基于局部归一化互信息联合约束的三维头颈部CT/MR配准方法
3D Head and Neck CT/MR Registration Method Based on Local Normalized Mutual Information Joint Constraint
商奥雪 1王玉 1王明泉 1贾虎 1成向北 1李文波1
作者信息
- 1. 中北大学信息与通信工程学院,山西 太原 030051
- 折叠
摘要
针对头颈部图像的复杂解剖结构和不同模态图像的差异,提出一种改进的多模态配准方法.该方法采用局部归一化互信息作为相似性测度,以捕捉局部图像区域的相似性,并减少不同区域间的强度差异对配准结果的影响.此外,为解决MR图像易受到噪声和伪影的影响产生畸变的问题,引入F范数约束配准过程,抑制不必要的畸变.实验结果表明,所提方法在降低均方误差、提高配准精度和鲁棒性方面均有所提高.
Abstract
An improved multimodal registration method is proposed for the complex anatomical struc-ture of head and neck images and the differences of different modal images.This method uses local normal-ized mutual information as a similarity measure to capture the similarity of local image regions and reduce the influence of intensity differences between different regions on the registration results.In addition,in or-der to solve the problem that MR images are susceptible to distortion caused by noise and artifacts,we in-troduce an F-norm constrained registration process to suppress unnecessary distortion.Experimental re-sults show that the proposed method improves the mean squared error,the registration accuracy and ro-bustness.
关键词
非刚性配准/互信息/F范数正则化/归一化/多模态配准Key words
non-rigid registration/mutual information/F-norm regularization/normalized mutual in-formation/multi-modal registration引用本文复制引用
基金项目
山西省重点研发计划资助项目(201803D121069)
山西省应用基础研究项目面上自然基金资助项目(201801D121162)
出版年
2024
国家科技期刊平台
NSTL
万方数据