[1]许 璐,余顺园.基于关注区域分割和 SURF 算子的医学图像配准[J].计算机技术与发展,2021,31(11):216-220.[doi:10. 3969 / j. issn. 1673-629X. 2021. 11. 035]
　XU Lu,YU Shun-yuan.Medical Image Registration Based on Region of Interest Segmentation and SURF Detector[J].,2021,31(11):216-220.[doi:10. 3969 / j. issn. 1673-629X. 2021. 11. 035]
点击复制

基于关注区域分割和 SURF 算子的医学图像配准()
分享到：

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:
31
期数:
2021年11期
页码:
216-220
栏目:
应用前沿与综合
出版日期:
2021-11-10

文章信息/Info

Title:
Medical Image Registration Based on Region of Interest Segmentation and SURF Detector
文章编号:
1673-629X(2021)11-0216-05
作者:
许 璐; 余顺园
安康学院 电子与信息工程学院,陕西 安康 725000
Author(s):
XU Lu; YU Shun-yuan
School of Electronics and Information Engineering,Ankang University,Ankang 725000,China
关键词:
医学图像配准; 关注区域分割; SURF 算子; 双重匹配; RANSAC 算法
Keywords:
medical image registration; region of interest segmentation; SURF detector; dual matching; RANSAC algorithm
分类号:
TP39
DOI:
10. 3969 / j. issn. 1673-629X. 2021. 11. 035
摘要:
医学图像配准在术前模拟、术中导航、病理检测和疗效跟踪方面应用广泛。 然而,受拍摄环境和拍摄区域的限制,直接获取的待配准医学图像不仅包含患病部位的身体组织密度信息,还包含了诸如 CT 扫描仪的部件、患者的衣服、一些特殊的传感器等无关干扰信息, 这些非关注区域对医学图像配准造成了一定的干扰。 针对上述问题,文中提出了一种基于关注区域分割的医学图像配准算法。 首先对待配准的医学图像进行关注区域分割,去除无关干扰信息,突出显示病患诊断必须的、医生重点关注病患部位;然后基于特征匹配的思想,利用 SURF( speeded up robust features) 检测特征描述子,采用双重匹配策略实现特征点的配对,最后基于 RANSAC( random sample consensus) 算法去除误匹点,实现医学图像自动准确的配准。 实验结果表明,提出的关注区域预分割思路能够有效提升特征检测与定位的准确性,医学图像特征点的匹配正确率在 90% 以上。
Abstract:
Medical image registration is widely used in preoperative simulation, intraoperative navigation, pathology detection and therapeutic effect tracking. However,due to the limitations of the shooting environment and shooting areas,the initial medical image to be registered not only contains the density information of the body tissue of the diseased part,but also contains some irrelevant information such as CT scanner components,patient clothes,and some special sensors. These non-interest areas greatly cause interference to medical image registration. As to the above problem,we propose a medical image registration algorithm based on region of interest segmentation.Firstly,the region of interesting is segmented to remove the irrelevant information. As a result,the parts that are necessary for patient diagnosis and the doctor’s focus are highlighted. Then the medical images are registrated based on feature matching. The feature descriptors are achieved with SURF (speeded up robust features) detector,and the feature points are mached based on the dual matching strategies. Finally,the RANSAC ( random sample consensus) algorithm is adopted to remove mismatch points,and the automatic and accurate registraton of medical images is achieved. The experiment shows that the proposed pre-segmentation of the region of interest can effectively improve the accuracy of feature detection and positioning,and the matching accuracy of feature points in the medical image registration process is above 90% .
更新日期/Last Update: 2021-11-10