Deep Learning-Based Continuous QT Monitoring to Identify High-Risk Prolongation Events After Class III Antiarrhythmic...

Area of Science:

• Cardiology
• Medical Technology
• Artificial Intelligence in Medicine

Background:

• Drug-induced QT prolongation is a risk during outpatient care for patients on Class III antiarrhythmics.
• Insertable cardiac monitors provide continuous data but are limited to single-lead configurations.
• Accurate QT/QTc measurement is crucial for managing antiarrhythmic drug safety.

Purpose of the Study:

• To develop and validate a deep learning system (3DRECON-QT) for reconstructing spatial information from single-lead ECGs.
• To quantify QT/QTc intervals and identify high-risk QT prolongation using this system.
• To assess the system's performance in continuous monitoring and real-world outpatient cohorts.

Main Methods:

• Developed 3DRECON-QT, a multitask encoder-decoder model, to ingest single-lead ECGs and predict 12-lead ECGs and QT/QTc.
• Trained and tested the model on large health system and external center datasets, including a public dofetilide-loading dataset.
• Validated performance in a real-world cohort of outpatients on dofetilide or sotalol and in patients with insertable cardiac monitor recordings.

Main Results:

• 3DRECON-QT achieved high accuracy in classifying prolonged QTc (AUC 0.942 internal, 0.943 external) with low mean absolute error.
• Continuous monitoring predictions correlated well with ground truth (r=0.851) and accurately identified QTc changes.
• In outpatients, 16.5% showed high-risk QTc prolongation, associated with a 4.24-fold increased risk of ventricular arrhythmias detected by 3DRECON-QT (AUC 0.94).

Conclusions:

• A single-lead deep learning approach can achieve guideline-level QT/QTc measurement accuracy.
• Enables continuous QTc surveillance from nonstandard ECG vectors, enhancing safety monitoring for Class III antiarrhythmics.
• Identifies clinically meaningful outpatient QTc prolongation linked to a significantly higher risk of serious ventricular arrhythmias.