다양한 심박수 범위에 적용 가능한 AF 분류 심박수 적응형 CNN 모델

Abstract

The early diagnosis of atrial fibrillation (AF), a critical subtype of arrhythmia in middle-aged individuals, is essential for preventing severe cardiac events. In this study, we propose a heart rate adaptive convolutional neural network model (HRA-CNN) for AF classification. To enhance the generalizability of the model across various heart rate ranges, transfer learning and data augmentation strategies were employed. Transfer learning and data augmentation are commonly used techniques in deep learning to improve the generalizability of models. By leveraging pre-trained models and augmenting the training data, the model can better capture the underlying patterns and variations in the data, leading to improved performance on new, unseen data. The HRA-CNN model shows promise in improving AF detection accuracy across different heart rate ranges. The study results will be important in determining the effectiveness of this approach and its potential implications for early AF detection, ultimately leading to improved patient outcomes.