Electrocardiogram classification method and device based on wavelet transform and DCNN (Deep Convolutional Neural Networks)

Abstract

The invention provides an electrocardiogram classification method and device based on wavelet transform and DCNN (deep convolutional neural networks), and belongs to the field of biological medicinesand modes. According to an automatic classification method and device of electrocardiogram signals based on the wavelet transform and the deep convolutional neural networks, by utilizing wavelet functions, the electrocardiogram signals are decomposed into sub-signals of different frequency scales, and the sub-signals are filtered in sectional modes and are subjected to wavelet reconstruction; by utilizing 24 layers of convolutional neural networks, characteristic extraction is carried out by adopting crossed-size convolution kernels, data overfitting is prevented by adopting dropout and BatchNormalization when characteristic information is transmitted; and finally, classification is carried out by adopting a softmax classifier. The method is already verified on an ECG data set provided by2017 PhysioNet / CinC Challenge, and the accuracy rate is 0.871, and an F1 score is 0.8652. A research proves that noises of the electrocardiogram signals can be eliminated more effectively through thewavelet transform, and extraction of multi-layer characteristics can be performed by utilizing the 24 layers of convolutional neural networks, and meanwhile, a receptive field can be improved by increasing the sizes of the convolution kernels, so that the classification performance of a model is improved.

Description

technical field [0001] The invention relates to the fields of biomedicine and pattern recognition, in particular to an electrocardiogram classification method and device based on wavelet transform and DCNN. Background technique [0002] Atrial fibrillation (AF) is the most common sustained tachyarrhythmia in clinical practice. The main hazard of atrial fibrillation is the increased risk of vascular embolism, which is one of the main causes of ischemic stroke. Atrial fibrillation is manifested on the electrocardiogram as the disappearance of sinus P waves in each lead, the shape and amplitude of QRS waves are basically the same as sinus rhythm, and the R-R interval is absolutely unbalanced. The ECG automatic analysis and classification system can provide great help to doctors in diagnosing heart diseases, and is of great significance to improving medical efficiency, reducing medical costs, and preventing and diagnosing heart diseases. [0003] In recent years, extensive res...

AI Technical Summary

Problems solved by technology

In view of the weak ECG signal and the insufficient level of extracted features leading to insufficient recognition and classification accuracy, improvements are mainly made in the following aspects:

Images