Abstract

Diseases of Cardiovascular (CVDs) remain a significant world health concern, highlighting the urgent need for early and accurate detection. This paper explores the integration of deep learning (DL) and machine learning (ML) to automate cardiovascular disease (CVD) detection in Electrocardiogram (ECG) images. Our proposed methodology exploits advancements in artificial intelligence to bolster the diagnostic capabilities of healthcare systems. We preprocess raw ECG signals to extract pertinent features, subsequently serving as input for both ML and DL models. ML methods such as Random Forests (RF), Support Vector Machines (SVM), and K-Nearest Neighbours (KNN) are initially employed for feature-based classification. Furthermore, deep learning models, including Convolutional Neural Networks (CNNs) and Long Short-Term Memory Networks (LSTMS), are utilized to obtain intricate patterns and physical dependencies within the ECG data. A diverse and extensive dataset of ECG images is utilized to train and validate our proposed models, ensuring robust performance across various cardiovascular conditions. This research signifies a significant stride towards integrating advanced technologies into cardiology, with the ultimate goal of enhancing patient outcomes alleviating healthcare burdens. In this work, we capitalize on publicly accessible ECG image datasets from cardiac patients to harness the power of deep learning techniques to predict the four main cardiac abnormalities: irregular heartbeat, myocardial infarction, history of myocardial infarction, and normal person classes. First, we investigate the transfer learning strategy leveraging SqueezeNet and AlexNet, two deep neural networks that have already undergone training. Next, we present a new architecture for convolutional neural networks (CNNs) specifically designed for the prediction of cardiac abnormalities.