Research On ECG Data Classification Algorithm Based On Neural Network

Heart disease has been one of the major diseases that threaten human life for many years,and the number of deaths due to heart disease worldwide continues to grow.Electrocardiogram is a visual reflection of cardiac activity.Extracting and analyzing ECG data from an electrocardiogram is of great significance for diagnosing and preventing heart disease.A large amount of ECG data is generated every day in the world,and due to the limited number of medical experts,a large amount of ECG data cannot be processed in time.Most of the current neural network algorithms for ECG classification only focus on improving accuracy and neglecting the computational cost.Deploy and run on the device.To this end,this paper studies the ECG data classification algorithm based on neural network,and builds an ECG storage and annotation platform.The main research contents are as follows:(1)For the current neural network algorithm parameters are too large,the classification time is long and so on.This paper proposes a lightweight neural network algorithm Lite Net(Liteweight Neural Network)based on neural network.The algorithm adopts a unique convolution structure and convolution method to make it fully extract the data features and improve the classification efficiency of the algorithm.In order to verify the classification performance of Lite Net,this paper uses two different ECG data sets to train and test the algorithm.The experimental results show that Lite Net has achieved good classification results in the dataset MIT-BIH and dataset CCDD,respectively.97.87% and 92.5% accuracy.At the same time,the design of the other three models compared with Lite Net fully demonstrates the structural advantages of Lite Net.(2)The storage format for massive ECG data is not standardized and the diagnostic efficiency of ECG is low.This paper constructs an ECG storage and labeling platform,which stores ECG data in the standard format of HL7-a ECG(Health Level 7 Annotation Electrocardiogram),which is conducive to the unification of storage format.At the same time,according to the characteristics of Mongo DB database,each ECG data sample is segmented for efficient storage.Finally,the trained Lite Net model is deployed to the platform,and the automatic annotation function is provided for the ECG expert to provide auxiliary diagnosis and improve work efficiency.In this paper,a new neural network algorithm is designed for the classification of ECG data.This algorithm reduces the computational cost and the computational time consumption while ensuring the accuracy of data classification.On this basis,the ECG storage and labeling platform is constructed to realize the rational storage of ECG data and provide manual and automatic labeling functions one by one,which provides an effective means for the diagnosis of ECG data.