计算机断层增强扫描图像结合动脉期纹理分析对胃肠道间质瘤危险度分级的评估价值

Application value of computed tomography enhanced image combined with arterial phase texture analysis in the risk of gastrointestinal stromal tumor

目的探讨CT增强图像结合动脉期纹理分析对胃肠道间质瘤(GIST)危险度分级的评估价值。方法回顾性收集2016年1月至2019年7月兰州大学第一医院经病理和免疫组织化学检查确诊的81例原发性单发GIST患者的CT图像和临床资料。81例GIST患者中,极低危7例,低危18例,中危19例,高危37例。在图像存档与传输系统工作站中测得CT增强扫描动脉期、静脉期、延迟期肿瘤实质CT值,强化峰值为动脉期、静脉期、延迟期最大CT值。利用FireVoxel软件提取14组CT纹理分析参数,包括最小值、最大值、平均值、中位数、标准差、峰度、偏度、不均匀性、熵值,以及10%、25%、50%、75%、90%分位像素值。统计学方法采用单因素方差分析和Kruskal-WallisH检验,相关性采用Spearman相关分析,以ROC曲线和AUC值评估诊断效能,二分类logistic回归分析评估联合诊断效能。结果极低危、低危、中危、高危GIST动脉期CT值和强化峰值[分别为(51.57±12.16)、(61.89±20.95)、(60.32±21.33)、(82.35±17.36) HU和(71.71±6.57)、(74.83±15.34)、(77.37±18.06)、(85.38±16.58) HU]比较差异均有统计学意义(F=10.299、2.772,P=0.001、0.047),而不同危险度分级GIST静脉期、延迟期CT值比较差异均无统计学意义(F=0.757、1.082,P=0.522、0.362)。极低危、低危、中危、高危GIST增强动脉期CT纹理分析参数中最小值和熵值(分别为881.57±229.75、942.00±73.82、934.21±66.71、788.18±307.38和3.45±0.36、3.68±0.59、3.79±0.50、4.06±0.62)比较差异均有统计学意义(F=2.759、3.425,P=0.047、0.021),而其他参数比较差异均无统计学意义(P均>0.05)。ROC曲线分析结果显示,最小值、熵值、动脉期CT值、强化峰值鉴别极低危、低危、中危、高危GIST的AUC值(95%CI)分别为0.643(0.516~0.769)、0.701(0.579~0.823)、0.509(0.362~0.656)、0.661(0.530~0.792)。二元logistic回归分析显示,熵值和强化峰值联合诊断的AUC值(95%CI)为0.829(0.731~0.927),灵敏度为77%,特异度为80%。Spearman相关分析显示,熵值、强化峰值与GIST危险度分级均呈正相关(r=0.406,P<0.01;r=0.341,P=0.002)。结论 CT增强图像结合动脉期CT纹理分析可以作为GIST危险度分级鉴别诊断的定量分析方法,可以更好地为GIST临床治疗提供参考依据。

Objective To explore the evaluation value of computed tomography(CT)-enhanced image combined with arterial phase texture analysis in the risk classification of the gastrointestinal stromal tumor(GIST).Methods From January 2016 to July 2019,in The First Hospital of Lanzhou University,the CT images and clinical data of 81 patients diagnosed with primary single GIST confirmed by pathology and immunohistochemistry were retrospectively collected.Among 81 GIST patients,seven cases were extremely low-risk,18 cases were low-risk,19 cases were medium-risk,and 37 cases were high-risk.The CT values of arterial phase,venous phase and delayed phase were measured in the image archiving and transmission system.The peak value of enhancement was the maximum CT value in arterial phase,venous phase and delayed phase.Fourteen sets of CT texture analysis parameters,including minimum,maximum,mean,median,standard deviation,kurtosis,skewness,heterogeneity,entropy,and 10%,25%,50%,75%,90%quantile pixel values were extracted by FireVoxel software.One way analysis of variance and Kruskal-Wallis H tests were used for statistical analysis.Spearman correlation analysis was used for correlation analysis,receiver operating characteristic curve(ROC)and area under the curve(AUC)were used to evaluate the diagnostic efficacy and binary logistic regression analysis for joint diagnosis.Results The CT values in the arterial phase and the enhanced peak of extremely low-risk,low-risk,medium-risk and high-risk GIST were(51.57±12.16),(61.89±20.95),(60.32±21.33),(82.35±17.36)HU and(71.71±6.57),(74.83±15.34),(77.37±18.06),(85.38±16.58)HU,respectively,and the differences were statistically significant(F=10.299 and 2.772,P=0.001 and 0.047).However,there were no significant differences in CT values in the venous phase and delayed phase among GIST of different risk classification(F=0.757 and 1.082,P=0.522 and 0.362).The minimum value and the entropy of CT texture analysis parameters in the enhanced arterial phase of extremely low-risk,low-risk,medium-risk,and high-risk GIST were 881.57±229.75,942.00±73.82,934.21±66.71,788.18±307.38 and 3.45±0.36,3.68±0.59,3.79±0.50,4.06±0.62,and the differences were statistically significant(F=2.759 and 3.425,P=0.047 and 0.021).While there were no significant differences in other parameters(all P>0.05).The ROC analysis showed that the AUC values and 95%confidence interval(CI)of minimum,entropy,CT value in the arterial phase and enhanced peak value in the differential diagnosis of extremely low-risk,low-risk,medium-risk and high-risk GIST were 0.643(0.516 to 0.769),0.701(0.579 to 0.823),0.509(0.362 to 0.656),and 0.661(0.530 to 0.792),respectively.The binary logistic regression analysis showed the AUC value(95%CI)of the joint diagnosis of entropy and the enhanced peak value was 0.829(0.731 to 0.927);the sensitivity was 77%,and the specificity was 80%.Spearman correlation analysis showed that the entropy value,peak enhancement value and GIST risk were both positively correlated(r=0.406,P<0.01;r=0.341,P=0.002).Conclusions CT-enhanced image combined with arterial phase CT texture analysis can be used as a quantitative analysis method to identify the risk of GIST,which can provide a better reference for clinical treatment of GIST.