Imbalanced Heartbeat Classification Using Easyensemble Technique And Global Heartbeat Information

Heartbeat classification, also known as arrhythmia detection, is essential for early prevention of cardiovascular diseases (CVD). However, in clinical, the physician checks the ECG signal beat-by-beat for diagnosis, which is time-consuming and laborious. Recently, some scholars have proposed computer-aided heartbeat classification methods, while these methods mainly focus on using the local information of heartbeat and ignore the role of the global information where the heartbeat is located. Meanwhile, the heartbeat is highly imbalanced, resulting in poor performance in the minority categories of existing methods. Based on these issues, in this study we propose a new method for classifying imbalanced heartbeat using EasyEnsemble technique with global heartbeat information. By testing on the MIT-BIH arrhythmia database using the inter-patient scheme, the experimental results show that the global heartbeat information is useful for heartbeat classification. Meanwhile, compared with the existing methods, our method can not only achieve the best overall performance, but also can significantly improve the performance of minority categories while maintaining the good performance of majority categories.