Hybrid Image Processing and Deep Learning Framework for Automated Coronary Artery Disease Classification Using Echocardiogram and Angiography

In this study, we present an integrated image processing and deep learning framework for the automated classification of coronary artery disease (CAD) using echocardiogram and angiography images. The proposed approach begins with image enhancement techniques to improve visual quality and highlight diagnostically relevant structures. Following enhancement, K-means segmentation is applied to accurately isolate the coronary artery regions of interest, reducing background noise and enhancing feature localization. The segmented images are then processed using the pre-trained VGG16 convolutional neural network, which serves as a robust deep feature extractor. These high-dimensional deep features are subsequently utilized for binary classification into two categories: CAD and Normal. The choice of echocardiogram and angiography images ensures complementary diagnostic perspectives, improving classification reliability based on data availability. The combination of classical image processing methods with advanced deep learning techniques enables precise detection of CAD while minimizing human intervention. Experimental results demonstrate the model’s potential to achieve high accuracy, making it a promising decision-support tool for cardiologists. This work underscores the importance of hybrid methodologies in medical image analysis and highlights the role of deep learning in facilitating early and accurate diagnosis of CAD, ultimately contributing to improved patient outcomes and treatment planning.

Keywords: Disease Prediction, Deep Learning, VGG16, Segmentation, Classification.