A convolutional neural network achieved an F1 score of 83% on the test set for classifying short ECG segments, compared to 79% for a state-of-the-art feature-based classifier.
Does a Convolutional Neural Network improve the F1-score for detecting arrhythmia from short ECG segments compared to a feature-based classifier?
Deep learning algorithms using ResNet architectures can effectively classify short ECG recordings into normal and arrhythmic classes without requiring hand-engineered features, outperforming traditional feature-based classifiers.
Absolute Event Rate: 83% vs 79%
The diagnosis of cardiovascular diseases such as atrial fibrillation (AF) is a lengthy and expensive procedure that often requires visual inspection of ECG signals by experts. In order to improve patient management and reduce healthcare costs, automated detection of these pathologies is of utmost importance.
Andreotti et al. (Thu,) conducted a other in Arrhythmia (Atrial Fibrillation) (n=12,186). Convolutional Neural Network (ResNet) vs. Feature-based classifier was evaluated on F1 score for classifying ECG segments into four classes (AF, normal, other, noise) on the hidden test set. A convolutional neural network achieved an F1 score of 83% on the test set for classifying short ECG segments, compared to 79% for a state-of-the-art feature-based classifier.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: