Heart failure & cardiomyopathiesDiagnostic or prognostic study
Development and external validation of an AI-ECG algorithm for estimating elevated serum NT-proBNP levels
- Development and external validation of an AI-ECG algorithm
- convolutional neural network incorporating residual and attention-based layers
- external validation at two tertiary cardiovascular centres
Authors' conclusion
An AI-enabled ECG model can identify patients with elevated NT-proBNP levels with good accuracy in both internal and external validation cohorts. This approach may help identify patients who should undergo confirmatory NT-proBNP testing, particularly when biomarker testing is delayed, unavailable, or not routinely performed. Prospective studies are warranted to define whether this strategy provides incremental clinical value and can be integrated into clinical pathways.
Study context
- Population
- Adult patients with standard 12-lead electrocardiograms (ECGs) and NT-proBNP measurementsadult patients in training (84,895 ECG-NT-proBNP pairs) · internal validation 8,545 · external validation 679
- Participants
- 40,762 patients
- Comparison
- AI-ECG score (nine-level ECG-BNP score) against measured NT-proBNP levels at prespecified thresholds (>250, >500, >1000 pg/mL)
Performance
As reported in the abstract
Spearman ρ
0.85
P
P < 0.001
Further findings
Discrimination across thresholds (internal validation)
- AUROC
- > 0.92
- internal validation
Discrimination (external validation)
- AUROC for >250 pg/mL
- 0.866 (95% CI 0.838-0.894)
- AUROC for >500 pg/mL
- 0.882 (95% CI 0.857-0.907)
- AUROC for >1000 pg/mL
- 0.885 (95% CI 0.859-0.914)
- external validation
Performance across key clinical subgroups
- Performance was consistent across key clinical subgroups
Original publication
Summarised from the published abstract; every number is checked against it