Adaptive Despeckling and Heart Disease Diagnosis by Echocardiogram using Optimized Deep Learning Model

Abstract

ABSTRACT Echocardiography is a significant diagnostic tool intended to interpret and identify different categories of heart diseases. The classification of heart condition by utilising the CAD on echocardiogram is an essential and primary step, which should be performed in a precise manner. This process aims to achieve a strong differentiation among abnormal and normal heart conditions through machine learning algorithms. As the heart disease diagnosis using echocardiography is still a complex process, it can be adopted with different deep learning approaches for identification. For solving these issues, this paper develops a novel heart disease diagnosis using echocardiogram images and video frames. As speckle noise is the main problem in echocardiogram images, noise reduction is the initial phase of the process. This is performed by the adaptive speckle-noise reduction technique. Then, the hybrid pattern extraction is adopted using two well-performing techniques termed local binary pattern and local directional texture pattern. This extracted hybrid pattern is subjected to the optimised Convolutional Neural Network (CNN). Here, optimised CNN is accomplished by the distance-based cat swarm optimisation for enhancing the diagnosis accuracy of heart disease. Finally, the proposed work shows the efficiency of the developed echocardiogram-based CAD heart disease diagnosis system by conducting the experiments.