A Novel Automatic Detection Method Of Atrial Fibrillation By Extracting ECG Fusion Feature

Atrial Fibrillation(AF)is one of the most common arrhythmias in the clinic.Traditional atrial fibrillation detection often requires professional physicians to visually inspect the long-term continuous ECG.This process is not only time-consuming,but also extremely relied on the physician's experience.In addition,it is difficult to timely monitor and diagnosis of atrial fibrillation ECGs due to the limited number of professional physicians.Therefore,it is significant to study of the automatic AF detection method,where the effective AF feature extraction using ECG is a key step to achieve good detection results.It is well known that there are two characteristics in AF ECGs,P wave absence and absolute irregularity of RR intervals.Based on such observation,this paper proposes a new fusion feature extraction method using AF ECGs.The structure of this paper is shown as follows.The first chapter systematically summarize the background and significance of atrial fibrillation automatic AF detection,and explains the importance of feature extraction using ECGs.A brief introduction to the basic knowledge of ECG,the characteristics of AF ECG and the existing methods of feature extraction method using AF ECGs are given.The second chapter introduces the principle of wavelet transform and adjustable quality factor wavelet transform in systematically.Two new measurements,that is,confidence divergence distance and confidence divergence index are defined respectively on the basis of T-lag scatter plots.Then the novel P-wave-absence-based feature extraction method is proposed.In the third chapter,a new R peak detection method is first proposed through applying dynamic threshold method.Then,an RR-interval-irregularity-based feature extraction method is proposed based on the substring length probability distribution entropy and variation coefficient.Finally,a new fusion feature extraction method is proposed,where the P-wave-absence-based feature and RR-interval-irregularity-based feature are combined.The fourth chapter applies the proposed fusion feature extraction method and extreme learning machine to conduct the automatic AF detection,so as to verify the feasibility and effectiveness of the proposed fusion feature extraction method.