摘要
目的探讨CT纹理分析对鉴别乏脂肪肾错构瘤与肾透明细胞癌的价值。方法回顾性分析经手术病理证实的16例乏脂肪肾错构瘤与79例肾透明细胞癌的CT增强图像;通过纹理分析的方法测得其平均值、标准差、熵、不均匀度、峰值、偏度等定量参数,并进行统计学分析。结果两位观察者测得的CT纹理分析定量参数的一致性分析结果如下:平均值、标准差、熵、不均匀度、峰值及偏度的Cronbachα系数分别为:0.97、0.93、0.97、0.94、0.56、0.68。纹理分析定量参数中差、熵、不均匀度三个参数两组间比较结果具有显著差异:标准差(t=3.60,P<0.01),熵(t=4.80,P<0.01),不均匀度(t=3.86,P<0.01);利用标准差鉴别两组肿瘤的曲线下面积、阈值、敏感性、特异性、准确性分别为:0.78、45.46、70.9%、81.2%、72.6%;利用熵鉴别两组肿瘤的曲线下面积、阈值、敏感性、特异性、准确性分别为:0.82、4.50、84.8%、68.8%、82.1%;利用不均匀度鉴别两组肿瘤的曲线下面积、阈值、敏感性、特异性、准确性分别为:0.80、0.09、70.9%、81.2%、72.6%。利用多参数联合鉴别肾乏脂肪错构瘤与肾透明细胞癌的效能:标准差+熵鉴别两组肿瘤的曲线下面积、敏感性、特异性、准确性分别为:0.83%、75.0%、81.0%、80.0%;标准差+不均匀度鉴别两组肿瘤的曲线下面积、敏感性、特异性、准确性分别为:0.81%、81.2%、70.1%、71.6%;熵+不均匀度鉴别两组肿瘤的曲线下面积、敏感性、特异性、准确性分别为:0.83、81.2%、70.1%、72.6%;标准差+熵+不均匀度鉴别两组肿瘤的曲线下面积、敏感性、特异性、准确性分别为:0.84、87.5%、69.6%、72.6%。结论 CT纹理分析的部分定量参数(标准差、熵、不均匀度)可用于鉴别乏脂肪错构瘤与肾细胞癌。
Objective To evaluate CT texture analysis to differentiate fat-poor angiomyolipoma（ fp-AML） from clear cell renal cell carcinoma（ ccRCC） on enhanced computed tomography（ CT） images. Methods Retrospective analysis of 16 patients with fp-AML and 84 patients with ccRCC,with all cases having been confirmed by operative pathological examination,underwent CT enhancement examination. Through using the method of CT texture analysis,we can get many parameters： Mean,StDev,Entropy,Inhomogenity,Skewness,Kurtosis,and so on. Then,using SPSS16. 0 software to analyze these quantitative parameters. Results There are statistically significan differences of St Dev,Entropy and Inhomogenity between fp-RAML and ccRCC： StDev（ t = 3. 60,P〈0. 01）; Entropy（ t = 4. 80,P〈0. 01）; Inhomogenity（ t =3. 86,P〈0. 01）. When using StDev to distinguish fp-RAML and ccRCC,its AUC,cut-off point,sensitivity,specificity and accuracy respectively is 0. 78,45. 46,70. 9%,81. 2%,72. 6%; when using Entropy to distinguish fp-RAML and ccRCC,its AUC,cut-off point,sensitivity,specificity and accuracy respectively is 0. 82,4. 50,84. 8%,68. 8%,82. 1%; when using Inhomogenity to distinguish fp-RAML and ccRCC,its AUC,cut-off point,sensitivity,specificity and accuracy respectively is 0. 80,0. 09,70. 9%,81. 2%,72. 6%. When using StDev and Entropy to distinguish fpRAML and ccRCC,its AUC,sensitivity,specificity and accuracy respectively is ： 0. 83,75. 0%,81. 0%,80. 0%;when using StDev and Inhomogenity to distinguish fp-RAML and ccRCC,its AUC,sensitivity,specificity and accuracy respectively is ： 0. 81,81. 2%,70. 1%,71. 6%; when using Entropy and Inhomogenity to distinguish fp-RAML and ccRCC,the AUC,sensitivity,specificity and accuracy respectively is ： 0. 83,75. 0%,81. 0%,80. 0%; when using StDev,Entropy and Inhomogenity to distinguish fp-RAML and ccRCC,its AUC,sensitivity,specificity and accuracy respectively is ： 0. 84,87. 5%,69. 6%,72. 6%. Conclusion Some of the quantitative parameters of CT texture analysis（ StDev,Entropy,Inhomogenity） can be used to difrerentiate fp-AML from ccRCC.
出处
《临床放射学杂志》
CSCD
北大核心
2017年第7期993-997,共5页
Journal of Clinical Radiology
关键词
肾错构瘤
肾透明细胞癌
体层摄影术
X线计算机
纹理分析
Renal angiomyolipoma
Clear cell Renal Cell Carcinoma
Tomography
X-ray Computed
Texture Analysis
