Applications And Research Of Machine Learning Method In Early Identification Of Sudden Cardiac Death

Sudden cardiac death is a major public health problem which is a great threat to life.It is usually caused by a variety of fatal arrhythmias and these arrhythmias often lead to insufficient blood pumping,and finally lead to cardiac arrest.which often results in death without effective treatments.With the serious problem of population aging,the early identification and prevention of sudden cardiac death has become an urgent work.Highrisk patient mostly can not get effective treatment when sudden cardiac arrest occurs,which is leading to death.Therefore,the early identification of sudden cardiac death is particularly important.Early warning can lighten doctor's workload as well as provide sufficient treatment time for doctors and bring greater survival opportunities for patients.In this paper,three works were done with the help of the latest clinical research progress and machine learning method:1.Early identification of sudden cardiac death combining machine learning method and clinical risk indicators.Based on the latest clinical research,a series of clinic risk indicators extraction method of sudden cardiac death was designed based on the electrocardiogram(ECG)waveform detection.Then it combines with the classical machine learning classifier to identify sudden cardiac death.In addition,an independent and synthetically risk indicators of sudden cardiac death in clinical research are integrated,and this independent risk indicator is proposed to assist doctors in clinical diagnosis and risk assessment.2.Early identification of sudden cardiac death combining transfer learning and shallow ECG features.Deep learning method is introduced into the early identification of sudden cardiac death with the help of transfer learning method to solve the problem of the scarcity of effective data from high-risk patients.In addition,Shallow ECG signal features including heart rate features and ECG signals are used in this study which is different from the tedious feature extraction of ECG signal in related research,Finally,all the early identification methods of sudden cardiac death proposed in this paper are analyzed and compared.3.Application of lightweight convolution neural network in myocardial infarction detection.To explore one of the important causes of sudden cardiac death: myocardial infarction,a lightweight convolutional neural network model with satisfying performance is trained to realize the early identification of myocardial infarction,based on four lead ECG and single lead ECG respectively,according to the difference between clinical monitoring and single lead intelligent wearable device.In addition,in the study of single lead,A detailed comparative analysis is done for neural network layer number and data input length.All in all,machine learning is applied to the early identification of sudden cardiac death and early detection of myocardial infarction,moreover the results are satisfying.It is preliminarily confirmed that the machine learning method based on clinical research can effectively identify sudden cardiac death at early stage.At the same time,myocardial infarction as main cause of sudden cardiac death is studied.Limited by the time and amount of data in this study,this paper has made a preliminary discussion and analysis.Further verification of results and conclusions will need large clinical data for testing and statistical evaluation in the future.