Left Ventricular Movement Feature Extraction: A New Method for Classifying Heart Condition in Four-Chamber and Two-Chamber Views

Abstract

The American heart association recommends examining four heart angles, i.e., the long axis, short axis, two-chamber, and a four-chamber section of the left ventricle, during the inspection of hearts' condition. Currently, left ventricle observation to assess the heart's condition is commonly done manually. No studies have performed an extraction of features from the movement of the left ventricular wall in the four-chamber and two-chamber views. The presence of an automatic method in observing and evaluating the left ventricle condition will help experts diagnose heart conditions. Methods: This study proposes an automatic tracking method for the left ventricular heart cavity, thereby obtaining the value of heart movement to build a classification system. It involves several stages: image preprocessing, image segmentation, tracking, feature extraction, classification, and validation. Results: The initial preprocessing stage produces images that have clear differences between the heart wall and the left ventricle. This clear distinction becomes the characteristic in separating the left ventricle from the heart wall, thus obtaining a contour using the segmentation method. The obtained contour line is then used as a good-feature using the intersection line. This research uses 24 good-features points which are only defined in the first frame. Furthermore, all frames will be processed using the optical flow Lucas-Kanade method to track the movement of the heart wall. Conclusion: Each good-feature produces four features values: direction (positive and negative) and distance (positive and negative); thus, a total of 96 features are obtained for the entire process. In its implementation, the support vector machine method obtains the highest accuracy value of 94,762 % and 94,977 % with validation techniques, k-folds, and leave-one-out. Significance: This study proposes a method for extracting features from left ventricular movement.