Automatic arrhythmia analysis method based on neural network of compressed image

Abstract

The invention discloses an automatic arrhythmia analysis method based on neural network of a compressed image. The method comprises that a multi-channel electrocardiogram sample is generated in two sampling manners; and obtained 600-dimension electrocardiosignals are spliced along the second dimension, original electrocardiosignals in lead II are equivalent to a 4*600*1-dimension electrocardiosignal sample, input signals of four channels are input to a merging layer and merged, and 600*4-dimension signals are output. An image coding layer between a merging layer unit and a convolution layer unit is used to convert the electrocardiosignals from 1D code into 2D image. A three-layer convolution layer unit is connected after the image coding layer in series; the convolution layer unit comprises a convolution layer which extracts features of 2D electrocardiosignal coding images via 2D convolution and an excitation unit and a pooling layer connected in series successively; the excitation unit in serial connection serves as a full connection layer of softmax; output is carried out; parameters of the deep neural network are learned and samples are identified automatically; and the problemthat a present arrhythmia analysis system cannot satisfy the clinical requirement for accuracy is solved.

Description

technical field [0001] The invention relates to the technical field of medical signal processing, more precisely, the invention relates to an automatic arrhythmia analysis method based on a compressed graph neural network. Background technique [0002] In recent years, auxiliary diagnostic equipment for electrocardiograms has developed rapidly. With the advancement of science and technology in the information field, especially with the development of pattern recognition technology, the function of electrocardiogram equipment is no longer just to obtain electrocardiogram signals and print electrocardiograms, but to Mining the effective data in the electrocardiogram and developing in the direction of automatic identification and statistical heart beat information. The analysis equipment with the function of automatic heartbeat recognition can provide doctors with more intuitive and effective ECG information, effectively save diagnosis time and improve the diagnosis efficiency ...

AI Technical Summary

Problems solved by technology

[0005] The purpose of the present invention is to solve the problem that the existing arrhythmia analysis system is not enough to meet the accuracy requirements of clinical applications, and to provide an automatic arrhythmia analysis method based on the compressed graph neural network

Images