B-spline based Motion Tracking of Cardiac Surface in Minimally Invasive Surgery

摘要

This paper investigates the problem of tracking heart motion of patients when performing minimally invasive cardiac surgeries. In the existing literature, most algorithms for tracking the heart motion are based on a region or feature of the heart surface. A disadvantage of the former method is consuming time, and hence difficult to meet the demand of clinical surgery. The focus of this paper is on the latter, i.e. a feature of the heart surface. First, we employ the BRIEF feature for describing and matching the binary features on the cardiac surface. Based on the matching result, we then propose a new, fast and stable tracking algorithm via the Bspline interpolation. The sparse motion field of the heart image is obtained by matching the binary features, and the motion of the target point is built by using the B-spline interpolation method. In this way, positions of the target points on the subsequent heart images can be obtained. Finally, we use the triangulation algorithm to compute the 3D coordinates of the target point. The novelty of our method includes the binary features for describing the heart surface and an efficient Bspline algorithm for 3D tracking of the beating heart using stereo images from a calibrated stereo endoscope. The proposed tracking method is evaluated on in vivo images acquired by a DaVinci surgical robotic platform. Many different regions of the heart surface over 600 images have been tracked. The data is analyzed by Fourier transformation to reveal the heart motion. The experimental results demonstrate the effectiveness of the presented method.