摘要
目的探讨纹理组学预测乳腺癌基因标志物Ki-67表达水平的可行性。方法回顾分析40例经手术或穿刺、病理证实为乳腺癌,并有相关免疫组织化学Ki-67表达水平的临床资料。根据2011年乳腺癌国际指南中Ki-67表达水平14%作为分级标准,将病例分为高表达组24例,低表达组16例。选取病灶T2WI序列最大层面,运用Mazda软件沿病灶边缘勾画感兴趣区。通过Mazda软件自带的特征参数提取法,交互信息提取法(MI)、Fisher系数提取法和分类错误概率联合平均相关系提取法(POE+ACC),3种方法各提取10个纹理参数,选取在3种提取方法中出现2次或2次以上的纹理参数。应用SPSS软件对服从正态分布的参数行两独立样本t检验,不服从正态分布的参数行Wilcoxon秩和检验。对有统计学差异的参数和Ki-67表达水平行Spearman相关性分析,并绘制受试者工作特征(ROC)曲线。结果采用Mazda软件最终提取到6个特征参数,长游程补偿水平方向(HorzlLngREmph)、长游程补偿垂直方向(VertlLngREmph)及熵值(Entropy)S(5,-5)具有统计学意义(P值均<0.05),而第99百分位数(Perc.99%)、参数σ(Sigma)和游程灰度非均匀性45°方向(45dgrGLevNonU)不具有统计学差异(P值均>0.05)。熵值S(5,-5)与Ki-67表达水平呈正相关,r值为0.353,P值为0.026。熵值、长游程补偿水平及垂直方向的曲线下面积分别为0.751、0.724、0.724,阈值分别为8.279、8.556、8.416,对应的敏感性、特异性分别为83.3%、75.1%;91.7%、62.5%;87.5%、62.5%(P值<0.05),熵值诊断效能最高。结论基于T2WI的长游程补偿值及熵值有助于定量预测乳腺癌Ki-67表达水平,熵值与Ki-67呈正相关性,可为临床治疗乳腺癌提供新的切入点。
Objective To investigate the the feasibility of texture analysis to predict the expression level of Ki-67 in breast cancer.Methods 38 patients who were confirmed as breast cancer by surgical treatment and pathological examination,and had the expression level of ki-67 were analyzed retrospectively.The cases were divided into two groups based on the expression level of ki-6714%,including 23 cases in the high-level group and 15 cases in the low-level group according to the recommendations from the International Ki-67 in Breast Cancer Working Group.The Region of interest(ROI)was placed on T2WI images with maximum lesion and texture analysis was performed by Mazda software.The feature selection methods included MI,Fishers coefficient,and POE+ACC,which was used to identify the most significant texture features in discriminating different levels of Ki-67 and each of the method extracts 10 parameters.Finally,six parameters were extracted that appear twice or more in the three methods.Two independent sample t-tests were conducted for parameters that conform to normal distribution,and Wilcoxon rank sum test was conducted for parameters that do not conform to normal distribution.Spearman correlation analysis was performed,and an ROC was established.Results In the 6 parameters extracted from the texture analysis among the 2 groups,Horzl_LngREmph,Vertl_LngREmph and S(5,-5)Entropy had statistical difference(P<0.05).However,Perc.99%、Sigma and 45dgr_GLevNonU had no statistical difference(P>0.05).S(5,-5)Entropy was positively correlated with the expression level of ki-67,with a correlation coefficient of 0.353(P<0.05).The AUC for S(5,-5)Entropy,Horzl_LngREmph,and Vertl_LngREmph were 0.751,0.724,and 0.724,respectively,with optimal thresholds of 8.279,8.556,and 8.416,and Sensitivity and Specificity values of 83.3 and 75.1%,91.7%and 62.5%,and 87.5%and 62.5%.Conclusion The LngREmph and Entropy could contribute to quantitatively predicting the expression level of ki-67 index,with which Entropy has a positive correlation,and provide a new entry point for clinical treatment of breast cancer.
作者
张兵
吴晶涛
叶靖
陆大军
袁为标
曹正业
沈力
徐圆
ZHANG Bing;WU Jingtao;YE Jing(Department of Imaging,the Affiliated Jianhu Hospital of Nantong University,Yancheng,Jiangsu Province 224700,P.R.China)
出处
《临床放射学杂志》
CSCD
北大核心
2020年第4期680-684,共5页
Journal of Clinical Radiology
基金
国家自然基金项目(编号:81571652)
关键词
乳腺癌
KI-67
磁共振成像
纹理分析
Breast cancer
Ki-67
Magnetic resonance imaging
Texture analysis
