Abstract 4370574: Artificial Intelligence-Enabled Electrocardiography Demonstrates Strong Diagnostic Performance for Diastolic Dysfunction, Heart Failure with Preserved Ejection Fraction, and Left Atrial Enlargement: A Meta-Analysis
Bibliographic record
Abstract
Introduction: Artificial intelligence (AI)-enabled electrocardiography (ECG) is redefining cardiac diagnostics by uncovering structural and functional abnormalities that human clinicians may not detect. Notably, AI models have shown promise in characterizing left ventricular diastolic dysfunction (LVDD), heart failure with preserved ejection fraction (HFpEF), and left atrial enlargement (LAE). However, the diagnostic performance of these models varies across studies. This meta-analysis evaluates the pooled performance of ECG-based AI models in detecting LVDD, HFpEF, and LAE. Research Question: What is the pooled diagnostic accuracy of AI-enabled ECG models for detecting (1) LVDD, (2) HFpEF, and (3) LAE? Methods: We systematically searched MEDLINE and Embase through May 2025 for studies evaluating AI-based ECG models for the diagnosis of LVDD, HFpEF, or LAE. Studies were included if they reported or allowed derivation of 2×2 confusion matrix data. A bivariate random-effects meta-analysis was conducted to pool sensitivity, specify, and area under the summary receiver operating characteristic (SROC) curve. Heterogeneity was quantified using sample size-adjusted I 2 estimates. Results: ECG-based AI models demonstrated strong diagnostic accuracy for LVDD, HFpEF, and LAE. Seven studies assessed models for detecting LVDD and yielded a pooled sensitivity of 0.82 (95% CI: 0.81–0.84) and a specificity of 0.75 (95% CI: 0.62–0.85), with an area under the SROC curve (AUC) of 0.83. Between-study heterogeneity was minimal, with sample size–adjusted I 2 estimates ranging from 1.5% to 3.1%. For HFpEF, three studies yielded a pooled sensitivity of 0.88 (95% CI: 0.77–0.95) and a specificity of 0.65 (95% CI: 0.42–0.82), with an AUC of 0.86. Between-study heterogeneity was minimal, with sample size–adjusted I 2 estimates ranging from 0.7% to 2.2%. Similarly, for left atrial enlargement (LAE), three studies yielded a pooled sensitivity of 0.83 (95% CI: 0.80–0.85) and a specificity of 0.80 (95% CI: 0.67–0.89), with an AUC of 0.86. Between-study heterogeneity was minimal, with sample size–adjusted I 2 estimated at 1.5%. Conclusion: AI-enabled ECG models demonstrate robust diagnostic performance for detecting LVDD, HFpEF, and LAE, with strong overall accuracy. These findings highlight the potential of AI as a scalable tool to enhance early detection of structural heart disease and support broader adoption of AI-driven screening in clinical practice.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".