能谱CT平扫图像直方图纹理分析对肾盂旁囊肿与肾盂积水的鉴别价值  

Value of histogram texture analysis of energy spectrum CT to distinguish parapelvic cysts from hydronephrosis

作　　者：陶健 马海彦 梁甜甜 曾宪春[1] 王荣品[1] TAO Jian;MA Haiyan;LIANG Tiantian;ZENG Xianchun;WANG Rongping(Department of Radiology of Guizhou Provincial People’s Hospital,International Exemplary Cooperation Base of Precision Imaging for Diagnosis and Treatment,Guiyang 550002,China)

机构地区：[1]贵州省人民医院放射科/精准影像诊疗示范型国际科技合作基地,贵州贵阳550002

出　　处：《中国中西医结合影像学杂志》2024年第2期202-206,共5页Chinese Imaging Journal of Integrated Traditional and Western Medicine

基　　金：国家自然科学基金项目(82160349)。

摘　　要：目的:探讨能谱CT平扫单能量图像结合直方图纹理分析对肾盂旁囊肿与肾盂积水的鉴别价值。方法:回顾性收集25例肾盂旁囊肿(肾盂旁囊肿组)和18例肾盂积水患者(肾盂积水组)。应用能谱成像分析平台测量2组40~140 keV(以10 keV为间隔)11个不同单能量CT值,并分别行统计学比较。单能量CT值差异有统计学意义的,在AW4.5工作站上重建相应单能量图像,采用GEOmni-Kinetics软件进行直方图纹理分析,选择病灶最大层面勾画ROI,生成基于灰度直方图纹理参数,包括最小值,最大值,均值,标准差,变异度,偏度,峰度,均一性,能量,熵,第5、10、25、50、75、90、95百分位数。采用单因素分析比较2组间参数的差异,基于ROC曲线评价各参数的鉴别价值。采用多因素logistic回归模型进行特征筛选并构建最终的预测模型,绘制最终模型的ROC曲线并分析其价值。结果:40~50 keV条件下,肾盂旁囊肿组的单能量CT值大于肾盂积水组,差异均有统计学意义(均P<0.05),2组60~140 keV单能量CT值差异均无统计学意义(均P>0.05)。40 keV条件下,2组均值,峰度,能量,熵,第5、10、25、50、75百分位数差异均有统计学意义(均P<0.05),余参数差异均无统计学意义(均P>0.05);50 keV条件下,2组均值,第10、25、50百分位数差异均有统计学意义(均P<0.05),余参数差异均无统计学意义(均P>0.05)。ROC曲线分析显示,40 keV图像纹理参数鉴别肾盂旁囊肿与肾盂积水的AUC为0.696~0.756,均大于50 keV图像纹理参数的AUC。多因素logistic回归分析显示,40 keV图像纹理参数的峰度、能量和第25百分位数是肾盂旁囊肿与肾盂积水的独立预测因子,三者联合诊断鉴别肾盂旁囊肿与肾盂积水的AUC为0.915,敏感度为93.3%、特异度为83.3%。结论:基于能谱CT平扫40 keV图像的直方图纹理参数构建的多因素logistic回归模型能有效鉴别肾盂旁囊肿与肾盂积水。Objective:To explore the value of energy spectrum CT monochromatic images combined with histogram texture analysis for identifying parapelvic cysts and hydronephrosis.Methods:Retrospectively collected data were obtained from 25 patients with parapelvic cysts and 18 patients with hydronephrosis.The spectral imaging analysis platform was applied to assess monoenergy CT values from 40 to 140 keV at 10 keV increments,and the monoenergy CT values of the two groups were compared.If there were significant differences in monoenergy CT values,the corresponding monoenergy images were reconstructed on AW 4.5 workstation.GE Omni-Kinetics software was used to conduct histogram texture analysis,the largest layer of lesion was selected to outline the ROI,and texture parameters were generated based on grayscale histograms,including the minimum value,maximum value,mean value,standard deviation,variability,skewness,kurtosis,uniformity,energy,entropy,and the 5th,10th,25th,50th,75th,90th,95th percentiles.Single factor analysis was used to compare the differences in each parameter between the two groups.ROC curve was used to analyze the diagnostic value of each parameters.A multivariate logistic regression model was used to screen features and construct the final prediction model,and ROC curve of the final model was drawn,and its value was analyzed.Results:At 40~50 keV,the monoenergy CT values of parapelvic cysts were greater than those of hydronephrosis,the differences were statistically significant(both P<0.05).There were no significant differences in the monoenergy CT values between the two groups at 60~140 keV.At 40 keV,the mean value,kurtosis,energy,entropy and the 5th,10th,25th,50th,75th percentiles were significantly different between the two groups(all P<0.05),while the other parameters were not significantly different(all P>0.05).At 50 keV,the mean value and the 10th,25th,50th percentiles were significantly different between the two groups(all P<0.05),while the other parameters had no significant differences(all P>0.05).ROC

关 键 词：肾盂旁囊肿 肾盂积水 体层摄影术 X线计算机 纹理分析 诊断 鉴别 

分 类 号：R692.7[医药卫生—泌尿科学] R816.7[医药卫生—外科学]