Key result
Deep residual networks using random signal segmentation speed PVC localization ~2.9x with limited accuracy loss.
Why the study?
Existing PVC origin localization methods require time-consuming R-peak localization and wavelet transform processes, creating a need for more efficient localization given large ECG datasets.
Does random signal segmentation improve the efficiency of localizing the origins of PVCs compared to traditional heartbeat localization algorithms?
Does random signal segmentation improve the efficiency of localizing the origins of PVCs compared to traditional heartbeat localization algorithms?
Random signal segmentation combined with a deep residual network offers a faster alternative to traditional ECG processing for localizing PVC origins, achieving a 2.86x speedup with minimal accuracy loss.
May accelerate PVC origin localization in EP labs; leaves open prospective clinical validation before adoption.
Localization of origins of premature ventricular contraction (PVC) is significant in the treatment of ventricular arrhythmia. Existing localization methods usually adopt a heartbeat localization algorithm in which R-peak localization and wavelet transform processes are rather time-consuming. Considering the fact that ECG is usually long and the overall quantity of ECG is large, efficient localization of origins of PVC is desired. In this paper, we propose to use random signal segmentation to process ECG so that the R-peak localization and wavelet transform processes are discarded with improved efficiency. The study included the dataset from 843 patients with spontaneous PVCs, which were clustered and labeled corresponding to 4 regions in the entire ventricle. The dataset contains a total number of 76221 groups, each incorporating 250 sampling points in 12 leads. To obtain the optimal classification model, 75% (sample group = 57166) of the whole dataset is randomly selected for training, while the remaining 25% (sample group = 19055) are selected for testing. The deep residual network (ResNet) is used to estimate the performance of classification methods. Experimental results show that the proposed scheme achieves a speedup of 2.86x while with limited accuracy loss.
No takes yet. Share an insight, caveat, or question.
Li et al. (2021) studied Premature ventricular contraction (PVC) (n=843). Random signal segmentation with deep residual network (ResNet) vs. Existing heartbeat localization algorithms (R-peak localization and wavelet transform) was evaluated on Speedup and accuracy of classification. Random signal segmentation with a deep residual network achieved a 2.86x speedup in localizing the origins of premature ventricular contractions with limited accuracy loss.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: