论文部分内容阅读
目的 探讨常规MRI三维纹理分析(TA)鉴别透明细胞肾细胞癌(ccRCC)、乳头状肾细胞癌(pRCC)和嫌色肾细胞癌(ChRCC)的价值.方法 回顾性分析经病理证实的74 例ccRCC、14 例pRCC和7 例ChRCC患者资料,手动勾画ROI分别提取肾皮髓质期、实质期及延迟期增强T1WI图像以上肿瘤三维纹理特征,筛选后应用非线性判别分析(NDA)方法进行分类分析,计算诊断准确率、敏感度、特异度及ROC曲线下面积(AUC).结果 常规MRI三维纹理分析鉴别ccRCC与pRCC的诊断准确率、敏感度、特异度可达97.73%、100%和85.71%,鉴别ccRCC与ChRCC、pRCC与ChRCC的诊断准确率、敏感度、特异度可达100%;NDA方法对三个图像集的纹理特征判别均获得较高的诊断效能(AUC 0.886～1.000).结论 常规MRI三维纹理分析有助于鉴别ccRCC、pRCC和ChRCC.“,”Objective To evaluate the diagnostic performance of three-dimensional (3D) texture analysis (TA) in conventional MRI for the classification of clear cell (ccRCC), papillary (pRCC) and chromophobe renal cell carcinoma (ChRCC). Methods A retrospective review was performed on patients with ccRCCs (n＝74), pRCCs (n＝14) or ChRCCs (n＝7) confirmed by pathology. Corticomedullary phase (CMP), nephrographic phase (NP) and delayed phase (DP) CE-MR images obtained from all the patients were used for texture analysis. 314 3D texture features were extracted from each of the three image series, and 30 important features were selected separately for each pair of ccRCCs, pRCCs and ChRCCs. Texture analysis was performed using nonlinear discriminant analysis (NDA). Classification accuracy, sensitivity, specificity and area under the receiver operator characteristics curve (AUC) were calculated. Results For ccRCC vs pRCC, the classification accuracy, sensitivity, specificity of 3D texture analysis in conventional MRI were up to 97.73%, 100% and 85.71%; for ccRCC vs ChRCC and pRCC vs ChRCC, the classification accuracy, sensitivity, specificity were all up to 100%. For all the pairs of ccRCCs, pRCCs and ChRCCs, classification performed the best in NDA (AUC 0.886～1.000) within each of the three image series. Conclusion 3D texture analysis in conventional MRI can be a reliable quantitative approach for differentiating ccRCC from pRCC or ChRCC and pRCC from ChRCC.