A Two-Dimensional Arrhythmia Diagnosis Method Aiming to the Application of Mobile Terminal Devices

Electrocardiogram is an important basis for the diagnosis of arrhythmia. Intelligent and efficient diagnosis of arrhythmia diseases requires research in computer-aided methods. In this paper, we propose a pair of arrhythmia diagnosis methods based on lightweight two-dimensional networks. One takes MobileNetV3-Small as the backbone, while the other takes MobileNetV3-Large (with transfer learning involved) as the backbone. According to the experiments, our method can reach the overall accuracy of 0.9935 at the highest, and the results obtained in precision, recall, and specificity are also satisfactory. At the same time, since the network we developed has a relatively small file size, it is suitable for the application in mobile terminal devices. This method provides direction for mobile applications of arrhythmia detection equipment.