AUTHOR=Song Min-Seo , Lee Seung-Bo 

TITLE=Comparative study of time-frequency transformation methods for ECG signal classification

JOURNAL=Frontiers in Signal Processing

VOLUME=Volume 4 - 2024

YEAR=2024

URL=https://www.frontiersin.org/journals/signal-processing/articles/10.3389/frsip.2024.1322334

DOI=10.3389/frsip.2024.1322334

ISSN=2673-8198

ABSTRACT=This study highlights the growing need for automated electrocardiogram (ECG) signal classification using deep learning to overcome the limitations of traditional electrocardiogram (ECG) interpretation algorithms, which can lead to misdiagnosis and are inefficient. The application of convolutional neural networks (CNN) to ECG signals is gaining significant attention owing to their exceptional image classification capabilities. However, this study was conducted to address the lack of standardized methods for converting 1D ECG signals into 2D-CNN-compatible input images by employing time-frequency methods and selecting hyperparameters associated with these methods, particularly the choice of function. Furthermore, the effects of fine-tuned training, a technique in which pre-trained weights are adapted to a specific dataset, on 2D-CNNs for ECG classification, were investigated. The experiments were conducted using the MIT-BIH Arrhythmia Database, focusing on the classification of premature ventricular contractions (PVCs) and abnormal heartbeats originating from the ventricles. Several CNN architectures pre-trained on ImageNet were employed and fine-tuned using the proposed ECG datasets. This study provides crucial insights into CNNs for ECG classification, underscoring the significance of fine-tuning and hyperparameter selection in image transformation methods. The findings of the study provide valuable guidance to researchers and practitioners, ultimately improving the accuracy and efficiency of ECG analysis using 2D-CNNs. Future research avenues may include advanced visualization techniques and extension of CNNs to multi-class classification, thereby expanding their utility in medical diagnosis.