Diagnosis of Arrhythmia Based on Multi-scale Feature Fusion and Imbalanced Data

Evidence suggests that Electrocardiogram (ECG) analysis plays an important role in the diagnosis of arrhythmia and the prevention of cardiovascular diseases. Extracting disease-related signals from ECG to improve the diagnostic efficiency of arrhythmia is still a challenging problem at present. In this paper, we propose a network framework based on multi-scale feature fusion and imbalanced data to classify and diagnose arrhythmias. The original ECG signals are denoised by improved wavelet threshold and the data issue is optimized by adaptive oversampling algorithm ADASYN, which can effectively improve the learning ability of the model for different types of samples. At the same time, in order to improve the classification efficiency of the model, we construct a deep learning framework based on the fusion of artificial features, convolutional depth features and temporal features. In the MIT-BIH database, the best accuracy of this method is 99.47%, which is higher than previous advanced methods. The average accuracy of 99.40% was obtained by cross-validation for the generalization performance of the model. The results show that the proposed method has excellent performance in ECG feature extraction and diagnosis of arrhythmias.