Development and External Validation of an AI-ECG Algorithm for Estimating Elevated Serum NT-proBNP Levels

Area of Science:

• Cardiology
• Biomarkers
• Artificial Intelligence

Background:

• N-terminal pro-B-type natriuretic peptide (NT-proBNP) is crucial for heart failure management but requires blood tests.
• Artificial intelligence (AI) offers a non-invasive alternative using electrocardiograms (ECGs).

Purpose of the Study:

• To develop and validate an AI model that estimates NT-proBNP levels from standard 12-lead ECGs.
• To assess the accuracy of AI-ECG in identifying elevated NT-proBNP.

Main Methods:

• A convolutional neural network with residual and attention layers was trained on 84,895 ECG-NT-proBNP pairs.
• Internal validation used 8,545 patients; external validation included 679 patients from two centers.
• Model performance was evaluated using AUROC for NT-proBNP thresholds (>250, >500, >1,000 pg/mL).

Main Results:

• The AI-ECG score strongly correlated with measured NT-proBNP (Spearman ρ=0.85) in internal validation.
• External validation showed high discrimination across thresholds (AUROC >0.866).
• Model performance remained consistent across various clinical subgroups.

Conclusions:

• An AI-enabled ECG model accurately identifies patients with elevated NT-proBNP levels.
• This approach can aid in selecting patients for confirmatory NT-proBNP testing.
• Further prospective studies are needed to confirm clinical utility and integration into care pathways.