Canonical text (the exact bytes fingerprinted)

TRUVACE RECORD VERSION
record: TRV-2026-0773
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-08-15T06:22:31.207206Z
status: published
lens: g_space
sector: health
headline: Development and External Validation of an AI-ECG Algorithm for Estimating Elevated Serum NT-proBNP Levels
dek: Background N-terminal pro-B-type natriuretic peptide (NT-proBNP) is a cornerstone biomarker for the diagnosis and management of heart failure, but its use may be limited by the need for blood testing and laboratory infrastructure. Artificial intelligence (AI) applied to electrocardiograms (ECGs) may offer a widely accessible, non-invasive approach to estimate NT-proBNP levels. Methods We developed a convolutional neural network incorporating residual and attention-based layers to estimate NT-proBNP levels from s…
gain_title: A convolutional neural network using standard 12-lead ECGs can estimate elevated NT-proBNP levels with strong correlation and good discrimination in internal and external validation.
problem_title: (none)
trace_subject: (none)
gain_reading: A convolutional neural network using standard 12-lead ECGs can estimate elevated NT-proBNP levels with strong correlation and good discrimination in internal and external validation.
gain_evidence: An AI-enabled ECG model can identify patients with elevated NT-proBNP levels with good accuracy in both internal and external validation cohorts.
problem_reading: (none)
problem_evidence: (none)
quick_read: Investigators built and validated an AI model that estimates serum NT-proBNP levels from routine 12-lead ECGs, training on nearly 85,000 ECG-lab pairs and testing internally in over 8,500 patients and externally in 679 patients at two tertiary centers.

The approach could provide a non-invasive, widely accessible triage tool to flag patients who should undergo confirmatory NT-proBNP testing when lab testing is delayed or unavailable, but its real-world clinical utility and pathway integration remain unproven pending prospective studies.
limitation: Prospective evidence of incremental clinical value and integration into clinical pathways is still lacking, requiring further studies.
tag: Evidence-backed gain
key_points: Model trained on 84,895 ECG-NT-proBNP pairs from 40,762 adult patients with 8,545 patients held out for internal validation. | External validation conducted in 679 patients at two tertiary cardiovascular centers with consistent performance across key clinical subgroups. | Model outputs a nine-level ECG-BNP score evaluated at thresholds >250, >500, and >1,000 pg/mL with AUROC, calibration, sensitivity, specificity, PPV/NPV.
rundown: Researchers developed a convolutional neural network incorporating residual and attention-based layers to generate a nine-level ECG-BNP score from standard 12-lead ECGs, trained on 84,895 pairs from 40,762 adults.

Internal validation in 8,545 held-out patients showed Spearman c1=0.85 correlation, and external validation in 679 patients at two tertiary cardiovascular centers yielded AUROCs of 0.866 to 0.885 across thresholds, with consistent performance across subgroups.
sources:
- peer_reviewed | European Heart Journal - Quality of Care and Clinical Outcomes | https://doi.org/10.1093/ehjqcco/qcag128 | 2026-08-13
prev: 0000000000000000000000000000000000000000000000000000000000000000

sha256

2c20ca336834df859afd85ce484ea6d78d53d2a75669ceebdcbbcf3ee048db98

previous

0000000000000000000000000000000000000000000000000000000000000000