Electrocardiogram data classification method and system combining feature extraction and inception network

Abstract

The invention discloses an electrocardiogram data classification method and system combining feature extraction and an inception network. The method comprises the steps that 1, normalizing filtered electrocardiogram data; step 2, segmenting the processed electrocardiogram data obtained in the step 1; 3, performing feature extraction on the plurality of heartbeat segments obtained in the step 2; step 4, constructing an inception network suitable for the electrocardiogram data; 5, obtaining relative wavelet energy and wavelet energy on the whole electrocardiogram data and absolute average deviation of wavelet energy and heartbeat interval time of all heartbeat segments by calculation; and step 6, taking the integrated features as input of a pre-constructed multi-layer perceptron, training the multi-layer perceptron to a preset convergence condition, and obtaining a final classification model. According to the method, the model not only extracts very valuable expert features, but also considers deep features of the electrocardiogram data, so that the purpose of enhancing the robustness and classification capability of the model can be achieved.

Description

technical field [0001] The invention belongs to the technical field of electrocardiographic data classification, and in particular relates to a method and system for classifying electrocardiographic data combined with feature extraction and an inception network. Background technique [0002] In medical practice, ECG (Electrocardiogram) is a relatively cheap and reliable tool. Through the analysis of a large amount of ECG data, it can help doctors discover many heart diseases, such as atrial fibrillation, myocardial infarction, and acute hypotension. Due to the rapid development of new sensing technologies, the ECG data to be analyzed is not only large in number but also complex in structure; in addition, the analysis of long-term ECG records is very cumbersome and time-consuming. Therefore, machine learning methods are used to classify ECG data. And diagnosis becomes a trend. [0003] At present, there are many methods based on feature selection to extract features from ECG...

AI Technical Summary

Problems solved by technology

There are some problems with the method described in this document: first, the method divides the ECG data into individual heartbeat beats, which will cause a certain degree of information loss, and the extracted features are not comprehensive enough; second, the proposed deep neural network is used in the model The final classification task is simply used as a classifier. In fact, in addition to some expert features, the ECG data also contains deeper features, and the deep neural network can be designed as an extractor of deep features.

Images