Research Of Real-time Automatic Registration In Optical Surgical Navigation

With the appearance of surgical navigation system,surgery is getting more minimally invasive and accurate.Based on computer technologies,surgical navigation system can show the accurate corresponding relationship between image data and the anatomical structure of the patient in the virtual scene with real-time,so that the doctor can get the spatial location of the surgical instruments visually during surgery.Nowadays,surgical navigation system in developed areas such as Europe,USA has been widely used,but used little in domestic clinical practice.The main reasons for this phenomenon are the high cost,complex process of operations,out line with Chinese doctor’s work habits and so on.Thus,for the purpose of extending the application of surgical navigation system,and with the view of practical point,our research group has built a near infrared optical surgical navigation system with self-owned intellectual property rights,and the workflow of the system is designed more simple and convenient.This paper mainly studies the automatic registration technology in surgical navigation system,to solve the problems of complex process of operations and out line with Chinese doctor’s working habits.The main contributions of this paper are:1.Based on the requirement of automatic registration,a new kind of marker is designed.The marker is designed considering from two aspects on material and structure to ensure that it can be automatically extracted in both image space and real space.The marker has three parts: the base,the pillar and a reflective ball,the base and the pillar are designed to be integrated structure,with high CT value materials,so that the markers can be separated out from image with some simple image operations.As the marker is designed to a special shape,the coordinates of markers can be automatically obtained with the method of model matching in the image space.In the real space,optical tracking system can capture the near infrared light reflected by the reflective ball,then the coordinates of markers can be computed with the method of threshold searching and gray weighted average algorithm.2.This paper proposes automatic methods to extract markers coordinates in the image space and real space respectively.In the image space,the extraction of markers coordinates is based on the markers’ special materials and shape structure.The high CT value character of the materials makes it possible to separate markers out from the image easily by threshold segmentation,image filter,image smoothing and other image processing operations.Then the model point cloud is constructed according to the marker’s shape structure,the extraction of markers coordinates in the image space is based on the matching between the model point cloud and the markers point cloud.Firstly,the coarse matching is performed by means of aligning their symmetric axis.Then the ICP algorithm is used to modify the result,the optimal matching result can be achieved.Compared with using ICP algorithm alone,this method has advantages such as small computation amount,fast and avoiding local optima.In the real space,the extraction of markers coordinates depends on the light spots which is formed from the near infrared light reflected by the reflective balls.According to the actual conditions that the workplace scenario of the optical tracking system in surgical navigation system is relatively simple,the markers coordinates in two dimension images can be computed by two steps: find the right light spots in left and right images respectively by threshold searching first,then compute the coordinates with the gray weighted average method.After getting the two dimension coordinates,stereo matching is processed with the method of epipolar constrain,then the three dimension coordinates of markers are acquired.In the process,maybe false matching will present,which can be eliminated according to the spots’ geometric information.As the method is proposed from the practical application scene of surgical navigation system,it has the advantages of simple arithmetic,high feasibility and high real control,which meets the requirement of localization algorithm in surgical navigation system.3.A quick,accurate spatial registration method is proposed based on point distance and ICP algorithm.The coarse matching is done by computing the distance of one marker to the other markers for every marker in both image space and real space at first,then sorting the distances of each marker and finding the corresponding marker which have the very similar distance network with the other marker in the other space.Secondly,based on coarse matching,the ICP algorithm is introduced to do precise match.The method overcomes the shortcomings of using the sole algorithm in spatial registration,such as sole algorithm cannot meet the requirement of briefness in calculation,high precision and robustness simultaneously.The method can take account of both matching accuracy and efficiency,and has robustness.