Fast, Accurate, and Robust Segmentation of Long-Term Electrocardiograms Through Multi-Point Iterative Warping and Dynamic Template Generation

Accurate beat-wise ECG segmentation is critical for extracting clinically relevant biomarkers such as QT and TpeakTend intervals. However, traditional methods often struggle with precision in the presence of morphological variability and signal drift. We present an Enhanced Two-Dimensional Warping (E2DW) framework that extends the original 2DW algorithm through two key innovations: dynamic template adaptation, which employs drift detection to continuously update beat templates in response to evolving morphology, and multi-point grid alignment, which generalizes the warping process via a vectorized multi-point optimization strategy. Evaluated on the QT Database and a full-night clinical sleep study, E2DW outperformed wavelet-based methods and the baseline 2DW in segmentation accuracy, robustness, and computational efficiency. These results demonstrate reliable beat-by-beat segmentation of long ECG recordings, underscoring E2DW’s potential as a high-fidelity tool for both clinical and research applications.