Position Detection Of QRS Wave Of ECG Based On Unet Network

The electrocardiogram can reflect the electrical activity of the heart,so the electrocardiogram can be used to diagnose various heart diseases.However,ECG signals will be interfered by various noises during the acquisition process.Therefore,ECG signal preprocessing and detection of characteristic waveforms have always been research hotspots in the field of medical signal processing.This research studies and improves the commonly used ECG signal denoising algorithm and QRS wave position detection algorithm,and proposes a new deep learning algorithm.The main work of this study is as follows:Firstly,according to the Pan＿Tompkins algorithm,the study proposes a new detection algorithm based on the dynamic threshold QRS wave position,improves the pre-processing in the Pan＿Tompkins algorithm,and the pre-processing part of the improved algorithm can simply and effectively remove the noise present in the ECG signal.The features of QRS waves are obvious,which is conducive to subsequent detection.The detection part follows the innovation of the Pan＿Tompkins algorithm,that is,the concept of dynamic update threshold and backtracking is also used.At the same time,the experimental parameters are reselected based on the experimental data: backtracking condition value and the average RR interval.The results show that the accuracy of the QRS wave position detection algorithm based on dynamic threshold has reached 71.85%,while the accuracy of the Pan＿Tompkins algorithm is only29.10%.Secondly,the research uses a two-stage strategy to detect the position of the QRS wave of the ECG signal.First,the sensitive area where the QRS wave is located is detected,and then the specific position of the QRS wave is located in the detected sensitive area.The algorithm used to detect the sensitive area where the QRS wave is located is studied,and the Unet segmentation network is proposed.At the same time,In order to enhance the features useful for QRS wave position detection in ECG signals,suppress the features that are not useful for QRS wave detection in ECG signals,the SE block is introduced in the Unet segmentation network,and a comparative experiment is performed based on the proposed Unet segmentation network and the SE＿Unet segmentation network.The results show that the accuracy rates of the Unet detection algorithm and the SE＿Unet detection algorithm are 81.45% and 82.35%,respectively.Finally,based on the structural characteristics of the recurrent neural network and the observation that the electrocardiogram can be regarded as a time series of waveforms,the recurrent residual convolutional neural network,namely the R2 Unet segmentation network,and the SE＿R2Unet segmentation network is also proposed.And the Unet detection algorithm,R2 Unet detection algorithm and its improved algorithm were analyzed and compared.The results show that the R2 Unet detection algorithm and the SE＿R2Unet detection algorithm have correct rates of 84.95% and 86.80%,respectively.