Electrocardio diagnosis model and electrocardio detection device

Abstract

The invention relates to an electrocardio diagnosis model, and the model is characterized by comprising the following steps: S1, collecting N pieces of resting 12-lead electrocardiogram data, and enabling the number of atrial fibrillation electrocardiogram data to be equal to that of non-atrial fibrillation electrocardiogram data; S2, performing preprocessing: if the signal sampling frequency is lower than 200Hz, re-sampling is carried out firstly, so that the sampling frequency reaches 200Hz or above, and then filtering is carried out; S3, training a deep learning network: the deep learning network comprises at least three convolutional layers as a feature extraction module, and at least comprises two full connection layers as a classification module; S4, optimizing parameters of the deep learning network through the loss value of the loss function under minimization, and obtaining all weights and offsets in the whole deep learning network. The invention provides a loss function, which not only reduces the error between the input and the tag, but also restrains the error between the outputs of the feature layers of different types of signals.

Description

technical field [0001] The application belongs to the technical field of electrocardiographic signal processing, and in particular relates to a signal optimization method of an electrocardiographic simulation model and an electrocardiographic detection device. Background technique [0002] When the deep learning model is used for auxiliary screening of atrial fibrillation, there are only two possible results, atrial fibrillation or non-atrial fibrillation. The reason for obtaining atrial fibrillation and non-atrial fibrillation is that the new ECG data is closer to the characteristics of the atrial fibrillation data in the training data, or is closer to the non-atrial fibrillation data in the training data. However, it is difficult to obtain all types of AF data, especially all types of non-AF data. Therefore, some of the learned network features may not be the key to distinguish between atrial fibrillation and non-atrial fibrillation. Therefore, for non-atrial fibrillatio...

AI Technical Summary

Problems solved by technology

Therefore, for non-atrial fibrillation ECG signals that have not participated in training, the trained deep learning model has a greater chance of making mistakes

In addition, because the ECG signal is nonlinear and varies from person to person, it is possible that the new ECG signal may be quite different from all signal features in the training samples; traditional deep learning networks often only reduce the total network The error between the output of the label and the label, at most introduces restrictions on the weights and coefficients

At this time, the probability of the new ECG signal being classified as atrial fibrillation and non-atrial fibrillation is close, and some interference may lead to misjudgment

Images