基于SIFT算子的脑立体定向手术标记点自动定位方法  

Marker localization in robot-assisted stereotactic neurosurgery based on SIFT operator

作　　者：曹彤[1] 王烁[1] 倪自强[2] 刘达[2] 

机构地区：[1]北京科技大学机械工程学院北京100086 [2]北京航空航天大学机器人研究所北京100191

出　　处：《高技术通讯》2014年第2期165-170,共6页Chinese High Technology Letters

基　　金：863计划(2009AA045301)资助项目

摘　　要：为提高脑立体定向手术中的图像注册效率,提出了一种基于尺度不变特征转换(SIFT)算子的标记点自动定位新方法。该方法使用了A、B两类人工标记点,先用SIFT算子对B类标记点进行粗检测,然后进行B类标记点的精确检测、A类标记点的检测,最后进行标记点的双目匹配和三维重建。实验表明,该方法在实现自动定位的同时能很好地适应相机与患者距离变化、光照变化等情况,克服相机与患者距离、光照等变化对自动定位精度的影响。标记点的漏检率低于0.43%,误检率均低于0.85%,平均定位误差在0.54mm以内。In order to improve the efficiency of image registration in robot assisted stereotactic neurosurgery, an approach for automatic detection and location of the markers glued on the head of patients based on the SIFT（ scale invariant feature transform）operator was developed. This approach adopts two sorts of artificial markers, named A and B. It uses the SIFI＂ operator to conduct the rough detection of B markers, and then, performs the accurate detection of B markers and the exact detection of A markers, and finally, the binocular match and 3D reconstruction of markers. The correct rate and positioning accuracy of the approach were tested by experiment, and the results show that the approach can overcome the influences of the camera-patient distance and the illumination on the localization accura- cy. The marker false rate is below 0.43%, lost rate is below O. 85%, and average location error is less than 0.54mm.

关 键 词：脑立体定向手术 图像注册 SIFT算子 人工标记点 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]