AF Recognition Algorithm Based on Fusion Transformer Model

Atrial fibrillation (AF) is a common arrhythmia. Recognizing the condition accurately is crucial for the early diagnosis and treatment of cardiovascular diseases. This study proposes a fusion transformer model that integrates convolutional neural network (CNN), convolutional block attention module (CBAM), and transformer models to improve the recognition performance of deep learning methods in detecting AF. The proposed approach focuses more comprehensively on features to improve the accuracy of AF prediction. To verify the effectiveness of the method, single-lead ECG data from patients with AF and a healthy control group are obtained from the MIT-BIH Atrial Fibrillation Database (AFDB). The experimental results show that the proposed method achieves an excellent accuracy rate of 99.3%.