基于体模和反卷积算法的PET呼吸运动模糊校正  

Correcting Respiratory Motion Blur in PET Images Based on Phantom and Deconvolution Algorithm

作　　者：孟祥宏[1] 陈仰纯[2] 郑冬琴[1] 

机构地区：[1]暨南大学物理系,广州510632 [2]广州医科大学第一附属医院PET-CT中心,广州510230

出　　处：《中国生物医学工程学报》2013年第6期655-662,共8页Chinese Journal of Biomedical Engineering

摘　　要：呼吸运动会导致PET图像出现运动模糊,影响肿瘤诊断的准确性和放射治疗的精确性。本研究结合高频正弦振动和反卷积技术提出了一种校正PET图像运动模糊的方法。高频正弦振动用于模拟肺部肿瘤的呼吸运动。首先使用雷当变换从运动模糊图像的伪倒谱中识别模糊运动方向,然后将运动模糊图像的模糊方向旋转到垂直模糊方向,利用双谱识别模糊幅度,最后使用Richardson-Lucy迭代算法对运动模糊图像进行校正。体模实验显示,通过校正后PET图像估算出的肿瘤体积和肿瘤内平均标准摄取值(SUV)更接近真实值,与未校正的模糊运动图像相比,其校正后的肿瘤体积误差从40%下降到10%,SUV误差从28%下降到4%。结果表明所使用的方法能够有效校正呼吸运动模糊,使肿瘤诊断更加准确。Motion blur in PET images induced by respiratory motion may influence the accuracy of tumor diagnosis and treatment. In this paper, a method of correcting the motion blur, which combined high-frequency sinusoidal vibration model and deconvolution technology, was proposed. The high-frequency sinusoidal vibration model is used to simulate the tumor motion due to respiration. The blur direction was identified by performing Radon transform on the quasi-cepstrum of motion blurred image. The blur extent was identified from bispectrum of motion blurred image rotated to make the motion direction vertical. Based on this, Richardson-Lucy iterative algorithm is used to correct respiratory motion blurs. Experiments on phantom images show that both the tumor volume and mean standard uptake value （SUV） derived from restored images are closer to the true values. Compared to un-correction motion blur images, the error of tumor volume estimation was decreased from 40% to 10% , and the error of SUV is decreased from 28% to 4%. The results demonstrate that the proposed correction approach can improve the accuracy of PET quantification by correcting the respiratory motion blurs.

关 键 词：呼吸运动 模糊校正 点扩散函数 正电子发射断层摄影术 

分 类 号：R318[医药卫生—生物医学工程]