PulseExploreJournal ClubResearchersJournals
Instagram
HomeJournal ClubExplore
Synapse
⌘+K
Synapse
August 15, 2026European Heart Journal - Quality of Care and Clinical Outcomes

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

View Full Paper
Ask AI
Bookmark
Share

Why the study?

NT-proBNP is a cornerstone biomarker for heart failure, but its use may be limited by the need for blood testing and laboratory infrastructure.

Does an AI-ECG algorithm accurately estimate elevated serum NT-proBNP levels in adult patients?

Population

40,762 adult patients for model training, 8,545 for internal validation, and 679 for external validation

Comparison

AI-ECG algorithm estimation vs measured serum NT-proBNP levels

Design

Model development and external validation study across two tertiary cardiovascular centers

Key result

An AI-enabled ECG model accurately identified patients with elevated NT-proBNP levels >250 pg/mL in external validation (AUROC 0.866; 95% CI 0.838-0.894).

Authors

CICiro IndolfiCSCarmen SpaccarotellaAPAlberto Polimeni

Discussion

Loading...

Member takes

Overview

May flag patients for confirmatory NT-proBNP testing when labs are unavailable; leaves open incremental value pending prospective studies.

Key Points

  • To develop and externally validate a deep learning model that non-invasively estimates serum NT-proBNP levels using standard 12-lead ECG signals.
  • Developed a convolutional neural network with residual and attention layers trained on 84,895 ECG–NT-proBNP pairs from 40,762 adult patients, using 8,545 held-out patients for internal validation to produce a nine-level ECG-BNP score.
  • Conducted external validation in 679 patients across two tertiary cardiovascular centers, assessing diagnostic discrimination (AUROC with 95% CIs), calibration, sensitivity, specificity, and predictive values across prespecified NT-proBNP cutoffs (>250, >500, and >1,000 pg/mL).
  • Internal validation revealed a strong correlation between predicted and measured NT-proBNP levels (Spearman ρ=0.85, p<0.001) and high discrimination across all thresholds (AUROC >0.92).
  • External validation demonstrated robust discrimination with AUROCs of 0.866 (95% CI 0.838–0.894) for >250 pg/mL, 0.882 (95% CI 0.857–0.907) for >500 pg/mL, and 0.885 (95% CI 0.859–0.914) for >1,000 pg/mL, with consistent subgroup performance.

Study Design

Type

Observational (n=41,441)

Multicenter

Yes

Structured PICO

Does an AI-ECG algorithm accurately estimate elevated serum NT-proBNP levels in adult patients?

P
Population
41,441 adult patients across training, internal validation, and external validation cohorts used to develop and validate an AI-ECG algorithm for estimating NT-proBNP levels.
E
Exposure
AI-ECG algorithm (convolutional neural network incorporating residual and attention-based layers) generating a nine-level ECG-BNP score
C
Comparator
Measured serum NT-proBNP levels via blood testing
O
Outcome
Discrimination for prespecified NT-proBNP thresholds (>250, >500, >1,000 pg/mL) assessed by AUROCsurrogate

Main Result

Effect estimate: AUROC 0.866 (95% CI 0.838-0.894)

An AI-enabled ECG model can accurately identify patients with elevated NT-proBNP levels, providing a widely accessible and non-invasive screening tool when blood testing is unavailable.

Limitations

  • Prospective studies are warranted to define whether this strategy provides incremental clinical value and can be integrated into clinical pathways.

Cite This Study

Indolfi et al. (2026) conducted an observational in Elevated NT-proBNP (n=41,441). AI-ECG algorithm vs. Measured NT-proBNP levels was evaluated on Discrimination for elevated NT-proBNP >250 pg/mL (AUROC 0.866, 95% CI 0.838-0.894). An AI-enabled ECG model accurately identified patients with elevated NT-proBNP levels >250 pg/mL in external validation (AUROC 0.866; 95% CI 0.838-0.894).

synapsesocial.com/papers/6a8019db75c2e31742c86108https://doi.org/10.1093/ehjqcco/qcag128
View Full Paper
Ask AI
Bookmark
Share

Also Consider

Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context:

  1. 1Artificial Intelligence-Enabled Electrocardiography Detects B-Type Natriuretic Peptide and N-Terminal Pro-Brain Natriuretic Peptide2023 · 10 citations
  2. 2Artificial intelligence-powered ECG is superior to NT-proBNP in predicting left ventricular systolic dysfunction in atrial fibrillation with rapid ventricular response2026
  3. 3Remote and noninvasive monitoring of childhood cancer survivors for elevated NT-proBNP using Apple Watch ECG2026
  4. 4Deep Learning for Cardiac Overload Estimation ― Predicting B-Type Natriuretic Peptide (BNP) Levels From Heart Sounds and Electrocardiogram ―2025 · 11 citations
  5. 5Detection of hospital-admitted heart failure regardless of ejection fraction using artificial intelligence-enhanced electrocardiograms2025 · 1 citations