Automatic Identification Of Premature Ventricular Contraction Based On Fusion ECG Features

Premature Ventricular Contraction(PVC)is one of the most common arrhythmia diseases.The diagnosis of PVC in clinics is mainly based on inspection of abnormal electrocardiogram(ECG)patterns,including the broader and malformed QRS complex,P wave disappearance,T wave with opposite direction and compensatory pauses.However,the long-term ECG is often required due to the sudden of PVC onset,which makes the traditional visual examination of PVC EEG by a trained cardiologist is a time-consuming and subjective process.In order to overcome the limitations of traditional diagnostic methods,there has been a hot research topic of the automatic PVC identification using ECGs in recent years.In this paper,a novel automatic identification method of PVC is proposed,which is based on a new fusion ECG feature.Firstly,a new automatic peak points detection method is designed for the QRS complex in the ECG.Secondly,according to the detected peak points,the 9 different PVC-ECG features are extracted,which describe the characteristics of PVC-ECG patterns from different points of view.Then the feature selection method is employed to select key features from the proposed 9 features and feature fusion method is used construct the fusion feature.Finally,combining with BP neural network,an automatic identification method based on the fusion feature is proposed to differentiate PVC-ECGs and normal ECGs.This paper uses some ECG data records in the CPSC2018 database,and using numerical experiments to verify the feasibility and effectiveness of the proposed method.Numerical experimental results show that the accuracy,false detection rate,and missed detection rate of the method proposed in this paper in PVC automatic recognition are 97.46%,1.37%,and 3.41%,respectively.