Tree structures segmentation of coronary arteries using CURVES method for interventional surgical simulation

Visualization model of the coronary vasculature is of utmost importance for the diagnosis of the coronary heart diseases, as well as the planning and navigation of the intravascular surgery. To protect the cardiologists and operation staff against the ionizing radiation, surgical robots are designed and come to assist the practitioners during the interventional procedure. Robotic surgical simulation aims to provide effective, economic and convenient support for the learners with the surgical details as real as possible. In building this system, the geometric model of the blood vessels especially the coronary arteries are the key part of the virtual anatomic scenario. Because of the complex topologies, the segmentation of the coronary arteries is full of challenges. We developed in this paper a semi-automatic approach for this challenging work. The approach is based on an improved geodesic active contours model called CURVES. Firstly, the region that contains the whole heart in the original images are completely extracted. Secondly, the extracted volumetric data is smoothed and thresholded in order to remove noises and irrelevant details. Next the image features are generated by calculating the gradients pixel-wisely, while the initial contours are generated by a modified fast marching method. Then the contour evolution is provoked to segment the boundaries of the coronary arteries. Finally the surface model is visualized after information is organized by using the marching cubes method. Experimental results showed the capability of the proposed approach in the segmentation of the coronary arteries tree.