A Robust Circular Control Point Detector for Bi-Planar Spine Surgery Navigation System

In this paper, we propose a robust circular control point detector for bi-planar spine surgery navigation system. First, this detector divides the regions into three cases, including real control points, possible control points, and false control points, based on support vector machine in which quantity of edge pixels, closed-loop attribute, and distance criterion are selected as the features. Next, the valid control points are extracted from possible control points based on improved HVCD (Horizontal and vertical search for circle detection) method, which has improved minor-radius circles detection in two ways: one is parameter adjustment for search principle and the other is the promotion of the edge segment detector. Finally, the effective control points could be obtained and classified into two classifications by using the distance density clustering algorithm. Experimental results demonstrate that both detectors proposed in this paper and EDCircles could obtain effective control points on images of spine model bone while the detector proposed in this paper is obviously superior to EDCircles on images of swine bone and swine which have strong background interference. In addition, the parameters used in this paper are fixed for all the images.