Noise Detection in Electrocardiography Signal for Robust Heart Rate Variability Analysis: A Deep Learning Approach.

Heart rate variability (HRV) analysis is widely used to assess the sympathetic and parasympathetic tones. However, the quality of the derived HRV features is heavily dependent on the accuracy of QRS detection. Noisy electrocardiography (ECG) signals, such as those measured by wearable ECG patches, can lead to inaccuracies in the QRS detection and significantly impair the HRV analysis. Hence, it is critical to employ noise detection algorithms to identify the corrupted segments of the ECG signal and discard them from the analysis. This paper proposes a convolutional neural network to distinguish between usable and unusable ECG segments where usability is defined based on the accuracy of QRS detection. The results indicate that the proposed method has significantly lower error rates compared to both the baseline method (HRV analysis on the noisy signals) and a noise detection method based on four ECG signal quality indices and a support vector machines classifier.