Electrocardiogram scatter diagram anomaly identification and classification method based on multi-modal neural network

Abstract

The invention discloses an ECG scatter diagram anomaly recognition and classification method based on a multi-mode neural network. The method comprises the steps that 1, an ECG data set with an artificial mark is obtained; 2, the electrocardio data is drawn into an electrocardio scatter diagram with time sequence characteristics, and labels are set for electrocardio scatter diagram samples; 3, establishing an ECG scatter diagram anomaly recognition and classification model based on the multi-modal neural network; 4, training a multi-modal neural network by using the ECG scatter diagram sample and the label thereof; 5, the R peak position of the electrocardiogram data to be analyzed is positioned and recorded, and an electrocardiogram scatter diagram is drawn according to the positioned position; and 6, carrying out anomaly classification on the ECG scatter diagrams by utilizing the ECG scatter diagram anomaly identification and classification model. According to the method, one-dimensional electrocardio data can be converted into a two-dimensional electrocardio scatter diagram, and the abnormal electrocardio data can be recognized and classified in combination with the multi-mode neural network model.

Description

technical field [0001] The invention belongs to the technical field of intelligent auxiliary medical diagnosis, in particular to a method for identifying and classifying abnormalities in an electrocardiogram scattergram based on a multimodal neural network. Background technique [0002] The traditional dynamic electrocardiogram is to record the long-term electrocardiographic activity of the human body through an electrocardiograph, and analyze and process it through a computer, which provides an important basis for clinical preliminary screening. However, the intuitive electrocardiogram analysis method has defects such as huge amount of data, difficult feature extraction, and low judgment and recognition rate. Since the dynamic electrocardiogram is often affected by the body's position, activity, mood, sleep and other factors during the monitoring process, the detection results need to be combined with relevant clinical data before comprehensive analysis can be carried out t...

AI Technical Summary

Problems solved by technology

However, the current research on artificial intelligence-based ECG data is limited to artificial intelligence analysis of one-dimensional original ECG data, and one-dimensional ECG data often has shortcomings such as large data volume, redundant information, and scattered key information.

At the same time, the artificial intelligence judgment model based on one-dimensional ECG data usually only focuses on the local information of the ECG data, and it is easy to ignore the problems reflected by the ECG data from a macro perspective, resulting in judgment errors

Images