CT检查纹理分析对胃肠间质瘤c⁃KIT Exon 11基因突变的预测价值

Predictive value of CT texture analysis for c⁃KIT Exon 11 gene mutation in gastrointestinal stromal tumors

目的探讨CT检查纹理分析对胃肠间质瘤c⁃KIT Exon 11基因突变的预测价值。方法采用回顾性描述性研究方法。收集2012年1月至2020年5月国内3家医疗中心收治的285例(湖州市中心医院93例、安徽省阜阳市人民医院92例、上海市光华中西医结合医院100例)GIST患者的临床病理资料;男130例,女155例;中位年龄为62岁,年龄范围为25~82岁。285例患者中,190例c⁃KIT Exon 11基因突变,设为突变组;95例c⁃KIT Exon 11基因未突变,设为未突变组。患者术前均行CT检查。观察指标:(1)两组患者临床病理特征比较。(2)CT检查纹理参数观察者一致性检验。(3)两组患者CT检查纹理参数比较。(4)临床病理参数与CT检查纹理参数的诊断效能。采用ICC分析检验纹理参数观察者的一致性(ICC>0.80为观察者一致性良好),取ICC>0.80纹理参数的平均值,用于进一步分析。采用Shapiro Wilk检验计量资料的正态性,正态分布的计量资料以x±s表示,组间比较采用独立样本t检验;偏态分布的计量资料以M(P25,P75)表示,组间比较采用Mann⁃Whitney U检验。计数资料以绝对数表示,组间比较采用χ2检验。等级资料比较采用Kruskal⁃Wallis秩和检验。采用多因素Logistic回归(五折交叉验证)将差异有统计学意义的参数进行联合,采用受试者工作特征曲线(ROC)分析评估参数对胃肠间质瘤c⁃KIT Exon 11基因突变的预测效能,其中2012年1月至2018年1月数据用于内部测试,2018年2月至2020年5月数据用于外部验证。采用Delong检验比较曲线下面积(AUC)。结果(1)两组患者临床病理特征比较:突变组患者CD34表达水平(阴性、弱阳性、阳性、强阳性),CD117例表达水平(阴性、弱阳性、阳性、强阳性)分别为0、15、50、125例和5、15、55、115例;未突变组上述指标分别为5、55、5、30例和5、45、15、30例;两组患者上述指标比较,差异均有统计学意义(Z=11.53,12.32,P<0.05)。(2)CT检查纹理参数观察者一致性检验:CT检查平扫纹理参数的中位值、标准差、集群突出、平均值、偏度、峰度、一致性、能量值、熵值、惯性、逆差距、Haralick的ICC值分别为0.77、0.79、0.75、0.83、0.89、0.85、0.83、0.81、0.85、0.87、0.82、0.84;动脉期分别为0.79、0.71、0.74、0.83、0.81、0.86、0.87、0.80、0.84、0.82、0.83、0.81;静脉期分别为0.74、0.72、0.70、0.88、0.81、0.84、0.89、0.82、0.83、0.84、0.86、0.88。(3)两组患者CT检查纹理参数比较:突变组患者CT检查平扫纹理参数偏度、逆差距,CT检查动脉期纹理参数平均值、能量值,CT检查静脉期纹理参数偏度、一致性、能量值、逆差距、Haralick分别为0.78±0.38、4.22(3.18,5.38),5.51±1.87、0.013(0.007,0.020),0.66±0.20、4.40±1.41、0.004(0.001,0.007)、0.23(0.15,0.28)、0.24±0.10;未突变组患者上述指标分别为0.59±0.25、5.16(4.35,6.10),3.96±1.04、0.020(0.014,0.027),0.45±0.12、6.22±2.09、0.005(0.003,0.008)、0.25(0.21,0.28)、0.18±0.06;两组患者上述指标比较,差异均有统计学意义(t=-2.17,Z=-6.87,t=-4.03,Z=-11.43,t=-2.21、-3.92,Z=-2.06、-4.18,t=-3.02,P<0.05)。(4)临床病理参数与CT检查纹理参数的诊断效能。内部测试数据集ROC分析结果显示:临床病理诊断参数中CD34及CD117表达水平的AUC分别为0.75、0.71;临床病理参数联合AUC为0.75;单项CT检查纹理参数AUC为0.54~0.67;CT检查纹理参数联合AUC为0.79;临床病理参数及CT检查纹理参数联合AUC为0.89。外部验证数据集ROC分析结果显示:临床病理诊断参数中CD34及CD117表达水平的AUC分别为0.77、0.72;临床病理参数联合AUC为0.78;单项CT检查纹理参数AUC为0.51~0.73;CT检查纹理参数联合AUC为0.80;临床病理参数及CT检查纹理参数联合AUC为0.88。Delong检验结果显示:临床病理参数及CT检查纹理参数联合AUC高于临床病理参数联合,差异有统计学意义(Z=2.13,P<0.05)。临床病理参数及CT检查纹理参数联合AUC与CT检查纹理参数联合比较,差异无统计学意义(Z=1.66,P>0.05)。结论CT检查纹理分析可无创预测胃肠间质瘤c⁃KIT Exon 11基因突变状态,且将临床病理参数与CT检查纹理参数联合诊断的预测效能更优。

Objective To investigate the predictive value of computed tomography(CT)texture analysis for c⁃KIT Exon 11 gene mutation in gastrointestinal stromal tumors(GIST).Methods The retrospective and descriptive study was conducted.The clinicopathological data of 285 GIST patients who were admitted to three domestic medical centers(93 cases in Huzhou Central Hospital,92 cases in Fu Yang People′s Hospital and 100 cases in Shanghai GuangHua Hospital of Integrated Traditional Chinese and Western Medicine)from January 2012 to May 2020 were collected.There were 130 males and 155 females,aged from 25 to 82 years,with a median age of 62 years.Of the 285 patients,190 cases with mutated c⁃KIT Exon 11 gene and 95 with non⁃mutated c⁃KIT Exon 11 gene were allocated into mutation group and non⁃mutation group.All patients underwent preoperative CT examination.Observation indicators:(1)comparison of clinicopathological features between groups;(2)consistency check of CT texture parameters between observers;(3)comparison of CT texture parameters between groups;(4)diagnostic efficiency of clinicopathological parameters and CT texture parameters.The intraclass correlation coefficient(ICC)analysis was used for consistency check of CT texture parameters between observers,and ICC>0.80 was defined as good consistency between observers.The means of CT texture parameters with ICC>0.80 were used for further analysis.The Shapiro Wilk was used for normality test of measurement data.Measurement data with normal distribution were represented as Mean±SD,and comparison between groups was analyzed using the independent sample t test.Measurement data with skewed distribution were represented as M(P25,P75),and comparison between groups was analyzed using the Mann⁃Whitney U test.Count data were described as absolute numbers,and comparison between groups was analyzed using the chi⁃square test.Comparison of ordinal data was analyzed using the Kruskal⁃Wallis rank sum test.The multiunivariate Logistic regression was used to combine parameters with significant differences,and the receiver operating characteristic curve was used to evaluate the predictive value of parameters for c⁃KIT Exon 11 gene mutation in GIST.Data from January 2012 to January 2018 were used for internal test,and data from February 2018 to May 2020 were used for external validation.The Delong test was used to compare areas under curve.Results(1)Comparison of clinico⁃pathological features between groups:cases negative,weakly positive,positive,strongly positive for CD34 expression and cases negative,weakly positive,positive,strongly positive for CD117 expression were 0,15,50,125 and 5,15,55,115 for the mutation group,the above indicators were 5,55,5,30 and 5,45,15,30 for the non⁃mutation group,respectively.There were significant differences in the above indicators between the two groups(Z=11.53,12.32,P<0.05).(2)Consistency check of CT texture parameters between observers:in plain CT,the ICC of texture parameters median value,standard deviation,cluster prominence,mean value,skewness,kurtosis,uniformity,energy,entropy,Inertia,inverse difference moment and Haralick were 0.77,0.79,0.75,0.83,0.89,0.85,0.83,0.81,0.85,0.87,0.82,0.84;in arterial phase,the above indicators were 0.79,0.71,0.74,0.83,0.81,0.86,0.87,0.80,0.84,0.82,0.83,0.81;in venous phase,the above indicators were 0.74,0.72,0.70,0.88,0.81,0.84,0.89,0.82,0.83,0.84,0.86,0.88.(3)Comparison of CT texture parameters between groups:skewness and inverse difference moment in plain CT,mean value and energy in arterial phase,skewness,uniformity,energy,inverse difference moment and Haralick in venous phase of the mutation group were 0.78±0.38,4.22(3.18,5.38),5.51±1.87,0.013(0.007,0.020),0.66±0.20,4.40±1.41,0.004(0.001,0.007),0.23(0.15,0.28),0.24±0.10,respectively,versus 0.59±0.25,5.16(4.35,6.10),3.96±1.04,0.020(0.014,0.027),0.45±0.12,6.22±2.09,0.005(0.003,0.008),0.25(0.21,0.28),0.18±0.06 of the non⁃mutation group.There were significant differences in the above indicators between the two groups(t=-2.17,Z=-6.87,t=-4.03,Z=-11.43,t=-2.21,-3.92,Z=-2.06,-4.18,t=-3.02,P<0.05).(4)Diagnostic efficiency of clinicopathological parameters and CT texture parameters:ROC analysis of internal test data showed that CD34 and CD117 had AUC of 0.75 and 0.71,combined clinicopathological parameters had AUC of 0.75,single CT texture parameters had AUC of 0.54-0.67,combined CT texture parameters had AUC of 0.79,clinicopathological parameters combined with CT texture parameters had AUC of 0.89.ROC analysis of external test data showed that CD34 and CD117 had AUC of 0.77 and 0.72,combined clinicopathological parameters had AUC of 0.78,single CT texture examination had AUC of 0.51-0.73,combined CT texture parameters had AUC of 0.80,clinicopathological parameters combined with CT texture parameters had AUC of 0.88.Delong test showed that the combination of clinicopathological parameters and CT texture parameters had a higher AUC than combination of clinicopathological parameters(Z=2.13,P<0.05).The combination of clinicopathological parameters and CT texture parameters had no significant difference in the AUC compared with the combination of CT texture parameters(Z=1.66,P>0.05).Conclusion CT texture analysis can early and non⁃invasively predict the mutation status of c⁃KIT Exon 11 gene in GIST,the combination of clinicopathologic parameters and CT texture parameters has better predictive performance.