New Method For Classifying Heart In Multiview Echocardiographic Images

Abstract

Echocardiography is a test that uses high-frequency sound waves to describe the structure of the heart. Echocardiography is used by doctors to analyze the movement of the walls in the heart chambers and identify heart disease. Several images, including the long-axis, short-axis, 2-chamber and 4-chamber left ventricle, can be used to check heart function. Many studies that have been carried out, including cardiac evaluation, are still carried out conventionally and require a certain level of accuracy. In this research, several methods proposed to achieve object extraction are used to build a classification system, the steps start with image enhancement, segmentation, tracking, extraction, output characteristics, validation and classification. Imaging enhancement aims to improve the echocardiographic image, thereby clarifying the edges of the heart wall. In addition, the images are reprocessed to separate the left ventricle from the heart wall and generate ventricular contours, at the segmentation stage. The contours are obtained by looking for the good features on each heart wall. In this approach, good features are identified only on the first image of the left ventricular slice. The good feature points used are 24 point which will be grouped into 6 segments. In addition, all images will be processed using the optical flow method to track the movement of the walls of the heart. Optical flow tracing will generate direction and distance feature extraction values that can be used to describe the resulting data features and find a suitable classification algorithm that is combined using different validation techniques, namely K-fold and Leave-one-out. In its implementation, Classifier Support Vector Machine (SVM) with rbf core achieves the highest accuracy. The SVM classification algorithm with validation techniques, namely k-fold cross-validation and leave-one-out cross-validation, reaches an accuracy value of 100% and 100%.