Synapse
⌘+K
Synapse
PulseExploreClubsResearchersJournals
Instagram
HomeClubsExplore
September 6, 2026European Heart Journal - Digital HealthOpen Access

Automated deep learning echo interpretation matches manual analysis for HFpEF diagnostic accuracy.

View Full Paper
Ask AI
Bookmark
Share

Why the study?

Diagnosis of HFpEF using the HFA-PEFF and H2FPEF scores remains clinically challenging and relies on echocardiographic assessment.

Does automated deep learning interpretation of echocardiograms perform similarly to manual measurements in diagnosing HFpEF?

Population

102 test HFpEF patients, 129 ambulatory HFpEF patients, and 427 diagnostic validation patients (182 HFpEF, 245 non-HFpEF)

Comparison

Automated deep learning echocardiogram interpretation vs manual measurements for HFA-PEFF and H2FPEF scores

Design

Cohort validation study

Key result

Automated deep learning interpretation of echocardiograms yielded similar diagnostic accuracy to manual analysis for HFpEF (AUC for HFA-PEFF 0.70 vs 0.71; AUC for H2FPEF 0.78 vs 0.75).

Authors

CVConstantijn S. VenemaMVMax F.G.H.M. VennerAAAnouk Achten

Discussion

Loading...

Member takes

Overview

Automated scoring may match manual HFpEF diagnostic accuracy; extends deep learning echo validation but leaves prospective adoption open.

Key Points

  • To evaluate whether diagnostic scoring based on automated deep learning interpretation of echocardiograms matches the performance of manual measurements in diagnosing heart failure with preserved ejection fraction (HFpEF).
  • Analyzed echocardiograms using an automated deep learning algorithm versus manual measurements across three cohorts: a test cohort (n = 102), an ambulatory validation cohort (n = 129), and a diagnostic validation cohort (n = 427; 182 HFpEF and 245 non-HFpEF).
  • Calculated HFA-PEFF and H2FPEF diagnostic scores, assessed correlations with pulmonary capillary wedge pressure (PCWP), and compared diagnostic accuracy via area under the receiver operating characteristic curve (AUC).
  • Automated and manual scores showed high correlation across cohorts for HFA-PEFF (r = 0.78 to 0.86) and H2FPEF (r = 0.96 to 0.98), exhibiting similar correlations with PCWP.
  • Diagnostic accuracy did not consistently differ between automated and manual HFA-PEFF (AUC 0.70 [95% CI: 0.66–0.74] vs. 0.71 [95% CI: 0.66–0.75]) or H2FPEF scores (AUC 0.78 [95% CI: 0.73–0.82] vs. 0.75 [95% CI: 0.71–0.80]).
  • Automated scoring categorized 20% of manual high-likelihood HFpEF cases as intermediate likelihood due to smaller estimated left atrial volumes.

Study Design

Type

Observational (n=658)

Structured PICO

Does automated deep learning interpretation of echocardiograms perform similarly to manual measurements in diagnosing HFpEF?

P
Population
658 patients across three cohorts (test, ambulatory validation, and diagnostic validation) evaluated for HFpEF using automated and manual echocardiographic analysis.
E
Exposure
Automated deep learning interpretation of echocardiograms to calculate HFA-PEFF and H2FPEF scores
C
Comparator
Manual echocardiographic measurements to calculate HFA-PEFF and H2FPEF scores
O
Outcome
Diagnostic accuracy using the area-under-the-curve (AUC) and correlations between automated and manual scoressurrogate

Main Result

Effect estimate: AUC 0.70 (HFA-PEFF) and 0.78 (H2FPEF) (95% CI 0.66-0.74)

Automated deep learning analysis of echocardiograms provides similar diagnostic accuracy to manual measurements for calculating HFA-PEFF and H2FPEF scores in patients with suspected HFpEF.

Cite This Study

Venema et al. (2026) conducted an observational in Heart Failure with Preserved Ejection Fraction (HFpEF) (n=658). Automated deep learning interpretation of echocardiograms vs. Manual echocardiographic measurements was evaluated on Diagnostic accuracy (area-under-the-curve) for HFA-PEFF and H2FPEF scores (AUC 0.70 (HFA-PEFF) and 0.78 (H2FPEF), 95% CI 0.66-0.74). Automated deep learning interpretation of echocardiograms yielded similar diagnostic accuracy to manual analysis for HFpEF (AUC for HFA-PEFF 0.70 vs 0.71; AUC for H2FPEF 0.78 vs 0.75).

synapsesocial.com/papers/6a9d1ec828139818eab21dfbhttps://doi.org/10.1093/ehjdh/ztag142
View Full Paper
Ask AI
Bookmark
Share