Transformer Encoder Driven Deep Learning Network for Single Lead ECG Arrhythmia’s

Electrocardiogram (ECG) analysis plays a crucial role in detecting arrhythmia, a condition associated with irregular heart rhythms that can lead to severe cardiac complications. This project introduces a Transformer Encoder Driven Deep Learning Network for single-lead ECG arrhythmia detection. The system leverages FREQ-ECG-Net, a frequency-aware deep learning network, and QuadPiltransformer encoder-driven deep learning network robust classification of ECG signals. The model is trained and evaluated using the ECG Images dataset from Kaggle, comprising diverse arrhythmia classes. The architecture efficiently captures both temporal and spectral features of ECG signals, enabling high classification accuracy. The framework is supported by a Flask-based backend, with an interactive front-end developed using HTML, CSS, and JavaScript. Users can register, log in, and submit ECG images for automated classification, receiving results with highlighted arrhythmia classes. Performance metrics, including accuracy, precision, recall, and F1-score, demonstrate the system’s reliability. This study emphasizes explainability, ease of use, and scalable deployment in healthcare monitoring applications. Overall, the proposed approach offers a significant advancement in automated arrhythmia detection, supporting clinicians in decision-making processes and reducing diagnostic effort.

Keywords: ECG, Arrhythmia Detection, Transformer Encoder, FREQ-ECG-Net, QuadPillar, Deep Learning, Single Lead ECG, Classification, Neural Networks, Flask.