Arrhythmia classification from single-lead ECG signals using the inter-patient paradigm
Felipe Meneguitti Dias,Henrique L M Monteiro,Thales Wulfert Cabral,Rayen Naji,Michael Kuehni,Eduardo José da S Luz
DOI: https://doi.org/10.1016/j.cmpb.2021.105948
Abstract:Background and objectives: Arrhythmia is a heart disease characterized by the change in the regularity of the heartbeat. Since this disorder can occur sporadically, Holter devices are used for continuous long-term monitoring of the subject's electrocardiogram (ECG). In this process, a large volume of data is generated. Consequently, the use of an automated system for detecting arrhythmias is highly desirable. In this work, an automated system for classifying arrhythmias using single-lead ECG signals is proposed. Methods: The proposed system uses a combination of three groups of features: RR intervals, signal morphology, and higher-order statistics. To validate the method, the MIT-BIH database was employed using the inter-patient paradigm. Besides, the robustness of the system against segmentation errors was tested by adding jitter to the R-wave positions given by the MIT-BIH database. Additionally, each group of features had its robustness against segmentation error tested as well. Results: The experimental results of the proposed classification system with jitter show that the sensitivities for the classes N, S, and V are 93.7, 89.7, and 87.9, respectively. Also, the corresponding positive predictive values are 99.2, 36.8, and 93.9, respectively. Conclusions: The proposed method was able to outperform several state-of-the-art methods, even though the R-wave position was synthetically corrupted by added jitter. The obtained results show that our approach can be employed in real scenarios where segmentation errors and the inter-patient paradigm are present.