Research And Implementation Of Ecg Diagnosis System Based On Deep Learning

Deep learning has been fully developed in the field of image process and speech recognition in recent years.With the aid of this technology,the detection and classification of electrocardiogram(ECG)is a hot topic in the field of electrocardiogram health.Among them,the ECG classification method based on Convolutional Neural Network(CNN)has reached a good standard in accuracy[1].Although the study on ECG classification has achieved initial results,there are still many complex ECG pathological diseases that are not ideal for clinical application.Based on a large number of ECG clinical data,this project selected data samples including 21 common ECG diseases for classification training.In addition,this paper is devoted to the application of deep neural network to solve the problem of ECG R wave detection,ECG interference repair,and build a real-time diagnosis platform for ECG.The main research work of this paper can be divided into the following parts:Firstly,an ECG R-wave detection algorithmic model based on Recurrent Neural Network(RNN)is designed in this paper.RNN has a broad range of applications in time sequence,including speech recognition,machine translation,trigger word detection and etc.In this paper,the neural network uses the Long Short-Term Memory(LSTM)unit,and also uses a convolution layer before the LSTM unit,which fully consider the information around the R point,to improve the accuracy.By experimenting in test dataset,the LSTM,which focus on both global and local features of the sequence compared with the traditional R-wave detection based on wavelet analysis[2],has more stability and reliability for feature extraction.Secondly,an ECG missing repair algorithmic model based on Generative Adversarial Network(GAN)is designed.In recent years,the algorithm has developed rapidly in the field of image generation and image transformation,which is benefit by the development and application of GPU and the computational efficiency of deep learning greatly improved.Due to poor contact between the patient and the ECG device and other electrical signal interference in the actual measure process,the ECG signal is incomplete and the R point is missing,a large error in the calculation of heart rate(HR)and heart rate variability(HRV).The algorithm is mainly based on the GAN network to repair the missing ECG fragments,so that it can correctly identify the R point and calculate the HR and HRV more accurately.The algorithm mainly repairs the missing ECG fragments based on the GAN,so that it can correctly identify the R points and give a more accuracy HR and HRV results.Thirdly,A Restful style Application programming Interface is constructed,which interfaces with medical grade electrocardiograph,smart bracelet,portable ECG patch,and etc.Combined with ECG R-wave detection,ECG filtering,ECG diagnosis and other algorithm modules,it provides real-time ECG diagnosis service.This system uses the Django Rest Framework,MySQL database,and is deployed on the Nginx server via reverse proxy.In the meantime,the system data can be queried,managed,displayed and counted by the front-end system based on Vue framework.