pedQTNet: A Deep Learning Approach to Estimate Corrected QT Intervals from Multi-Lead Conventional ECG Waveforms in...

Area of Science:

• Cardiology
• Artificial Intelligence
• Pediatric Health

Background:

• Long QT syndrome (LQTS) increases the risk of ventricular arrhythmias and sudden cardiac death in children.
• Accurate corrected QT intervals (QTc) measurement is crucial but challenging for non-specialists in pediatric electrocardiograms (ECGs).

Purpose of the Study:

• To develop and evaluate pedQTNet, a deep neural network model for estimating QTc and detecting LQTS in pediatric patients.
• To assess pedQTNet's performance against expert interpretations and a commercial algorithm.

Main Methods:

• Trained pedQTNet on 65,370 pediatric ECGs (ages 0-18) annotated by pediatric electrophysiologists (PEPs).
• Evaluated QTc estimation accuracy (mean absolute error) and LQTS detection (sensitivity, specificity, likelihood ratios).
• Compared pedQTNet performance to GE Healthcare's Marquette 12SL algorithm and PEPs using cross-validation and a prospective set.

Main Results:

• PedQTNet achieved a mean absolute error of 18.8 ms for QTc estimation.
• In cross-validation, pedQTNet demonstrated 85% sensitivity and 87% specificity for LQTS detection.
• In a prospective set, pedQTNet showed higher sensitivity (100%) than PEPs (71%) for LQTS detection.

Conclusions:

• PedQTNet provides accurate QTc estimation and reliable LQTS detection in pediatric patients.
• The model outperforms a commercial ECG algorithm and performs comparably to expert interpretation.
• PedQTNet offers a scalable, automated tool for pediatric ECG screening and LQTS risk assessment, enhancing cardiac care.