目的:探讨动脉增强后CT纹理指标熵(entropy,E)的变化(ΔE)是否提高肾乏脂肪血管平滑肌脂肪瘤(fat-poor renal angiomyolipoma,fpRAML)和肾透明细胞癌(clear cell renal cell carcinoma,ccRCC)的鉴别诊断价值.方法:纳入2019年1月—2021年12月在苏州大学附属第二医院经腹腔镜肾部分切除手术病理证实为fpRAML或ccRCC且行增强CT检查的患者205例,其中fpRAML组32例,ccRCC组173例.ImageJ软件提取肿瘤最大层面CT平扫和动脉期相同形状和大小感兴趣区(region of interest,ROI)内的每个像素CT值,计算出平扫时CT值(CTVnc)和纹理指标熵(Enc)、动脉增强后CT值(CTVa)和熵(Ea)以及动脉增强后CT值和熵的变化(ΔCTV和ΔE);比较这些指标在fpRAML组和ccRCC组的差异,用2组差异有统计学意义的变量分别建立不含ΔE的模型A和含ΔE的模型B,绘制受试者工作特征(receiver operator characteristic,ROC)曲线,获得曲线下面积(area under the curve,AUC),用AUC比较2个模型对fpRAML和ccRCC鉴别诊断价值,决策曲线分析、净重分类指数(net reclassification index,NRI)、综合判别改善指数(integrated discrimination improvement,IDI)等判断模型的优劣.结果:fpRAML组和ccRCC组的性别、CTVnc、Ea和ΔE比较均差异有统计学意义(均P＜0.05),鉴别fpRAML和ccRCC的AUC分别为0.71、0.70、0.84和0.90;以性别、CTVnc和Ea为变量建立的logistic模型A鉴别fpRAML和ccRCC的AUC为0.895,加入ΔE的logistic模型B的AUC增加到0.944(DeLong检验,P＜0.05),模型B的决策曲线优于模型A,NRI和IDI分别为0.137和0.076,显示出加入ΔE后模型B鉴别fpRAML和ccRCC有明显临床净收益.结论:动脉期增强CT纹理熵变化明显提高fpRAML和ccRCC的鉴别诊断价值.
Contrast-enhanced CT texture entropy change on arterial phase improves differential diagnostic value between fat-poor renal angiomyolipoma and clear cell renal cell carcinoma
Objective:To investigate if contrast-enhanced CT texture entropy change on arterial phase im-proves differential diagnostic value between fat-poor renal angiomyolipoma(fpRAML)and clear cell renal cell car-cinoma(ccRCC).Methods:Two hundred and five patients with pathologically confirmed ccRCC or fpRAML after laparoscopic partial nephrectomy and with contrast-enhanced CT performed in Second Affiliated Hospital of Soo-chow University were enrolled in this study,among which 173 patients were ccRCC and 32 patients were fpRAML.Pixel CT values were extracted from regions of interest of the same shape and size in non-contrast-en-hanced and arterial phase of contrast-enhanced CT on axial section with the largest transversal tumor diameter u-sing ImageJ software.The non-contrast CT value(CTVnc)and texture character entropy(Enc),arterial phase CT value(CTVa)and E(Ea)as well as the change of CTV and E(ΔCTV and ΔE)were calculated and recorded.These features were compared between fpRAML and ccRCC groups.Two sets of variables with statistically sig-nificant differences were employed to establish Model A without ΔE and Model B with ΔE,respectively.ROC curves were constructed.Differential diagnostic efficacy of both models was evaluated and compared with AUC,net re-classification index(NRI)and integrated discrimination improvement(1DI).Results:There were significant differences in sex,CTVnc,Ea and ΔE between fpRAML and ccRCC groups.The AUCs for sex,CTVnc,Ea and ΔE to differentiate fpRAML from ccRCC were 0.71,0.70,0.84 and 0.90 respectively.Logistic Model A with the variables of sex,CTVnc and Ea had lower AUC than Model B with ΔE added(0.895 vs 0.944,P＜0.05,DeLong test).Mode B showed better decision curve than Model A.Both NRI of 0.137 and ID1 of 0.076 demonstrated the positive clinical improvement of Model B with the ΔE variable to differentiate fpRAML from ccRCC.Conclusion:Contrast-enhanced CT texture entropy change on arterial phase significantly improves differential diagnostic value between fpRAML and ccRCC.
NETL
NSTL
万方数据
