AI-Based QRS Onset Detection in the Early Ventricular Activation Site ECGs.

Identifying the onset of the QRS complex is an important step for localizing the site of origin (SOO) of premature ventricular complexes (PVCs) and the exit site of Ventricular Tachycardia (VT). However, identifying the QRS onset is challenging due to signal noise, baseline wander, motion artifact, and muscle artifact. Furthermore, in VT, QRS onset detection is especially difficult due to the overlap with repolarization from the prior beat. In this study, 7706 captured bipolar pacing beats (Stim-QRS < 40 ms) pooled from 384 anatomically widely dispersed pacing sites of 15 patients were used for an attention-based Swin-Unet neural network. We also utilized a self-supervised pretraining technique using 88253 unannotated ECG records. The algorithm correctly identified most of the onsets for ECG signals with bipolar pacing-site ECG dataset, achieving a sensitivity of 0.958 and a 1.924 ± 4.275 milliseconds prediction error. Our algorithm also achieved a prediction error of 1.518 ± 8.702 milliseconds for the QT Database (QTDB), and a prediction error of 1.333 ± 7.575 milliseconds for the Lobachevsky University Electrocardiography Database (LUDB) public datasets. We also achieved high inter-dataset performance, which supports the practical performance of the method, with a sensitivity of 0.927 for QTDB and a sensitivity of 0.981 for LUDB. The AI model achieves accurate onset detection in paced ECGs with spike-removed inputs, providing a controlled, high-fidelity training setting for future efforts in generalizing to VT ECGs. The use of self-supervised pretraining further improves the detector's accuracy, showcasing the applicability of the approach and using unannotated ECG signals for downstream tasks.