摘要
为了提高甲状腺肿瘤检出的准确率,提出基于改进蝙蝠算法的甲状腺SPECT(single-photo emission computed tomogropby)-B超图像配准方法.针对甲状腺SPECT图像与B超图像灰度差异大,采用2类图像共有的甲状腺及肿瘤轮廓特征进行配准.采用阈值分割法提取SPECT图像中甲状腺及肿瘤轮廓;采用Shearlet变换与基于活动轮廓模型图割算法相结合的方法提取B超图像中甲状腺及肿瘤轮廓;以归一化互信息为相似性测度,以改进的蝙蝠算法为优化算法,优化配准所需的空间变换参数.实验结果表明,提取的B超图像中甲状腺及肿瘤轮廓更准确,改进的蝙蝠算法使配准具有更高的准确性和较好的鲁棒性.
In order to improve the accuracy in thyroid tumor detection, a novel registration method of SPECT image and B-type ultrasound image based on improved bat algorithm was put forward. In view of large difference in gray level of SPECT image and B-type ultrasound image, we registered two kinds of im- age using common the contour characteristic of thyroid and tumor. Threshold segmentation was used to extract the contour of thyroid and tumor of SPECT image. Segmentation combined Shearlet transform and GCBAC was used to extract the contour of thyroid and tumor of B-type ultrasound image. The normalized mutual information was used for similarity measure and an improved bat algorithm was used to optimize the space registration transform parameters. The experimental result showed that our method extracting the contour of thyroid and tumor of B-type ultrasound image was more accurate and improved bat algo- rithm could improve the accuracy and robustness of registration.
出处
《河北大学学报(自然科学版)》
CAS
北大核心
2015年第5期520-525,共6页
Journal of Hebei University(Natural Science Edition)
基金
河北大学医工交叉研究中心开放基金资助项目(BM201103)