基于全变分正则化去卷积的PET部分容积校正

发布时间：2018-09-05 05:49

【摘要】：目的提出全变分正则化去卷积方法并应用于PET图像部分容积校正。方法将全变分(toal variation,TV)引入到图像退化模型中,提出基于全变分正则化的Van Cittert(VC)和Richardson-Lucy(RL)去卷积方法。结果提出的方法分别应用于NCAT仿真数据、NEMA NU4-2008 IQ物理体模和肿瘤小鼠数据(均采用西门子小动物Inveon PET扫描得到)。相比于传统VC和RL去卷积方法,本文提出方法在仿真实验中校正图像有明显的去噪和保持边缘效果,同时在小鼠实验中肿瘤区域的活度值增加率为(10±1.8)%时,图像标准方差(standard deviation,SD)增加率分别从49.98%下降到14.26%和42.76%下降到4.70%。结论本方法能够在校正PET部分容积效应的同时有效抑制噪声增加,可望更为准确地诊断肿瘤。
[Abstract]:Aim to propose a total variational regularization deconvolution method and apply it to partial volume correction of PET images. Methods Total variation (toal variation,TV) was introduced into the image degradation model and a Van Cittert (VC) and Richardson-Lucy (RL) deconvolution method based on total variational regularization was proposed. Results the proposed method was applied to the physical model of NCAT NU4-2008 IQ and the data of tumor mice (obtained by Siemens small animal Inveon PET scan) respectively. Compared with the traditional VC and RL deconvolution methods, the proposed method has obvious denoising and edge preserving effects in the simulation experiments, and at the same time, the increase rate of the activity of tumor area in mice is (10 卤1.8)%. The increase rate of image standard variance (standard deviation,SD) decreased from 49.98% to 14.26% and 42.76% to 4.70%, respectively. Conclusion this method can be used to correct the partial volume effect of PET and to suppress the increase of noise effectively, which is expected to be more accurate in the diagnosis of tumor.
【作者单位】：南方医科大学生物医学工程学院;南方医科大学南方医院PET中心;
【基金】：国家自然科学基金(81101046,81371544) 教育部高等学校博士学科专项科研基金(20134433120017) 广东省医学科研基金立项课题(B2014239,B2014240) 广东省自然科学基金(2014A030310243)~~
【分类号】：R817

【参考文献】