AI-ECG heart failure predictions linked to echocardiographic traits
A study of 8,147 patients at Akershus University Hospital (Jan 2023–Jun 2025) found that AI-ECG predictions of heart failure correlate with echocardiographic measures of myocardial dysfunction, remodeling, and filling pressures, even beyond LVEF. External validation included 36,286 ECG-echocardiography pairs.
Key facts
- 8,147 patients underwent ECG and echocardiography within three days at Akershus University Hospital between 1 January 2023 and 1 June 2025.
- A previously validated AI-ECG model for heart failure detection was applied to all ECGs.
- Spearman's rank correlation quantified associations between echocardiographic parameters and AI-ECG risk.
- Subgroup analyses were performed by sex and left ventricular ejection fraction (LVEF).
- External validation included 36,286 ECG-echocardiography pairs.
- AI-ECG can detect heart failure not captured by LVEF.
- The cardiac phenotypes underlying AI-ECG predictions were unclear before this study.
- The study retrospectively analysed ECG and echocardiography data.
Entities
Institutions
- Akershus University Hospital