Ye Can, Coimbra Miguel Tavares, Vijaya Kumar B K
Department of Electrical & Computer Engineering, Carnegie Mellon University, USA.
Annu Int Conf IEEE Eng Med Biol Soc. 2010;2010:1918-21. doi: 10.1109/IEMBS.2010.5627645.
Computer-assisted cardiac arrhythmia detection and classification can play a significant role in the management of cardiac disorders. In this paper, we propose a new approach for arrhythmia classification based on a combination of morphological and dynamic features. Wavelet Transform (WT) and Independent Component Analysis (ICA) are applied separately to each heartbeat to extract corresponding coefficients, which are categorized as 'morphological' features. In addition, RR interval information is also obtained characterizing the 'rhythm' around the corresponding heartbeat providing 'dynamic' features. These two different types of features are then concatenated and Support Vector Machine (SVM) is utilized for the classification of heartbeats into 15 classes. The procedure is applied to the data from two ECG leads independently and the two results are fused for the final decision. Compare the two classification results and the classification result is kept if the two are identical or the one with greater classification confidence is picked up if the two are inconsistent. The proposed method was tested over the entire MIT-BIH Arrhythmias Database [1] and it yields an overall accuracy of 99.66% on 85945 heartbeats, better than any other published results.
计算机辅助心律失常检测与分类在心脏疾病管理中可发挥重要作用。本文提出一种基于形态学和动态特征相结合的心律失常分类新方法。对每个心跳分别应用小波变换（WT）和独立成分分析（ICA）来提取相应系数，这些系数被归类为“形态学”特征。此外，还获取RR间期信息以表征相应心跳周围的“节律”，从而提供“动态”特征。然后将这两种不同类型的特征连接起来，并利用支持向量机（SVM）将心跳分类为15类。该过程分别应用于来自两条心电图导联的数据，并将两个结果融合以做出最终决策。比较两个分类结果，若二者相同则保留该分类结果，若二者不一致则选取分类置信度更高的结果。所提方法在整个麻省理工学院 - 贝斯以色列女执事医疗中心心律失常数据库[1]上进行了测试，在85945个心跳上总体准确率达到99.66%，优于任何其他已发表的结果。
