A novel 6-layer CNN approach using 2-D time-frequency wavelet transforms of ECG signals achieved 95.3% accuracy in classifying arrhythmias, demonstrating potential for clinical application.
Supports CNN-based ECG classification research; leaves open prospective clinical validation.
A new method for classifying cardiac abnormalities is here proposed based on the electrocardiogram (ECG). The ECG may manifest abnormal heart patterns, which are generally known as arrhythmias. MIT-BIH arrhythmia database and AAMI standards are used for machine learning purposes considering the patient-oriented scheme. Heartbeat time intervals and morphological features processed by a 2-D time-frequency wavelet transform of ECG signals are combined into an image, which carries relevant information from each heartbeat. These dataset images are used as input to train and evaluate the classifier, which is essentially a 6 layers convolutional neural network (CNN), resulting in powerful artifact discrimination. The training set is artificially augmented to reduce the imbalance of the five heartbeat classes, achieving better results. A significant achieved overall accuracy of 95.3% of the proposed method, compared to some of the most relevant published methods, permits to expect effective results when applied to real patients.
No takes yet. Share an insight, caveat, or question.
Oliveira et al. (2019) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: