Key result
A deep learning model detected left ventricular dilatation from 12-lead ECGs with an AUROC of 0.810, significantly outperforming logistic regression and random forest methods.
Why the study?
Left ventricular dilatation and hypertrophy are risk factors for heart failure whose detection improves screening, prompting investigation into whether deep learning can detect them from a 12-lead ECG.
Does a deep learning model improve the detection of left ventricular dilatation and hypertrophy from 12-lead ECGs compared to machine learning models and conventional criteria?
Population
18,954 patients with ECG and echocardiographic data
Comparison
Deep learning model vs machine learning models and conventional ECG criteria
Design
Diagnostic model development and validation study
Authors
Loading...
May aid ECG-based HF screening; leaves open prospective validation before clinical use.
Observational (n=18,954)
No
Does a deep learning model improve the detection of left ventricular dilatation and hypertrophy from 12-lead ECGs compared to machine learning models and conventional criteria?
Effect estimate: AUROC 0.810 (95% CI 0.801-0.819)
Absolute Event Rate: 0.81% vs 0.77%
p-value: p=<0.001
Deep learning applied to standard 12-lead ECGs can detect left ventricular dilatation and hypertrophy with significantly higher accuracy than conventional machine learning models and standard ECG criteria.
Kokubo et al. (2022) conducted an observational in Left ventricular dilatation and hypertrophy (n=18,954). Deep learning model vs. Logistic regression and random forest methods was evaluated on Detection of left ventricular dilatation (AUROC) (AUROC 0.810, 95% CI 0.801-0.819, p=<0.001). A deep learning model detected left ventricular dilatation from 12-lead ECGs with an AUROC of 0.810, significantly outperforming logistic regression and random forest methods.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: