Synapse
⌘+K
Synapse
PulseExploreClubsResearchersJournals
Instagram
HomeClubsExplore
June 7, 2026Cardiology Research and PracticeOpen Access

AI-ECG shows good discriminatory ability for detecting HFpEF with an AUROC of 0.84.

View Full Paper
Ask AI
Bookmark
Share

Why the study?

Diagnosis of HFpEF remains challenging and requires multimodal testing, whereas artificial intelligence applied to the electrocardiogram offers a low-cost, scalable method to detect HFpEF.

Does AI-enhanced electrocardiography accurately diagnose heart failure with preserved ejection fraction or left ventricular diastolic dysfunction compared to recognized reference standards?

Comparison

AI/ML models applied to ECGs vs recognized reference standard

Design

Systematic review and meta-analysis

Key result

Artificial intelligence applied to the electrocardiogram (AI-ECG) demonstrated good discriminatory ability for detecting HFpEF, with a pooled AUROC of 0.84 (95% CI 0.78-0.88).

Authors

CMCian P. MurrayHTHugo C. TemperleyRDRob S. Doyle

Discussion

Loading...

Member takes

Overview

May aid HFpEF screening via routine ECG; leaves open prospective validation before clinical adoption.

Key Points

  • This research aims to evaluate the utility of AI-enhanced electrocardiography in diagnosing heart failure with preserved ejection fraction (HFpEF).
  • Conducted a systematic review and meta-analysis following PRISMA guidelines, registered with PROSPERO.
  • Searched multiple databases for studies evaluating AI/ML models applied to ECGs for HFpEF diagnosis.
  • Pooled AUROC values from eligible studies using a random-effects model with risk of bias assessed.
  • Included ten studies with over 270,000 participants; seven studies contributed to AUROC pooling.
  • The pooled AUROC was 0.84 (95% CI 0.78–0.88), indicating good discriminatory ability despite extreme heterogeneity (I² = 100%).
  • Risk of bias ranged from moderate to high, influenced by selective cohorts and inconsistent standards.

Study Design

Type

Meta-Analysis (n=270,000)

Multicenter

Yes

Structured PICO

Does AI-enhanced electrocardiography accurately diagnose heart failure with preserved ejection fraction or left ventricular diastolic dysfunction compared to recognized reference standards?

P
Population
Over 270,000 participants across 10 studies evaluating AI-enhanced electrocardiography for the diagnosis of HFpEF or LVDD.
E
Exposure
Artificial intelligence/machine learning models applied to electrocardiograms (AI-ECG)
C
Comparator
Recognized reference standard
O
Outcome
Diagnostic performance measured by Area Under the Receiver Operating Characteristic curve (AUROC)surrogate

Main Result

Effect estimate: AUROC 0.84 (95% CI 0.78-0.88)

AI-enhanced ECG demonstrates good discriminatory ability for detecting HFpEF (pooled AUROC 0.84), but current evidence is limited by high heterogeneity, retrospective designs, and a lack of prospective clinical validation.

Limitations

  • Predominantly retrospective evidence base
  • Methodologically heterogeneous
  • Variable reference standards
  • Insufficient external validation
  • Moderate to high risk of bias in several domains
  • No prospective, outcome-based studies
  • Real-world implementation remains untested
  • Predominantly retrospective
  • Selective cohorts and incomplete reporting

Cite This Study

Murray et al. (2026) conducted a meta-analysis in Heart failure with preserved ejection fraction (HFpEF) (n=270,000). Artificial intelligence applied to the electrocardiogram (AI-ECG) vs. Recognized reference standard was evaluated on Diagnostic performance (AUROC) (AUROC 0.84, 95% CI 0.78-0.88). Artificial intelligence applied to the electrocardiogram (AI-ECG) demonstrated good discriminatory ability for detecting HFpEF, with a pooled AUROC of 0.84 (95% CI 0.78-0.88).

synapsesocial.com/papers/6a250b8b7def13d035e1b935https://doi.org/10.1155/crp/4994662
View Full Paper
Ask AI
Bookmark
Share