosis and Risk Prediction. Circulation. Published 1 September 2025. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →