基于非局部自相似图像块字典学习的伪CT图像预测

发布时间：2018-03-27 05:23

本文选题：非局部自相似性　切入点：字典学习　出处：《信号处理》2017年03期

【摘要】：随着PET/CT技术的日益发展,其被广泛应用于现代放射治疗。但在采集数据过程中,对人体放射时间较长,辐射当量较大,增加了患者的痛苦,因此人们希望减少CT扫描中X射线的辐射。为解决这一问题,本文提出基于非局部自相似图像块字典学习的伪CT图像预测方法。首先,对训练CT与MRI图像进行图像分块,通过块匹配算法聚类CT图像块,并提取CT与MRI图像块的多尺度特征。其次,通过字典学习,获得MRI图像与CT图像的映射关系矩阵,并对CT图像块进行预测。最后,通过重构算法,从目标MRI图像中得到预测CT图像。仿真实验证明了提出算法相对基于图谱集算法的有效性,以及在现代放射治疗中利用MRI图像替代CT图像的应用前景。
[Abstract]:With the development of PET/CT technology, it is widely used in modern radiotherapy. However, in the process of collecting data, the radiation time of human body is longer, the radiation equivalent is larger, and the pain of patients is increased. Therefore, people want to reduce X-ray radiation in CT scanning. In order to solve this problem, a pseudo-CT image prediction method based on nonlocal self-similar image block dictionary learning is proposed in this paper. Firstly, the training CT and MRI images are divided into blocks. The block matching algorithm is used to cluster CT image blocks and extract the multi-scale features of CT and MRI image blocks. Secondly, through dictionary learning, the mapping relation matrix between MRI images and CT images is obtained, and the CT image blocks are predicted. Computed tomography (CT) images are obtained from target MRI images by reconstruction algorithm. The simulation results show that the proposed algorithm is more effective than the atlas set based algorithm and the application prospect of replacing CT images with MRI images in modern radiotherapy.
【作者单位】：天津大学电子信息工程学院;滨州学院信息工程学院;天津商业大学信息工程学院;
【基金】：国家自然科学基金项目(61340034) 滨州学院科研基金项目(BZXYG1112)
【分类号】：R814.42;TP391.41

【相似文献】