摘要
为了获得更多甲状腺肿瘤的诊断信息,文章提出了一种基于改进CPD算法的甲状腺肿瘤的SPECT图像和B超图像自动配准方法。首先,将参考图像(SPECT图像)和待配准图像(B超图像)分别通过阈值分割和图割的方法提取轮廓特征点;然后,用遗传算法和粒子群算法相结合的优化算法对CPD的权重参数进行自动寻优,利用改进的CPD算法对提取出的两组特征点进行匹配,得到两幅图像的空间变换参数;最后,将待配准图像按所求得的变换参数旋转、平移,从而获得配准后图像。实验结果表明,该方法能实现SPECT与B超甲状腺肿瘤图像的良好配准,具有参数少、精度高、鲁棒性好等特点。
In order to obtain more information on the diagnosis of thyroid tumor, an automatic registration meth-od of SPECT image and B-type ultrasound image of thyroid tumour based on improved CPD algorithm are pro-posed. Threshold segmentation and GCBAC segmentation are used to extract the outline’s feature points of the refer-ence image（SPECT image）and input image（B-type ultrasound image）.A new hybrid algorithm with combined GA and PSO is proposed to search optimization for CPD. The improved CPD algorithm is used to match both groups of feature points,and then the space transformation parameters of two images are obtained. The input image is rotated and shifted by the space parameters,then the registration image is obtained. The experiments show that the method can realize good registration of the thyroid SPECT image and B-type ultrasound image , and this meathod has the ad-vantages of less parameter, high precision and good robust.
出处
《激光杂志》
CAS
CSCD
北大核心
2014年第7期19-22,共4页
Laser Journal
基金
河北省卫生厅科研基金项目(20120395)
河北省教育厅科学技术研究重点项目(ZD20131086)
