Tracking Multidimensional Echocardiographic Image using Optical Flow

Abstract

Heart disease is still the leading cause of death in the world; therefore, a regular heart check needs to be done to find out the health of the heart. Echocardiographic tests that show heart rate movements are commonly used by doctors to analyze heart performance. Heart health conditions can be analyzed based on the movement of the heart wall and the area of the heart wall during diastolic or systolic conditions. This analysis process depends on the accuracy and observation experience of a doctor, and sometimes causes different analysis of each doctor. This study focuses on the use of a system that can track the movement of the heart wall using the Optical Flow from various points of view. First, the echocardiographic video will be captured into several frames. High boost filters and median filters are used to improve image quality. Next, the heart cavity wall will be segmented using the watershed method. After that, a good feature point is determined in the heart cavity. These good features point is the point that will be tracked. The tracking process in this research uses two methods of Optical Flow as a comparison, they are Optical Flow Lucas - Kanade and Optical Flow Farneback. Lucas-Kanade is one of the sparse Optical Flow method that can provide the flow vector of some interesting features within the frame. Whereas Farneback is one of dense Optical Flow method that tracks every pixel in a frame. Based on the experimental results of this study, with 30 data tested, the Lucas Kanade Optical Flow method has a higher average value than the Optical Flow Farneback method with the average accuracy of Optical Flow Lucas Kanade is 90.73% and an average accuracy of Optical Flow Farneback is 90.04%. In the experiment with each point of view, from 9 displacement data, Optical Flow Lucas Kanade has a higher average accuracy value than the Optical Flow Farneback.