Deep learning based ECG segmentation for delineation of diverse arrhythmias

TL;DR

The paper tackles the challenge of accurately delineating P, QRS, and T waves in ECGs under diverse arrhythmias, where waveform morphology and limited annotated data hinder reliability. It proposes a 1D U-Net–like segmentation model with post-processing and an arrhythmia-guided classification branch, trained on a large internal diverse dataset and validated on QTDB and LUDB, achieving competitive benchmark performance and improved robustness to arrhythmia types. Key contributions include the diverse training regimen, a two-stage delineation pipeline, and a classification-guided strategy that substantially reduces false P-wave detections in AFIB/AFL, with systematic analysis of arrhythmia impact on delineation. The approach demonstrates strong potential for real-time analysis of long Holter recordings and highlights the importance of diverse arrhythmia representation for generalizable ECG delineation.

Abstract

Accurate delineation of key waveforms in an ECG is a critical step in extracting relevant features to support the diagnosis and treatment of heart conditions. Although deep learning based methods using segmentation models to locate P, QRS, and T waves have shown promising results, their ability to handle arrhythmias has not been studied in any detail. In this paper we investigate the effect of arrhythmias on delineation quality and develop strategies to improve performance in such cases. We introduce a U-Net-like segmentation model for ECG delineation with a particular focus on diverse arrhythmias. This is followed by a post-processing algorithm which removes noise and automatically determines the boundaries of P, QRS, and T waves. Our model has been trained on a diverse dataset and evaluated against the LUDB and QTDB datasets to show strong performance, with F1-scores exceeding 99% for QRS and T waves, and over 97% for P waves in the LUDB dataset. Furthermore, we assess various models across a wide array of arrhythmias and observe that models with a strong performance on standard benchmarks may still perform poorly on arrhythmias that are underrepresented in these benchmarks, such as tachycardias. We propose solutions to address this discrepancy.

Figures (11)

• Figure 1: A schematic representation of an ECG signal measured in lead I or lead II with the main complexes indicated.
• Figure 2: Flow diagram for ECG delineation: an ECG input signal is segmented by a U-Net like model using an optional classification branch, and post-processed for noise, before producing final delineation results.
• Figure 3: Segmentation model architecture. Our architecture is similar to U-Net3+, but uses 1D convolutional blocks and has an additional classifier branch.
• Figure 4: Arrhythmia classification branch network architecture.
• Figure 5: Examples of transformations used for data augmentation. (a) Original, (b) baseline wander, (c) baseline shift, (d) resize, (e) powerline noise and (f) Gaussian noise.