Truace
Subscribe

AI reading heart ultrasound spots a hard-to-diagnose type of heart failure with good accuracy

Study title: Artificial Intelligence Applied to Echocardiography for Detection of Heart Failure with Preserved Ejection Fraction and Elevated Filling Pressures: A Diagnostic Test Accuracy Meta-Analysis

Researchers pooled results from eight studies covering 2770 participants to measure how well AI models applied to echocardiograms identify heart failure with preserved ejection fraction, elevated left ventricular filling pressure, or diastolic dysfunction. Databases were searched through June 2026 and models were compared with invasive or guideline-based reference standards.

Read the study abstract

Heart failure with preserved ejection fraction (HFpEF) accounts for nearly half of heart failure cases and remains difficult to diagnose non-invasively. Artificial intelligence (AI) applied to echocardiography may improve detection, but its aggregate diagnostic accuracy has not been quantified. We performed a diagnostic test accuracy meta-analysis of AI-enabled echocardiography for identifying HFpEF, elevated left ventricular filling pressure, or diastolic dysfunction. MEDLINE/PubMed, Embase, and CENTRAL were se…

By the Truvace desk1 min read1 sourceEvidence: Moderate · 70/100 — Systematic review or meta-analysis · n = 2,770 · 1 independent sourcePeer-reviewedTRV-2026-2311· cite & verify
Artificial Intelligence Applied to Echocardiography for Detection of Heart Failure with Preserved Ejection Fraction and Elevated Filling Pressures: A Diagnostic Test Accuracy Meta-Analysis
US Army 51682 090917-A-4142B-002 by U.S. Army photo by Spc. Jazz Burney, 3rd Infantry Brigade Combat Team, 25th Infantry Division Public Affairs. Public domain

In brief

The finding matters because HFpEF makes up nearly half of heart failure cases and is difficult to confirm without invasive testing. Yet the evidence base is small and inconsistent, so it remains unclear how the tools would perform in routine care across different reference standards and patient groups.

Main points

  1. Eight studies with 2770 participants were included, and 1317 (47.5%) had the target condition.
  2. The area under the summary ROC curve was 0.875, and the diagnostic odds ratio was 24.73.
  3. Positive and negative likelihood ratios were 4.84 and 0.20.
  4. Accuracy did not differ by AI architecture but varied significantly by reference standard.

The gain

Pooled across eight studies, AI applied to echocardiography detected HFpEF and elevated filling pressures with roughly 84% sensitivity and 83% specificity, offering a non-invasive route for a condition that is hard to diagnose.

The rundown

The team searched MEDLINE/PubMed, Embase, and CENTRAL from inception to June 2026 and included studies that tested AI models on echocardiography against invasive, guideline-adjudicated, or ASE/EACVI reference standards. Sensitivity and specificity were pooled with a bivariate random-effects model, and risk of bias was assessed with QUADAS-2.

Beyond pooled sensitivity and specificity, the analysis reported a summary ROC area of 0.875 and a diagnostic odds ratio of 24.73. No small-study effects were detected, suggesting publication bias was not evident in this set.

The authors conclude the results support diagnostic potential but call for larger, prospectively validated studies before routine clinical adoption, citing high heterogeneity and dependence on the reference standard.

Sources

  1. Peer-reviewedCardiology in Review2026-10-29

The debate