Inter-patient heartbeat classification based on region feature extraction and ensemble classifier

The electrocardiogram (ECG) is an important tool for detecting arrhythmia. To solve the limitations of visual inspection, computer-aided diagnosis appears and grows rapidly. Most of the reported researches for heartbeat classification were based on intra-patient dataset. Moreover, existing inter-patient researches were usually conducted for superclasses of arrhythmia. To classify specific types of arrhythmia, this study proposed an inter-patient heartbeat classification method based on region feature extraction and ensemble classifier. The proposed method is composed of four stages. In preprocessing stage, the ECG signal is filtered and proportionally segmented. Afterwards, heartbeats are divided into three regions and region features are extracted. Subsequently, the dimension of features is reduced and all the features are fused and normalized. Eventually, an ensemble classifier is employed for the classification of Normal (N), Left Bundle Branch Block (LBBB), Right Bundle Branch Block (RBBB), Atrial Premature Contraction (APC) and Ventricular Premature Contraction (VPC). The method was applied to a new dataset divided from MIT-BIH arrhythmia database. The obtained sensitivities for Normal, LBBB, RBBB, APV and VPC were 95.0%, 27.9%, 79.6%, 81.8% and 88.1%. A comparative experiment demonstrated that the proposed region feature extraction method improves the accuracy of arrhythmia classification. The new division of MIT-BIH arrhythmia database is also advised to other researchers. (C) 2019 Elsevier Ltd. All rights reserved.