王旭,刘磊,刘义军,李贝贝,范勇,童小雨,王诗耕.采集模式与能谱成像能级对CT影像组学特征稳定性的影响:体模研究[J].中国医学影像技术,2023,39(6):910~914
采集模式与能谱成像能级对CT影像组学特征稳定性的影响:体模研究
Impact of acquisition mode and gemstone spectral imaging energy level on stability of CT radiomics features: A phantom study
投稿时间:2022-10-28  修订日期:2023-01-19
DOI:10.13929/j.issn.1003-3289.2023.06.025
中文关键词:  体层摄影术,X线计算机  体模,显像术  影像组学
英文关键词:tomography, X-ray computed  phantoms, imaging  radiomics
基金项目:
作者单位E-mail
王旭 大连医科大学附属第一医院放射科, 辽宁 大连 116011  
刘磊 大连医科大学附属第一医院放射科, 辽宁 大连 116011 liuleidmu1989@163.com 
刘义军 大连医科大学附属第一医院放射科, 辽宁 大连 116011  
李贝贝 大连医科大学附属第一医院放射科, 辽宁 大连 116011  
范勇 大连医科大学附属第一医院放射科, 辽宁 大连 116011  
童小雨 大连医科大学附属第一医院放射科, 辽宁 大连 116011  
王诗耕 大连医科大学附属第一医院放射科, 辽宁 大连 116011  
摘要点击次数: 3174
全文下载次数: 528
中文摘要:
      目的 基于体模实验观察不同采集模式及能谱成像(GSI)能级对CT影像组学特征稳定性的影响。方法 依次向PH-75体模中置入模拟肝组织、脂肪及碘(4、8、12 mgI/ml共3种浓度)的等价材料棒。采用常规120 kVp及GSI模式、腹部扫描协议对体模进行3次扫描(噪声指数均为10),获得常规CT及虚拟单能量图像(VMI)[120 kVp-like (VMI74 ke V)及70 keV (VMI70 keV)]。手动勾画3种图像中各置入棒ROI,获得三维模型图,提取其7类共107个影像组学特征;计算各影像组学特征变异系数(CV),评价其稳定性:0 ≤ CV ≤ 5%为非常稳定,5%20%为不稳定。结果 各置入棒的组织形状特征均为非常稳定及稳定,即扫描模式及GSI能级变化对其影响均较小;其他6类特征主要为稳定及中等稳定,存在不稳定特征,即扫描模式及GSI能级变化对其影响较明显。针对肝脏软组织和脂肪组织棒,基于常规CT所获的影像组学特征多为非常稳定或中等稳定,基于GSI不同能级图像所获特征多为中等稳定或不稳定。对于4 mgI/ml及8 mgI/ml碘棒,基于VMI74 keV提取的影像组学特征中不稳定者分别占28.97%(31/107)及19.63%(21/107),基于常规CT及VMI70 keV多为非常稳定至中等稳定特征,不稳定特征在4 mgI/ml碘棒特征中分别占7.48%(8/107)及2.80%(3/107),在8 mgI/ml碘棒中分别占5.61%(6/107)及7.48%(8/107);对于12 mgI/ml碘棒,基于3种CT图像提取的影像组学特征均稳定,无不稳定特征。结论 不同采集模式及GSI能级均影响组织CT影像组学特征的稳定性,其影响程度与组织成分及强化程度相关。
英文摘要:
      Objective To observe the impact of different acquisition modes and gemstone spectral imaging (GSI) energy levels on the stability of CT radiomics features based on phantom experiments. Methods Rods simulating liver tissue, adipose tissue and iodine (4, 8 and 12 mgI/ml, respectively) were sequentially inserted into a PH-75 phantom. Conventional 120 kVp and GSI mode were used to scan the phantom with abdominal scanning protocol (noise index of 10) for 3 times to obtain conventional CT, 120 kVp-like (VMI74 keV) and 70 keV (VMI70 keV) virtual mono-energy images (VMI). ROI for each rod was manually delineated on 3 kinds images to obtain t 3D model maps, and 7 types tolly 107 radiomics features were extracted. The coefficient of variation (CV) of each radiomics feature was calculated to evaluate its stability:0 ≤ CV ≤ 5% was considered very stable, 5%20% was unstable. Results The shape features of all rods were very stable or stable features, indicating that the impact of acquisition mode and GSI energy level on shape features were slight. Among other 6 types of features, there were many stable and moderately stable features, but also some unstable features, indicating that the impact of acquisition mode and GSI energy level were obvious. For liver and adipose tissue, most of the radiomics features obtained based on conventional CT were very stable or moderately stable, while most of those obtained based on GSI images with different energy levels were moderately stable or unstable. For 4 mgI/ml and 8 mgI/ml iodine rods, the proportion of unstable features extracted based on VMI74 keV was 28.97% (31/107) and 19.63% (21/107), respectively, while those based on conventional CT and VMI70 keV were mostly very stable to moderately stable, the proportion of unstable features of 4 mgI/ml iodine rod was 7.48% (8/107) and 2.80% (3/107), of 8 mgI/ml iodine rod was 5.61% (6/107) and 7.48% (8/107), respectively. For 12 mgI/ml iodine rods, the radiomics features extracted based on all 3 kinds CT images were stable, no stable feature was found. Conclusion Different acquisition modes and GSI energy levels affected the stability of CT radiomics feature, with degrees related to tissue composition and enhancement.
查看全文  查看/发表评论  下载PDF阅读器