A defibrillation heart rhythm recognition device

Abstract

The embodiment of the present invention discloses a defibrillable cardiac rhythm recognition device based on sequential pooling convolutional neural network and recurrent neural network, which performs sliding window processing on the original ECG signal; realizes the electrocardiographic data in each sliding window through the convolution layer. Band-pass filtering and waveform slope calculation; extract the distribution characteristics of the slope of the ECG data in each sliding window; input the slope distribution characteristics into the recurrent neural network, and use the dynamic time series relationship of the slope distribution characteristics to identify defibrillable heart rhythms. The heart rhythm recognition model aims to minimize the recognition error, and jointly optimizes the preprocessing, feature extraction and classification of the ECG signal. While making full use of the dynamic information of the slope distribution of the ECG signal, it realizes the conversion from the original ECG signal to the rhythm category. Marked end-to-end learning can automatically adjust the parameters of band-pass filtering and slope calculation, thereby providing more accurate criteria for out-of-hospital cardiac defibrillation first aid, while reducing the need for ECG detection time and improving first aid efficiency.

Description

technical field [0001] The invention relates to the technical field of data identification, in particular to a defibrillable cardiac rhythm identification device based on sequential pooling convolutional neural network and recurrent neural network. Background technique [0002] According to the "China Cardiovascular Disease Report 2016" released by the National Center for Cardiovascular Diseases, the number of sudden cardiac deaths in my country exceeds 540,000 each year, which is equivalent to about 1,500 people dying of sudden cardiac death every day. Sudden cardiac death is mainly caused by ventricular fibrillation (VF) and ventricular tachycardia (VT). The onset of VF / VT often has no warning. During the onset, the electrical activity of the ventricles loses synchronization and the pumping function of the heart loses. If measures are not taken in time to convert the heart rhythm, the patient will die within a few minutes. [0003] Deploying an automatic external defibril...

AI Technical Summary

Problems solved by technology

Although the inventive scheme uses the order statistics of the waveform slope to better characterize the morphological difference of ShR / NshR, it needs to set the difference interval for slope calculation empirically, and the classification method adopted cannot take advantage of the dynamics of the ShR / NshR waveform slope distribution change information

Images