摘要
目的评估MRI纹理分析在直肠癌术前T分期中的应用价值。方法选取经手术及病理证实的61例直肠癌患者术前盆腔增强MRI图像资料,采用Omni-Kinetics软件,于轴位T_(1)WI、T_(2)WI及增强T_(1)WI上,在包含肿瘤实性成分最大的层面测量,软件自动生成直方图、形态学矩阵等4类纹理特征的纹理参数,比较pT_(1)-2期与pT3-4期直肠癌各参数的差异,并对有差异的参数绘制R0C曲线,评估其效能。结果统计得到峰度、能量及熵值等37种纹理特征参数,其中在T_(1)WI序列上有峰度、熵等5个纹理特征在2组数据中差异有统计学意义(P<0.05),其中熵具有较高的鉴别效能,AUC值为0.765,最佳阈值为5.34,敏感度65.9%,特异度80%;在T_(2)WI序列上有最大强度、峰度等19个纹理特征参数差异有统计学意义(P<0.05),其中最大强度具有较高的鉴别效能,AUC值为0.788,最佳阈值为2.185,敏感度58.5%,特异度95%;在T_(1)WI增强序列上有体积数等4个纹理特征差异有统计学意义(P<0.05),其中体积数具有较高的鉴别效能,AUC值为0.786,最佳阈值为2.075,敏感度70.7%,特异度85%。结论基于磁共振T_(1)WI、T_(2)WI及增强T_(1)WI图像的信号纹理分析有助于辅助直肠癌术前T分期的诊断,与pT_(1)-2期相比,pT3-4期肿瘤具有更复杂的纹理特征,其中基于T_(2)WI序列的最大强度值是最佳的分期指标。
Objective To evaluate the value of MR texture analysis in preoperative T staging of rectal cancer.Methods From November 2016 to November 2018,61 patients with rectal cancer confirmed by operation and pathology in the first hospital of jiaxing were analyzed retrospectively.Using omni kinetics software,on the axial T_(1)WI,T_(2)WI and enhanced T_(1)WI,on the level containing the most solid components of the tumor,the software automatically generated four kinds of texture features,such as histogram and morphological matrix.The parameters of pT_(1)-2 and pT3-4 were compared,and the ROC curve was drawn to evaluate the efficacy of these parameters.Results 37 texture feature parameters such as kurtosis,energy and entropy are obtained,among which 5 texture features such as kurtosis and entropy are statistically significant in the T_(1)WI sequence(P<0.05),in which entropy has higher discrimination efficiency,with AUC value of 0.765,the best threshold of 5.34,sensitivity of 65.9%,specificity of 80%.There are 19 texture characteristic parameters such as maximum intensity and kurtosis on T_(2)WI sequence have statistical significance(P<0.05),in which the maximum intensity has a higher discrimination efficiency,with AUC value of 0.788,the best threshold of 2.185,sensitivity of 58.5%,and specificity of 95%.There are 4 texture features such as volume number on T_(1)WI enhancement sequence have statistical significance(P<0.05),in which volume number has a higher discrimination efficiency,with AUC value of 0.786,the best threshold of 2.075,sensitivity of 70.7%,and specificity of 85%.Conclusion Signal texture analysis based on T_(1)WI,T_(2)WI and enhanced T_(1)WI images is helpful to the diagnosis of preoperative T stage of rectal cancer.Compared with pT_(1)-2 stage,pT3-4 stage tumor has more complex texture features,and the maximum intensity based on T_(2)WI sequence is the best staging index.
作者
孙延豹
黄震霄
黄益镌
林英楠
丁健
吴凡
SUN Yanbao;HUANG Zhenxiao;HUANG Yijuan;LIN Yingnan;DING Jian;WU Fan(Department of Radiology,First hospital of Jiaxing,Radiology Department,Jiaxing 314000,China)
出处
《医学影像学杂志》
2022年第4期643-647,共5页
Journal of Medical Imaging
关键词
直肠肿瘤
磁共振成像
纹理分析
肿瘤分期
Rectal tumor
Magnetic resonance imaging
Texture analysis
Tumor stage