Key result
A deep learning method using ResNet-18 and transfer learning achieved PVC recognition accuracies of 99.93% and 99.77% on the MIT-BIH and INCART datasets, respectively.
Why the study?
Manual assessment of long-term ECGs to identify premature ventricular contractions is time consuming and cumbersome for cardiologists.
Population
44,103 normal and 6423 PVC beats from MIT-BIH Arrhythmia plus 106,239 normal and 9987 PVC beats from INCART datasets
Comparison
ResNet-18 transfer learning model vs state-of-the-art methods
Design
Algorithm development and validation study using leave one subject out cross-validation
Authors
Loading...
May aid automated PVC screening on ECGs; leaves open prospective clinical validation before adoption.
A deep learning method using ResNet-18 and transfer learning achieved >99% accuracy for automatic PVC recognition on imbalanced ECG datasets.
Ullah et al. (2022) studied Premature ventricular contraction (PVC). ResNet-18 deep learning model with transfer learning was evaluated on Accuracy of PVC recognition using leave one subject out cross-validation. A deep learning method using ResNet-18 and transfer learning achieved PVC recognition accuracies of 99.93% and 99.77% on the MIT-BIH and INCART datasets, respectively.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: