基于ADC图纹理分析预测直肠癌美国肿瘤联合会分期

Texture analysis based on ADC images in predicting American Joint Committee on Cancer staging of rectal cancer

目的观察基于ADC图像纹理分析预测直肠癌美国肿瘤联合会(AJCC)分期的价值。方法回顾性分析102例直肠癌的ADC图像,其中AJCC分期Ⅰ~Ⅱ期49例、Ⅲ~Ⅳ期53例。采用MaZda软件自动提取纹理特征参数,并获得最佳纹理参数;以B11模块进行纹理特性分类分析,比较模块中原始数据分析(RDA)、主成分分析(PCA)、线性判别分析(LDA)及非线性判别分析(NDA)4种方法评估AJCC分期的最小错误率。采用组内相关系数(ICC)评估纹理特征值可重复性,筛选最佳纹理参数;以Spearman相关性检验分析最佳纹理参数与AJCC分期的相关性。采用ROC曲线分析纹理特征联合临床指标对AJCC分期的诊断效能。结果B11模块中,RDA、PCA、LDA及NDA分析法诊断AJCC分期的最小错误率分别为16.67%(17/102)、18.63%(19/102)、3.92%(4/102)和8.82%(9/102)。共16个纹理参数与AJCC分期存在相关性,其中Mean3D、S(1.0.0)SumAverg、Perc.50%与AJCC分期的rs较高,分别为-0.96、-0.96和-0.92(P均<0.01)。ROC曲线结果显示,纹理特征联合临床指标对诊断AJCC分期的AUC为0.80,敏感度和特异度分别为0.84和0.72。结论基于ADC图像纹理分析可评估直肠癌AJCC分期,为术前制定治疗方案提供帮助。

Objective To explore the predictive value of texture analysis based on ADC image in American Joint Committee on Cancer(AJCC)staging of rectal cancer.Methods ADC images of 102 patients with rectal cancer confirmed by pathology were retrospectively analyzed,including 49 cases of stageⅠ-Ⅱand 53 cases of stageⅢ-Ⅳaccording to AJCC staging.MaZda software was used to automatically extract texture feature parameters and to obtain the best texture parameters.B11 module was used to classify and analyze the texture characteristics,and the minimum error rate of different analysis methods(raw data analysis[RDA],principal component analysis[PCA],linear discriminant analysis[LDA]and nonlinear discriminant analysis[NDA])for identifying AJCC stages were compared.The intra-class correlation coefficient(ICC)was used to evaluate the repeatability of eigenvalues,and the best texture parameters were selected.Spearman correlation test was used to analyze the correlation of the best texture parameters and AJCC stages.The diagnostic efficiency of texture features combined with clinical indexes in AJCC staging by ROC curve analysis.Results In B11 module,the minimum error rate of RDA,PCA,LDA and NDA analysis for AJCC staging was 16.67%(17/102),18.63%(19/102),3.92%(4/102)and 8.82%(9/102),respectively.Sixteen texture parameters were correlated with AJCC staging,among which,|rs|between Mean3D,S(1.0.0)SumAverg,Perc.50%and AJCC staging were higher(rs=-0.96,-0.96 and-0.92,all P<0.01).ROC curve showed that AUC of texture features combined with clinical indexes for diagnosis of AJCC stage was 0.80,sensitivity and specificity was 0.84 and 0.72,respectively.Conclusion Texture analysis based on ADC images could be used to evaluate AJCC staging of rectal cancer and provide valuable information for preoperative treatment of rectal cancer.