Automated Segmentation of the Right Ventricle in 3D Echocardiography: A Kalman Filter State Estimation Approach

Abstract

As the right ventricle's (RV) role in cardiovascular diseases is being more widely recognized, interest in RV imaging, function and quantification is growing. However, there are currently few RV quantification methods for 3D echocardiography presented in the literature or commercially available. In this paper we propose an automated RV segmentation method for 3D echocardiographic images. We represent the RV geometry by a Doo-Sabin subdivision surface with deformation modes derived from a training set of manual segmentations. The segmentation is then represented as a state estimation problem and solved with an extended Kalman filter by combining the RV geometry with a motion model and edge detection. Validation was performed by comparing surface-surface distances, volumes and ejection fractions in 17 patients with aortic insufficiency between the proposed method, magnetic resonance imaging (MRI), and a manual echocardiographic reference. The algorithm was efficient with a mean computation time of 2.0 s. The mean absolute distances between the proposed and manual segmentations were 3.6 ± 0.7 mm. Good agreements of end diastolic volume, end systolic volume and ejection fraction with respect to MRI ( -26±24 mL , -16±26 mL and 0 ± 10%, respectively) and a manual echocardiographic reference (7 ± 30 mL, 13 ± 17 mL and -5±7% , respectively) were observed.

Semiautomated biventricular segmentation in three-dimensional echocardiography by coupled deformable surfaces

“…Finally, we previously presented a method for RV segmentation based on a statistical shape model and Kalman filter state estimation. 12 The method presented in this paper allows a coupled segmentation of endo-and epicardial borders of the LV and RV. The endo-and epicardial surfaces are represented as coupled deformable surfaces and fitted to the image using a real-time segmentation framework previously applied to the LV 5 and RV 12 separately.…”

“…12 The method presented in this paper allows a coupled segmentation of endo-and epicardial borders of the LV and RV. The endo-and epicardial surfaces are represented as coupled deformable surfaces and fitted to the image using a real-time segmentation framework previously applied to the LV 5 and RV 12 separately. Incompressibility of the myocardium is introduced by regularizing the myocardial volume during the cardiac cycle.…”

Three‐dimensional‐based landmark tracker employing a superpixels method for neuroscience, biomechanics, and biology studies

“…KF for motion analysis uses some observed measurements over time to estimate variables related to the motion. KF has frequently been employed to predict the position of objects in different fields, human tracking, mice tracking, or cardiovascular disease detection . The KF model assumes that a state of a system for frame number n evolves from the prior state at frame number n −1 .…”

“…KF has frequently been employed to predict the position of objects in different fields, human tracking, 49 mice tracking, 50 or cardiovascular disease detection. 51 The KF model assumes that a state of a system for frame number n evolves from the prior state at frame number n−1. 52 We extended usual 2D KF to a 3D KF considering three states and three measurements to update the coefficients of KF.…”

Segmentation of Carpal Bones Using Gradient Inverse Coefficient of Variation with Dynamic Programming Method

“…In boundary detection, dynamic programming (DP) method has been implemented in many applications including in biomedical imaging field applications [21], such as the segmentation of myocardium [22], right [23] and left ventricles [24], breast [25], and lung [26]. The basic idea of DP is to determine the optimum path between two given points, in which those two points are also optimum lying on the path [27].…”