Automatic atrium contour tracking in ultrasound imaging

Abstract

Cardiac ultrasound imaging tracking is an important issue in medical image analysis. The results of tracking greatly influence the diagnoses and judgment of physicians. However, the conventional active contour model is inappropriate for use in cardiac imaging tracking since the mitral and tricuspid valves rise and fall, leading to poor tracking and excessive convergence in overall contour during systoles and diastoles. In this study, a novel tracking approach, combining the active contour model and scale invariant feature transform, is proposed for use in cardiac ultrasound imaging tracking. In addition to preprocessing that removes some noise and detects initial cardiac edges automatically, the active contour model is used to segment and track cardiac regions. The scale invariant feature transform is proposed to determine and track the correct positions of heart valves. The performance of the proposed method is evaluated by testing with cardiac ultrasound image sequence and in comparison with five other tracking techniques in terms of several metrics, such as segmentation accuracy, AUC, the Dice coefficient, and modified Hausdorff distance. The results indicate that the proposed method is more accurate and effective in the tracking of cardiac imaging.