Laboratório de Instrumentação, Engenharia Biomédica e Física da Radiação (LIBPhys-UNL), Departamento de Física, Faculdade de Ciências e Tecnologia da Universidade Nova de Lisboa, Monte da Caparica, 2892-516 Caparica, Portugal.
Laboratório de Instrumentação, Engenharia Biomédica e Física da Radiação (LIBPhys-UNL), Departamento de Física, Faculdade de Ciências e Tecnologia da Universidade Nova de Lisboa, Monte da Caparica, 2892-516 Caparica, Portugal.
Comput Biol Med. 2017 Aug 1;87:322-334. doi: 10.1016/j.compbiomed.2017.06.009. Epub 2017 Jun 15.
Biosignals are usually contaminated with artifacts from limb movements, muscular contraction or electrical interference. Many algorithms of the literature, such as threshold methods and adaptive filters, focus on detecting these noisy patterns. This study introduces a novel method for noise and artifact detection in electrocardiogram based on time series clustering. The algorithm starts with the extraction of features that best characterize the shape and behaviour of the signal over time and groups its samples in separated clusters by means of an agglomerative clustering approach. The method has been tested in numerous datasets to reveal that it is independent on specific records and globally, the algorithm was able to successfully detect noisy patterns and artifacts with a sensitivity of 88%, a specificity of 92% and an accuracy of 91%, demonstrating a good performance in pattern detection based on morphological clustering. This algorithm can be applied to the detection and sectioning of multiple types of noise for more accurate denoising and adapted for signal classification.
生物信号通常会受到肢体运动、肌肉收缩或电干扰产生的伪迹的污染。文献中的许多算法，如阈值方法和自适应滤波器，都专注于检测这些噪声模式。本研究提出了一种基于时间序列聚类的心电图噪声和伪迹检测新方法。该算法首先提取最能表征信号随时间变化的形状和行为的特征，并通过凝聚聚类方法将其样本分组到单独的聚类中。该方法已经在许多数据集上进行了测试，结果表明它不依赖于特定的记录，并且全局上，该算法能够以 88%的灵敏度、92%的特异性和 91%的准确率成功检测噪声模式和伪迹，证明了基于形态聚类的模式检测的良好性能。该算法可应用于多种类型噪声的检测和分段，以实现更精确的去噪，并适用于信号分类。
