Method for obtaining risk predictions of sudden death with weight value matrices of artificial neural network

Abstract

The invention discloses a method for obtaining the weight value matrix of the artificial neural network for sudden death risk prediction. The sudden cardiac death database and the MIT-BIH normal sinus rhythm database are constructed into training data samples and cross-validation samples, and each layer of the artificial neural network is randomly set first. Input training data samples to iteratively correct the weight values ​​of each layer until the training error is less than a specified value, find the weight value matrix that can predict the risk of sudden death, and then use the weight value matrix to add the weight value matrix to the original artificial neural network to construct a new model. The artificial neural network, and then use the collected target human ECG signal as data, process the human ECG signal to obtain the target human feature vector X, and perform prediction calculations according to the target human feature vector X and the new artificial neural network, and finally Get the predicted value.

Description

technical field [0001] The invention relates to sudden death risk prediction, in particular to a method for obtaining a sudden death risk prediction artificial neural network weight value matrix. Background technique [0002] Sudden cardiac death (sudden cardiac death, SCD) refers to the natural death caused by cardiac causes, which is characterized by sudden loss of consciousness and occurs within 1 hour after the onset of acute symptoms. According to statistics, there are about 7 million SCD patients in the world every year, accounting for 1 / 4 of all deaths, which seriously threaten people's lives. At present, the average success rate of rescue in the world is less than 1%. [0003] Patients with sudden cardiac death are usually healthy (50% of cardiac arrests occur in individuals without known heart disease) or in stable condition, and there may be manifestations of heart disease before sudden death, but a considerable number of heart disease patients may have sudden deat...

AI Technical Summary

Problems solved by technology

[0006] Although SCD directly endangers the personal safety of patients and has very large harm and relatively serious consequences, the early identification technology of SCD in clinical medicine mainly lies in stratified long-term risk management and prediction, and short-term prediction before the occurrence of SCD Technology, which obviously lags behind modern treatment technology is in the process of exploration. The main difficulty and key of this short-term prediction of SCD is how to timely and accurately identify the population at high risk of sudden death, and take intervention measures to reduce the occurrence of sudden death

Images