ECG Data Asset Management And Application

Heart disease is a common clinical disease,and in recent years,with the changes in people’s lifestyles and eating patterns,the clinical incidence has increased.Arrhythmia is a relatively common anomaly,and arrhythmia is often a precursor to some of the more severe heart diseases.Therefore,timely detection of arrhythmia is of great importance for prevention of heart disease.Electrocardiogram(ECG)is a non-invasive and effective medical tool for observing heart rhythm and heart state.ECG can detect abnormal heart rate in time.Manually distinguishing arrhythmias is time-consuming and labor-intensive.If you can use modern deep learning techniques to classify arrhythmias,it can greatly reduce the burden of artificiality.However,the amount of data required for deep learning training is large.Currently,a single database on the Internet is difficult to meet the data volume requirements.The existing large number of paper ECGs are not convenient to use computer technology for processing,and the ECG data record specifications of different data sources may differ.Therefore,different data formats,unified standards,and unified stored ECG signals can provide strong support for ECG research.In the era of big data,referring to the idea of data asset management,this paper proposes a method of ECG data asset management,and applies the standardized ECG data to the arrhythmia model to detect the effect of data governance.In response to the problems raised previously,this article proposes the following solutions:(1)A fusion database is proposed.In order to solve the problem of poor intelligent analysis of deep learning models in clinical practice,this paper expands the network training and testing database.It was expanded based on the international standard database MIT-BIH,and ECG data of the local People’s Hospital were added.(2)Aiming at the problem of digitizing paper ECG,a method for digitizing paper ECG was proposed.Denoising and filtering baseline drift were performed on the digitized sample set.(3)According to the differences in the data of different ECG databases,a management strategy for establishing data assets such as consistent coordinates,label standardization,and normalization is proposed.(4)Establish an LSTM network for input data signals.Utilizing the characteristics of ECG signals that are periodic in time,the signal data of each cycle in the ECG data set is fed to the network as a sequence,with the diagnostic code as the output.Classification results.This paper finally solves the problem of normalization of data between different databases,the digitization of paper ECG,and converts a large number of traditional paper ECGs into electronic data to increase the training set of deep learning models.Finally,an LSTM model is designed to perform experiments on different anomaly classifications.Finally,test results under different classification standards are obtained,which proves the effectiveness of data governance.