摘要
目的:分析影像组学特征在不同设备不同扫描参数中的可再现性。方法:在3台CT设备上设置不同扫描参数:重建层厚(1、3、5 mm)、重建核函数(和CT设备有关)、曝光时间(50、100、300 mAs),扫描2种模体,提取影像组学特征并评估特征可再现性。结果:81.6%的特征在不同层厚条件下可再现性差(变异系数CV>10%);有8.8%的特征在3台设备中均满足要求(CV<10%)。46.4%的特征在不同重建核函数条件下可再现性差;有12.1%的特征在3台设备中均满足要求(CV<10%);占总数31.7%的特征只在单一设备中达到可再现性要求。54%的特征在不同曝光时间条件下可再现性差。结论:应用影像组学建立模型研究临床问题时,应充分考虑并避免扫描参数对影像组学特征可再现性的不利影响,以获得更好的临床研究结果。
Objective To analyse the reproducibility of radiomics features in different CT equipment with various scanning parameters.Methods Different scanning parameters were set on three CT machines:different reconstruction slice thickness(1,3,5 mm respectively),different reconstruction kernel function(related to specific CT machines),different exposure time(50,100,300 mAs respectively),and two kinds of phantom were scanned to extract radiomics features and evaluate their reproducibility.Results 81.6%of the features had poor reproducibility(CV>10%)under different slice thickness,and only 8.8%of the features in the three machines met the requirements(CV<10%);46.4%of the features have poor reproducibility under different reconstruction kernel functions,and only 12.1%of the features in three machines meet the requirements(CV<10%);31.7%of the features met the reproducibility requirements only in a single machine;54%of the features have poor reproducibility under different exposure time.Conclusion When using radiomics to establish a model to study clinical problems,the adverse effects of scanning parameters on the reproducibility of imaging histology features should be fully considered and avoided in order to obtain better clinical research results.
作者
张白霖
戴振晖
朱琳
杨耕
朱远湖
招什武
蔡春雅
李飞
王学涛
ZHANG Bailin;DAI Zhenhui;ZHU Lin;YANG Geng;ZHU Yuanhu;ZHAO Shiwu;LI Fei;CAI Chunya;WANG Xuetao(Radiotherapy Center,the Second Affiliated Hospital of Guangzhou University of Chinese Medicine,Guangzhou,510006,China)
出处
《中国医学物理学杂志》
CSCD
2021年第1期61-64,共4页
Chinese Journal of Medical Physics
基金
广东省科技厅自筹经费类项目(2017ZC0165)
广东省医学科学技术研究基金(A2019196)
广东省医学会医学工程学分会青年委员会科研课题(2019-GDMAZD-01)。
