Synapse
⌘+K
Synapse
PulseExploreClubsResearchersJournals
Instagram
HomeClubsExplore
July 2, 2026European Heart JournalOpen Access

AI-ECG predicts mortality modestly better than ESC-SCORE in primary prevention but lags in secondary prevention.

View Full Paper
Ask AI
Bookmark
Share

Why the study?

Cardiovascular risk scoring systems rely on traditional risk variables rather than cardiac structure and function, motivating an AI approach to predict mortality using standard 12-lead ECG data.

Does an AI model using a single 12-lead ECG predict long-term mortality comparably to the ESC-SCORE in patients with suspected chronic coronary syndrome?

Population

720 patients scheduled for invasive coronary angiography for suspected chronic coronary syndrome

Comparison

ECG-based AI model vs ESC-SCORE

Design

Registry-based cohort study

Key result

An AI model using a single 12-lead ECG predicted long-term mortality with an AUROC of 0.606 (vs 0.584 for ESC-SCORE) in primary prevention and 0.612 (vs 0.658) in secondary prevention.

Authors

SWS WegenerDGD GruenJPJoshua Prim

Discussion

Loading...

Member takes

Overview

AI-ECG modestly outperforms ESC-SCORE for mortality in primary prevention but underperforms in secondary; extends ECG-AI risk tools while needing validation.

Key Points

  • The aim is to predict mortality in cardiovascular patients using artificial intelligence and ECG data.
  • Analyzed data from 720 patients enrolled in a registry for coronary angiography.
  • Utilized a deep learning model previously trained on the PTB-XL dataset to predict mortality from ECG recordings.
  • Compared AI model performance to traditional risk scoring systems like ESC-SCORE.
  • For patients without CAD, the AI model achieved an AUROC of 0.606, compared to the ESC-SCORE at 0.584.
  • For patients with CAD, the AI model achieved an AUROC of 0.612, compared to the ESC-SCORE at 0.658.

Study Design

Type

Cohort (n=720)

Structured PICO

Does an AI model using a single 12-lead ECG predict long-term mortality comparably to the ESC-SCORE in patients with suspected chronic coronary syndrome?

P
Population
720 patients scheduled for invasive coronary angiography for suspected chronic coronary syndrome, with available ECG and ESC-SCORE variables, followed for long-term mortality.
E
Exposure
Deep learning artificial intelligence model applied to a single 12-lead ECG obtained at admission.
C
Comparator
ESC-SCORE (weighed for a German population).
O
Outcome
Long-term overall mortality.hard clinical

Main Result

Effect estimate: AUROC 0.606 vs 0.584 (primary prevention); 0.612 vs 0.658 (secondary prevention)

Cite This Study

Wegener et al. (2021) conducted a cohort in Suspected chronic coronary syndrome (n=720). AI model based on a single 12-lead ECG vs. ESC-SCORE was evaluated on Long-term mortality (AUROC 0.606 vs 0.584 (primary prevention); 0.612 vs 0.658 (secondary prevention)). An AI model using a single 12-lead ECG predicted long-term mortality with an AUROC of 0.606 (vs 0.584 for ESC-SCORE) in primary prevention and 0.612 (vs 0.658) in secondary prevention.

synapsesocial.com/papers/6a468172aa56e3314088d014https://doi.org/10.1093/eurheartj/ehab724.1132
View Full Paper
Ask AI
Bookmark
Share