What does this research mean for the field?

Four externally validated AI-ECG models showed AUC values from 0.83 to 0.96 for detecting left ventricular systolic dysfunction across various cohorts. Novelty: ClaimNovelty.SYNTHESIS. Consensus alignment: ConsensusAlignment.NEUTRAL.

What question did this study set out to answer?

The study aims to review AI-ECG models for LVSD detection and assess their performance through external validation.

February 8, 2026European Heart Journal

Assessing the external performance of AI-ECG for left ventricular systolic dysfunction: a review and multinational head-to-head comparison

View Full Paper

Ask AI

Bookmark

Key result

Validated AI-ECG models detect LVSD with an AUC up to 0.96 in heart failure patients.

Why the study?

Incomplete reporting of dataset characteristics and limited model sharing hinder independent external validation, leaving concerns about generalizability and bias of AI-ECG models for detecting LVSD.

Do AI-ECG models accurately detect left ventricular systolic dysfunction when externally validated in an independent cohort?

Population

1,306 individuals with paired ECG and CMR data

Comparison

Four shared AI-ECG models evaluated head-to-head

Design

Systematic review and independent external validation study

Authors

PCP M CroonYale University MBM J BoonstraAmsterdam University Medical Centers CAC P AllaartAmsterdam University Medical Centers

Discussion

Member takes

Implication

Supports deployment for LVSD screening; confirms generalizability beyond internal validation.

Key Points

The study aims to review AI-ECG models for LVSD detection and assess their performance through external validation.
Conducted a systematic review to identify AI-ECG models for LVSD detection.
Searched Ovid MEDLINE for relevant studies from 2010 onwards.
Contacted corresponding authors to share models for external validation.
Evaluated model performance using a cohort with paired ECG and cardiac MRI data.
Assessed performance metrics including AUC, sensitivity, and predictive values.
Identified 35 studies describing 51 AI-ECG models.
57% of models reported AUC values greater than 0.9 for internal validation.
AUC values for external validation ranged from 0.83 to 0.93 in the full cohort and 0.87 to 0.96 in the heart failure cohort.
Model performance remained stable across different sex and age groups, with slight reductions in specific ECG conditions.

Structured PICO

Do AI-ECG models accurately detect left ventricular systolic dysfunction when externally validated in an independent cohort?

Population

1,306 individuals with paired ECG and cardiac MRI (CMR) data, mean age 59 ±15 years, 35% female.

Intervention

Four AI-ECG models for left ventricular systolic dysfunction (LVSD) detection (from Korea, Taiwan, United States, and the Netherlands).

Comparator

Cardiac MRI (CMR) as the reference standard.

Outcome

Diagnostic performance for LVSD detection measured by area under the receiver operating characteristic curve (AUC), sensitivity, specificity, negative predictive value (NPV), and positive predictive value (PPV).surrogate

AI-ECG models for detecting left ventricular systolic dysfunction demonstrate strong and consistent diagnostic performance in independent external validation.

Cite This Study

Croon et al. (2025) studied this question. Four externally validated AI-ECG models for LVSD detection showed AUCs of 0.83-0.93 in a full cohort and 0.87-0.96 in a heart failure cohort.

synapsesocial.com/papers/698828d90fc35cd7a8848ad3 https://doi.org/10.1093/eurheartj/ehaf784.4385

Topics

Heart failure HFrEF treatment Artificial intelligence in cardiology

View Full Paper

Ask AI

Bookmark