Accurate delineation of key waveforms in an ECG is a critical initial step in extracting relevant features to support the diagnosis and treatment of heart conditions. Although deep learning based methods using a segmentation model to locate the P, QRS, and T waves have shown promising results, their ability to handle signals exhibiting arrhythmia remains unclear. This study builds on existing research by introducing a U-Net-like segmentation model for ECG delineation, with a particular focus on diverse arrhythmias. For this purpose, we curate an internal dataset containing waveform boundary annotations for various arrhythmia types to train and validate our model. Our key contributions include identifying segmentation model failures in different arrhythmia types, developing a robust model using a diverse training set, achieving comparable performance on benchmark datasets, and introducing a classification guided strategy to reduce false P wave predictions for specific arrhythmias. This study advances deep learning based ECG delineation in the context of arrhythmias and highlights its challenges.

翻译：准确提取心电图关键波形是支持心脏疾病诊断和治疗所需特征的关键初始步骤。尽管基于深度学习的分割模型在定位P波、QRS波和T波方面已展现出良好效果，但其处理心律失常信号的能力尚不明确。本研究在现有工作基础上，引入类U-Net分割模型进行心电波形界定，特别关注多种心律失常类型。为此，我们构建了包含多种心律失常波形边界标注的内部数据集用于模型训练与验证。主要贡献包括：识别不同心律失常类型中分割模型的失效模式、利用多样化训练集开发鲁棒模型、在基准数据集上取得可比性能、提出分类引导策略以降低特定心律失常的假阳性P波预测。本研究推动了基于深度学习的心律失常心电波形界定技术发展，并揭示了相关挑战。