Application of various CNNs for the Identification of Hypertrophic Cardiomyopathy

Hypertrophic Cardiomyopathy (HCM) is an abnormal condition of the heart due to the thickening of heart muscles. Early diagnosis and prognosis of HCM is appreciable as it can be a threaten to life if left untreated. The structural definitions of the heart are easily captured by Echocardiography. An automated system utilizing Echocardiography for the identification of HCM is thus beneficial to the well-being of the society. This study performs a comparative analysis of various Convolutional Neural Networks (CNNs), that are used for extracting the features which are relevant for the classification of HCM. Heart ultrasound images from 35 subjects each of HCM and normal are utilized for the study. Variants of Support Vector Machine (SVM) are used to perform the classification. Further, significant features are analyzed by feature ranking methods. Finally, it is observed from the result that ResNet50 and AlexNet are the best CNN models towards the identification of HCM.