CHO数字观察器在PET图像滤波方法评估中的应用

发布时间：2018-06-19 05:29

本文选题：高斯滤波器 + 非局部均值(NLM)滤波器　；参考：《中国生物医学工程学报》2017年06期

【摘要】：在临床应用中需要限制扫描时间和药物剂量,这往往会使正电子发射断层扫描(PET)的图像的分辨率变低,噪声变多。为提供可供临床诊断的图像,去噪是一个必须的手段,而在重建后增加一个滤波器是目前最常用的去噪方法。因此对不同滤波器滤波效果的比较是PET图像重建中的重要环节,其中最关键的是滤波参数的选取。目前采用的信噪比(SNR)以及恢复系数(RC)等评估方法可以用来非定量地选取参数,研究者们只能凭经验选取最优参数。而通道化霍特林观察器(CHO)作为一个比较通用的数字观察器,已被用于与PET图像质量相关的各种参数的选择,如重建算法参数、系统设计参数、临床协议参数等,然而其在评估不同滤波方法对图像重建质量的影响中的应用研究还比较少。通过比较CHO计算得到的ROC(receiver operating characteristic)曲线下面积(area under the ROC curve,AUC),选择两种常用的滤波器(即高斯滤波器和非局部均值(Non-Local Mean,NLM)滤波器)的最优参数,并评估它们在PET中的滤波效果。结果表明,对于13 mm球体,σ为1.1~1.4的高斯滤波器和f为0.5~0.9的NLM滤波器可以达到最大的检测能力值,而对于10 mm球体,σ为1.4~2.0的高斯滤波器和f为0.5~0.9的NLM滤波器可以达到最大的检测能力值。虽然两个滤波器所对应的AUC值都能高达0.9,但是NLM滤波器的AUC值高于高斯滤波器。通过IEC图像和病人图像也能发现,NLM滤波后的PET图像中的亮点比高斯滤波的更加清晰,噪声更少。该结论和传统滤波器评估方法得到的结论一致,这说明在PET的病灶检测任务中,CHO能够准确地比较这两种滤波器的性能。
[Abstract]:It is necessary to limit the scanning time and drug dose in clinical application, which often makes the resolution of PET image lower and the noise more frequent. In order to provide images for clinical diagnosis, denoising is a necessary method, and adding a filter after reconstruction is the most commonly used denoising method. Therefore, the comparison of filtering effects of different filters is an important step in PET image reconstruction, and the selection of filter parameters is the most important one. The SNR (SNR) and the restoration coefficient (RC) can be used to select parameters in an unquantifiable way, but researchers can only select the optimal parameters based on experience. As a general digital observer, channelized Hotlin observer has been used to select various parameters related to PET image quality, such as reconstruction algorithm parameters, system design parameters, clinical protocol parameters, etc. However, there are few researches on how to evaluate the effect of different filtering methods on image reconstruction quality. The optimum parameters of two kinds of filters (Gao Si filter and Non-Local mean value Non-Local mean NLM filter) are selected by comparing the area under the receiver operating curve and the AUC filter, and their filtering effect is evaluated. The results are as follows: (1) in this paper, the optimal parameters of the two kinds of filters (Gao Si filter and Non-Local mean value Non-Local mean value NLM filter) are selected. The results show that for the 13mm sphere, the Gao Si filter with 蟽 = 1.1 ~ 1.4 and the Gao Si filter with f = 0.5 ~ 0.9 can reach the maximum detection capability. For the 10 mm sphere, the Gao Si filter with 蟽 = 1.4g 2.0 and the Gao Si filter with f = 0.5 ~ 0. 9 can reach the maximum detection capability. Although the AUC value of the two filters can be as high as 0.9, the AUC value of the Gao Si filter is higher than that of the Gao Si filter. It can also be found that the bright spot of the filtered Gao Si image is clearer and less noise than that of the Gao Si filter. This conclusion is consistent with that obtained by the traditional filter evaluation method, which shows that Cho can accurately compare the performance of the two filters in the focus detection task of PET.
【作者单位】：杭州电子科技大学生物医学工程学院;明峰医疗系统股份有限公司;
【基金】：国家科技支撑计划项目课题(2012BAI13B06) 国家重点研发计划项目(2016YFC0104500) 国家自然科学基金(81671038)
【分类号】：R817;TP391.41

【相似文献】