pedQTNet, a deep neural network model trained on pediatric ECG waveforms, demonstrated high QTc estimation accuracy and reliable LQTS detection, outperforming a commercial algorithm and performing on par with expert pediatric electrophysiologists.
pedQTNet achieved a mean absolute error of 18.8 ms in estimating corrected QT intervals in 10-fold cross-validation.
pedQTNet predicted LQTS at a 470 ms threshold with 85% sensitivity and 87% specificity in 10-fold cross-validation.
In a prospective set of 200 ECGs, pedQTNet demonstrated higher sensitivity for LQTS detection than pediatric electrophysiologists.
pedQTNet outperformed the GE Healthcare Marquette 12SL commercial algorithm in QTc estimation and LQTS classification.
The study dataset was one of the largest pediatric ECG datasets used for deep learning QTc analysis, spanning a decade of clinical data.
The model was developed specifically for pediatric populations, addressing the challenge that accurate QTc measurement is particularly difficult for non-heart-rhythm specialists in children.
This research describes the development and testing of pedQTNet, an artificial intelligence system designed to measure the QT interval on electrocardiograms (ECGs) in children. The QT interval is a feature of heart electrical activity, and when it is abnormally long (a condition called Long QT Syndrome or LQTS), it can increase the risk of dangerous heart rhythms and sudden cardiac death. Measuring it accurately in children is difficult and typically requires specialist expertise. The researchers trained their AI model on over 65,000 ECGs from nearly 38,000 patients under 18 years old, using measurements made by expert pediatric heart specialists as the standard of comparison. The study found that pedQTNet was highly accurate at estimating QT intervals, with an average error of about 19 milliseconds, and was effective at identifying children with LQTS. Importantly, the AI outperformed a widely used commercial ECG software (GE Healthcare's Marquette 12SL) and matched or exceeded the performance of human expert specialists. In a separate prospective test of 200 ECGs, the AI correctly identified 100% of LQTS cases, compared to 71% identified by the specialist physicians, though this difference was based on a small number of positive cases. This research suggests that an AI tool like pedQTNet could help make reliable LQTS screening more accessible in clinical settings where pediatric heart specialists are not always available. By automating QT interval measurement from standard ECG recordings, the tool could support earlier identification of children at risk for serious heart rhythm problems, potentially improving outcomes at scale across pediatric healthcare settings.
Ruiz V, Asztalos I, Silva L, Shi L, Iyer V, Nash D, et al.. (2026). pedQTNet: A Deep Learning Approach to Estimate Corrected QT Intervals from Multi-Lead Conventional ECG Waveforms in Pediatric Patients.. Journal of medical systems. https://doi.org/10.1007/s10916-026-02386-1