Electrocardiogram Classification Based on Spatial-Channel Networks and Optimization Algorithms

Cardiovascular diseases are one of the leading causes of death globally, and the electrocardiogram (ECG) serves as a primary diagnostic tool, making automated ECG classification of significant clinical importance. Although previous studies have extensively utilized various neural network models for feature extraction and classification of ECG signals, these methods often face challenges such as insufficient accuracy, high computational demands, and lengthy processing times. This study aims to design a new arrhythmia classification framework that achieves high accuracy and low cost in diagnosing arrhythmias. We introduced Spatial and Channel-wise Convolutional (SCC) attention into Temporal Convolutional Networks (TCN), resulting in the SCC-TCN model, which significantly enhances classification accuracy and reduces computational costs. Additionally, by incorporating a stochastic walking mechanism, we improved the Rime optimization algorithm(RIME), effectively enhancing its performance. We also utilized the improved RIME algorithm to optimize the hyperparameters of the Spatial and Channel-wise Temporal Convolutional Networks(SCC-TCN) model, further improving its classification performance. Testing on the MIT-BIH Arrhythmia Dataset (MITAD) and the MIT-BIH Supraven-tricular Arrhythmia Dataset (SVDB) demonstrated that our method achieves classification accuracies of 99.40 % and 97.82%, respectively, validating the practical value of this approach.