Echocardiography Based Cardiac Ischemic Detection Using Red-CNN Classifier

Abstract

One of the most vital and hardest-working organs of the human body is the Heart. Nevertheless, a condition termed Cardiac Ischemia or CAD is engendered by the inadequate supply of blood along with oxygen to the heart muscle. So, with the aid of a Regularisation Dropout-centered Convolutional Neural Network (ReD-CNN) classifier, an efficient CAD detection system is proposed. At first, the input videos and the data are amassed from the Echonet Dynamic dataset. Next, frame conversion is undergone by the input video. For further processing, keyframes are extracted and preprocessed after frame conversion. Afterward, by means of a Contraction Expansion-Residual Network (CE-ResNet), the heart chambers are segmented, and then the features are extracted. Simultaneously, utilizing the Feynman Kan Formalism-based Tissue Doppler Imaging (FKF-TDI) approach, wall motion is estimated. Meanwhile, the features are extracted from the data and grouped via Average Distance-centric Fuzzy C Means (AD-FCM). After that, mapping of the features extracted from the frames and the grouped features occurs and the resultant data is divided into training, testing, along with validation stages. Then, utilizing the Cosine Similarity-centered Circulatory System Based Optimization (2(CS)BO) methodology, the Feature Selection (FS) is done. Therefore, to compute the Ischemic index, the features that are selected and the estimated wall motions are wielded. Lastly, for effective classification, the Ischemic index and the selected features are fed into the ReD-CNN classifier.  Finally, to ascertain the superiority of this technique, the proposed system’s performance is contrasted with the prevailing methodologies.