Medical decision-Making based on Combined CRISP-DM Approach and CNN Classification for Cardiac MRI

Abstract

In this study, a combined CRISP-DM technique with a deep convolutional neural network (CNN) for medical decision classification is used to perform evaluating of myocardial subjects as well as left ventricular (LV) volume estimate from cardiac magnetic resonance (CMR) images. The medical prognosis is strongly influenced by the measurement. The results of the proposed model were compared with those of a deep CNN built using the decision tree learning method. The findings demonstrate that the K-nearest neighbor classifier (k=1 with 88% accuracy) and deep CNN architecture with a decision tree classified topics with good accuracy (62 percent). The principal component analysis (PCA) approach was used to classify and optimize the important characteristics of roundness, centroid (px), roughness, and eccentricity. The estimated features acquired from the trained network had a strong correlation with the computation of the ejection fraction using grey-level pixels around the myocardium.