常规T2WI纹理分析预测早期宫颈鳞癌盆腔淋巴结转移的价值

Value of Conventional T2-Weighted Images Texture Analysis in Predicting Pelvic Lymph Node Metastasis in Early-Stage Cervical Cancer

目的初步探讨常规T2WI纹理分析预测早期宫颈鳞癌盆腔淋巴结转移的价值。方法回顾性分析103例经手术病理证实为早期宫颈鳞癌且行盆腔淋巴结清扫术患者的临床和影像资料。依据淋巴结病理结果分为淋巴结转移组和无淋巴结转移组。采用MaZda软件于轴位T2WI图像上手动勾画感兴趣区(ROI),提取宫颈病灶的纹理特征参数。通过软件特征选择方法:Fisher系数、交互信息(MI)、分类错误概率联合平均相关系数(POE+ACC)以及三者联合应用(FPM)各选取最佳纹理特征参数。采用B11统计模块的初始数据分析法(RDA)、主要成分分析法(PCA)、线性判断分析法(LDA)、非线性判断分析法(NDA)对各选取的纹理特征参数进行分类分析。结果以错判率的形式表述。由2名具有MRI诊断经验的医师综合评估影像资料,以盆腔淋巴结的病理结果为金标准,统计影像医师的错判率,并采用χ^2检验比较其与纹理分析结果的差异。结果103例中,淋巴结转移组31例,无淋巴结转移组72例。转移组中影像医师错判22例,无转移组错判22例,影像医师诊断淋巴结转移的错判率为42.72%(44/103)。纹理分析中,NDA联合FPM特征选择方法的错判率最低(12.62%),低于联合其他三种(Fisher、MI、POE+ACC)方法(18.45%、18.45%、14.56%)。采用纹理分析早期宫颈癌盆腔淋巴结转移的错判率较影像医师的错判率低,两者差异有统计学意义(χ^2=13.39,P<0.001)。结论常规T2WI序列纹理分析对早期宫颈鳞癌患者盆腔淋巴结转移有一定的预测价值,有望为临床治疗方案提供更可靠的客观依据。

Objective To evaluate the value of the conventional T2-weighted imaging(T2WI)texture analysis in predicting the pelvic lymph node metastasis in patients with early-stage cervical cancer.Methods The clinical and radiological data of 103 patients with pathological diagnoses of early-stage cervical cancer who were performed pelvic lymphadenectomy were analyzed retrospectively.According to the pathological results,the pelvic lymph nodes were divided into metastatic lymph node(MLN)group and non-metastatic lymph node(NLN)group.Texture features were extracted from manually drawn ROIs on axial T2WI by using MaZda software.The optimal texture parameters were selected according to four different methods including Fishers coefficient,mutual information(MI),classification error probability combined with average correlation coefficients(POE+ACC),and the combination of the above three methods.The selected texture parameters were analyzed by B11 statistic module including raw data analysis(RDA),principal component analysis(PCA),linear discriminant analysis(LDA)and nonlinear discriminant analysis(NDA).The results were demonstrated by misclassification rate.Two radiologists with ample experience in MRI diagnosis comprehensively evaluated the image data.Taking the pathological results of pelvic lymph nodes as the gold standard,the misjudgment rate of the radiologists was evaluated,and the difference between the results of the texture analysis and theχ^2 test were compared.Results MLN group included 31 patients(30.10%)and the remaining 72 patients were in NLN group(69.90%).There were 22 cases misclassified by radiologists in MLN and NLN group separately with misclassification rate of 42.72%(44/103).NDA combined with FPM had the lowest misclassification rate of 12.62%in comparison with the other three methods(Fisher 18.45%、MI 18.45%、POE+ACC 14.56%).Misclassification rate of the radiologists’analysis was higher than that of the texture analysis for predicting the pelvic lymph node metastasis in patients with early-stage cervical cancer and there was significant difference between them(χ^2=13.39,P<0.001).Conclusion Conventional T2WI texture analysis can be used to predict the pelvic lymph node metastasis in patients with early-stage cervical cancer and it will provide reliable and objective evidence for the clinical treatment in the future.