LMAU-Net: A Local Mask Attention Convolutional Neural Network for QRS Wave Detection

Abstract

The detection of QRS plays a crucial role in the analysis of ECG signals and the diagnosis of cardiac diseases. Although researchers have proposed many QRS detection algorithms with good performance, the noise robustness and generalization of these methods are still insufficient. To address these issues, we propose a new QRS wave detection model named LMAU-Net based on U-Net and local mask attention. The model is to introduce a novel local mask attention mechanism in the U-Net framework, which can effectively alleviate the influence of noise and non-QRS waves on detection. The proposed algorithm was tested on five commonly used ECG datasets and achieved 99.7% QRS detection precision. The proposed algorithm produces an error rate of no more than 0.1% on lownoise datasets. Furthermore, on two heavily noisy datasets, the sensitivity reaches over 98.72%. At the same time, it only takes 0.63s to detect a 30-minute long ECG signal, which verifies the real-time performance of the model. The LMAU - Net is a simple and efficient QRS detection algorithm that can provide fast QRS detection for telemedicine and embedded systems.