Published Paper Details:

Electrocardiogram (ECG)-based arrhythmia detection plays a pivotal role in modern cardiac healthcare, offering early diagnosis of potentially fatal heart conditions. With the advent of deep learning, especially Convolutional Neural Networks (CNNs), automatic detection of arrhythmias has significantly improved. This paper provides a comprehensive review of recent advancements in ECG arrhythmia detection, emphasizing the application of 2D CNNs on time-frequency representations of ECG signals. We present our implementation of a 2D CNN-based model evaluated on the MIT-BIH Arrhythmia Database, achieving an accuracy of 86.12%. Additionally, we explore emerging approaches including transformer-based architectures (e.g., ECG-BERT), self-supervised learning, and federated learning. These innovations aim to enhance model generalization, address data scarcity, and ensure patient data privacy. The study concludes with proposed directions for future research, including real-time deployment on wearable devices and hybrid models combining CNNs with transformers.