基于CT图像的影像组学特征鉴别肾透明细胞癌与血管平滑肌脂肪瘤

The radiomic features based on CT images ddistinguish clear cell renal cell carcinoma and angiomyolipoma

目的基于CT图像的影像组学特征分析鉴别肾透明细胞癌和血管平滑肌脂肪瘤。方法选取经手术病理证实的110例透明细胞癌患者及28例血管平滑肌脂肪瘤患者的术前增强CT扫描图像,采用ITK-SNAP软件逐层手工勾画整个病灶,通过Pyradiomcs软件获得病灶内影像组学特征进行统计学分析,绘制受试者工作特征(ROC)曲线筛选出对鉴别两者性能较好的特征参数。结果共筛选出29组诊断性能较好的参数,包括一阶特征(n=16)、形态特征(n=1)、灰度共生矩阵特征(n=7),灰度相关矩阵特征(n=2),灰度运行长度矩阵特征(n=3),其诊断两者的特异性、敏感度、准确率均较高。结论基于增强CT图像的影像组学特征参数对于鉴别肾透明细胞癌和血管平滑肌脂肪瘤有较好的诊断效能。

Objective To analyze the radiomic features analysis based on CT images so as to distinguish renal clear cell carcinoma(ccRCC)and angiomyolipoma(AML).Methods We retrospectively collected preoperative enhanced CT scan images of 110 ccRCC patients and 28 AML patients confirmed by surgery and pathology.The entire lesion was manually delineated layer by layer using ITK-SNAP software,and all radiomic features in the lesion were obtained by Pyradiomcs software.The features were statistically analyzed,and the receiver operating characteristic(ROC)curve was drawn to select the features with better performance in identifying the two.Results 29 groups of parameters with better diagnostic performance were screened out,including first order(n=16),Shape(n=1),grey level co-occurrence matrix(n=7),grey level dependence matrix(n=2),grey level run length matrix(n=3).The specificity,sensitivity and accuracy of diagnosis ccRCC and AML were all high.Conclusion The characteristics of radiological features based on enhanced CT images have better diagnostic efficacy for the differentiation of renal clear cell carcinoma and vascular smooth muscle lipoma.