Classification of congestive heart failure and arrhythmias based on fusion of multiple electrocardiographic features

Electrocardiogram (ECG) is a common non-invasive method for recording the electrical activity of the heart. By analyzing ECG results, doctors can promptly detect and monitor heart issues, thereby effectively guiding the formulation of treatment plans. This paper proposes a multi-feature fusion-based classification method for congestive heart failure and arrhythmia. The method comprehensively extracts features from ECG signals, starting with the extraction of Heart Rate Variability (HRV) features, including time-domain and frequency-domain features. Simultaneously, the Constant Q Transform (CQT) algorithm is used to convert one-dimensional ECG signals into frequency-domain images, and the Vision Transformer model is employed for deep learning feature extraction. The Recursive Feature Elimination (RFE) algorithm is then used to screen features for relevance, and finally, the multi-feature fusion of HRV and deep learning features is input into an XGBoost classifier to complete the classification of congestive heart failure and arrhythmia. The experimental results show a classification accuracy of 99.65%, outperforming existing methods and validating the effectiveness of this approach in improving ECG diagnostic accuracy.