2-D ECG Spectral Imaging: A New Approach to Arrhythmia Detection

Abstract- -Electro cardiogram (ECG) monitoring is critical for detecting heart arrhythmias associated with cardiovascular diseases. This report focuses on an image-based ECG heartbeat classification methodology for detecting arrhythmias by using deep convolutional neural networks (CNNs). The model processes grayscale ECG image inputs and classifies them into one of seven categories mentioned- normal or one of six arrhythmia types. A web application was developed to perform as an interface between the model and the user, allowing users to upload ECG images and receive a prediction. The report describes the end-to-end model development process including dataset preprocessing, CNN architecture design, model training and evaluation. Training approaches were explored on a local machine. Comparable performance was achieved demonstrating 96% prediction accuracy. Overall, the developed methodology showcases an effective computer aided arrhythmia detection system using deep learning and ECG imaging. Such artificial intelligence methods can play a vital role in early diagnosis, leading to timely precautionary checkups for a particular disease. Further optimizations of the model could elevate performance even higher to benefit real-world cardiology based disease and treatment.

DOI: /10.61463/ijset.vol.12.issue1.113