Automated machine learning-based algorithm for ECG monitoring during exercise

Electrocardiogram (ECG) is a commonly used method for monitoring the body. The analysis of alterations in electrocardiographic activity during physical exercise can evaluate an individual's cardiovascular health status. Designing an automatic adaptive ECG signal classification model that can accommodate different physiological characteristics and changes in ECG activity patterns in different states is of great practical significance. In this study, we proposed an innovative approach that transformed ECG signals into image data and automatically constructed optimal neural network models for different distributions of ECG detection data using the differentiable neural architecture search algorithm (PC-DARTS). This approach achieved accurate ECG signal classification in various scenarios. In this paper, ECG signal classification experiments were conducted on the PhysioNet MIT-BIH arrhythmia dataset and the PTB diagnostic ECG dataset, respectively, to verify the discrimination performance of the proposed method in different application scenarios. The results indicate that the algorithm proposed in this paper has higher accuracy and greater robustness compared to other methods. At the same time, the proposed method is able to cope with the classification challenges posed by different acquisition devices, different experimental environments and different subject populations, and has sufficient generalization performance. In the future, the research results are expected to be combined with new ECG monitoring devices to achieve efficient and accurate ECG detection and accelerate the application of ECG detection in more fields.