Why the study?
BNP and pBNP predict cardiovascular morbidity and mortality, and patients needing intensive monitoring may benefit from an AI-ECG capable of predicting these peptides.
Does an AI-enabled ECG accurately predict BNP/NT-proBNP levels and future all-cause mortality?
Population
47,709 development, 16,249 tuning, 4001 internal validation, and 6042 external validation ECGs
Comparison
ECG-BNP vs ECG-pBNP for estimating peptide levels and predicting all-cause mortality
Design
AI model development and internal/external validation study
Key result
An artificial intelligence-enabled electrocardiogram accurately detected abnormal BNP/pBNP levels with AUCs ≥0.85, providing sensitivities of 68.0-85.0% and specificities of 77.9-86.2%.
Authors
Loading...
May support non-invasive BNP estimation via AI-ECG; hypothesis-generating and requires prospective validation before practice change.
Observational (n=74,001)
Does an AI-enabled ECG accurately predict BNP/NT-proBNP levels and future all-cause mortality?
Effect estimate: AUC ≥0.85
An AI-enabled ECG can accurately estimate BNP and NT-proBNP levels (AUC ≥0.85) and predict all-cause mortality, offering a non-invasive tool for cardiovascular risk monitoring.
Liu et al. (2023) conducted an observational in Cardiovascular diseases (n=74,001). Artificial intelligence-enabled electrocardiogram (AI-ECG) was evaluated on Detection of mild (≥500 pg/mL) and severe (≥1000 pg/mL) abnormal BNP/pBNP (AUC ≥0.85). An artificial intelligence-enabled electrocardiogram accurately detected abnormal BNP/pBNP levels with AUCs ≥0.85, providing sensitivities of 68.0-85.0% and specificities of 77.9-86.2%.