肺癌B1场校正T1-mapping影像组学特征的可重复性研究
Study on the reproducibility of B1-corrected T1-mapping-based radiomics features of lung cancer
江建芹 1刘佳 2崔磊 2肖勇 1符益纲 1徐高峰 1胡春洪3
作者信息
- 1. 224000 江苏盐城,徐州医科大学盐城临床学院(盐城市第一人民医院)影像科
- 2. 226000 江苏南通,南通市第一人民医院影像科
- 3. 215002 江苏苏州,苏州大学附属第一医院影像科
- 折叠
摘要
目的:探讨基于B1场校正T1-mapping测量肺癌影像组学特征的一致性及可重复性.方法:回顾性搜集2020年6月—2022年3月在本院经病理确诊的70例肺癌患者的术前胸部T1-map-ping和T1加权容积内插屏气检查(T1-VIBE)图像.其中,有41例进行了T1-mapping重复扫描(间隔30 min).由2位测量者独立采用半自动勾画方法分别在T1-mapping和T1-VIBE图像上逐层沿着病灶边缘勾画ROI,融合生成全病灶的容积ROI(VOI),然后提取原始图像及不同变换预处理图像上病灶VOI的影像组学特征.采用组内相关系数(ICC)评价组学特征在观察者内、观察者间的一致性及重复扫描时的可重复性,以ICC≥0.80为一致性或可重复性好(满足临床要求).结果:自T1-mapping和T1-VIBE图像分别可提取1906个组学特征(包括一阶特征396个、形态特征14个及纹理特征1496个).基于T1-mapping提取的组学特征在观察者内及观察者间的ICC的中位数(Q1,Q3)分别为0.94(0.87,0.98)和0.91(0.83,0.96),基于T1-VIBE提取的组学特征在观察者内及观察者间的ICC的中位数(Q1,Q3)均为0.98(0.95,0.99);2个序列分别有1450(72.72％)和1768(91.76％)个组学特征同时满足观察者内和观察者间的一致性要求.T1-mapping重复扫描提取的组学特征的ICC的中位数(Q1,Q3)为0.83(0.69,0.93),有1002(52.57％)个组学特征同时满足重复扫描时的可重复性;其中,形状特征的可重复性表现最佳(ICC≥0.80者达100％),而一阶特征、灰度共生矩阵(GLCM)及灰度级区域矩阵(GLSZM)特征的可重复性相对较差(ICC≥0.80者仅占49.75％～53.98％).基于不同变换预处理图像提取的组学特征中,Gradient变换明显提高了原始特征的可重复性(ICC≥0.80者达86.05％),而采用Exponential、LBP2D、LBP3D、Logarithm和Square变换方法获得的特征降低了原始特征的可重复性(ICC≥0.80这仅占36.05％～50.00％).结论:T1-mapping的影像组学特征表现出中等可重复性,在临床应用之前需要首先对其进行可重复性评估.
Abstract
Objective:The aim of this study was to investigate the repeatability and reproducibili-ty of B1-corrected T1-mapping-based radiomics features of lung cancer.Methods:Chest images of T1-mapping and T1-volume interpolated breath-hold examination (VIBE)of 70 patients with pathologi-cally confirmed lung cancer in our hospital from June 2020 to March 2022 were collected and analyzed retrospectively.Among them,41 patients underwent repeated T1-mapping scans with an interval of 30min.Radiomics features were extracted from T1-mapping and T1-VIBE using the semiautomatic seg-mentation method to delineate the volume of interest by two independent observers based on the origi-nal and pre-processing different transformation images.Interclass correlation coefficients (ICC)was used to evaluate the repeatability of radiomics features within and between observers,as well as the re-producibility of repeated scans,and good repeatability and reproducibility was defined as ICC≥0.80 (meeting clinical requirements).Results:1906 features were extracted from T1-mapping and T1-VIBE,including 396 first-order features,14 morphological features,and 1496 texture features.The intra-and interobserver ICCs of T1-mapping-based radiomics features were 0.94 (0.87,0.98)and 0.91 (0.83,0.96);the intra-and inter-observer ICCs of T1-VIBE-based radiomics features were 0.98 (0.95,0.99);with 1450 (72.72％)and 1768 (91.76％)radiomics features meeting both intra-and inter-ob-server agreement,respectively.The ICC of repeated T1-mapping scans was 0.83 (0.69,0.93),with 1002 (52.57％)T1-mapping-based radiomics features meeting test-retest reproducibility at the same time.Among them,the shape features displayed the best reproducibility in repeated scanning (the ratio of ICC≥0.80 was 100％),while the first-order features,GLCM,and GLSZM texture features exhibi-ted relatively poor reproducibility (the ratios of ICCs with values ≥0.80 were 49.75％～53.98％).A-mong the features based on different transformations,the reproducibility of the original features was significantly improved in Gradient transformation (the ratio of ICC with values ≥0.80 was 86.05％),while the reproducibility of the original features were reduced in exponential,LBP2D,LBP3D,Lo-garithm,and Square transformations (the ratios of ICC with values ≥0.80 were 36.05％～50.00％).Conclusion:The radiomics features of T1-mapping showed moderate reproducibility,and agreement e-valuation is necessary before clinical application.
关键词
肺癌/影像组学/T1-mapping/容积内插屏气检查/可重复性Key words
Lung cancer/Radiomics/T1-mapping/Volume interpolated breath-hold examination/Reproducibility引用本文复制引用
基金项目
盐城市重点研发计划社会发展指导性项目(2022)(YCBE202211)
徐州医科大学附属医院科技发展基金(XYFY202304)
出版年
2024
NETL
NSTL
万方数据