基于增强CT静脉期纹理分析与胃间质瘤危险度分级及ki-67表达水平的相关性研究

The Value of Vein Phase Image Features and Texture Parameters Based on Enhanced CT in Predicting the Expression Level of Ki-67 in Gastric Stromal Tumors

目的 探讨基于增强CT静脉期纹理参数在预测胃间质瘤(gastric stromal tumor,GST)危险度分级及ki-67表达水平中的价值。方法 回顾性分析潍坊市人民医院经病理确诊的80例GST患者的影像学资料,依据美国国立卫生研究院(National Institutes of Health,NIH)的分级标准,将患者分别分为极低、低危组(极低危及低危组)和中、高危组(中危及高危组)以及Ki-67低表达组和高表达组。利用Image J软件测量并记录纹理分析参数,利用两独立样本t检验或Mann-Whitney U检验对各参数进行统计分析,利用受试者工作特征曲线(ROC)来评估各参数鉴别GST危险度分级及ki-67表达水平的效能。结果 极低、低危组与中、高危组在均值、标准差、相关性、逆差矩及熵之间的差异具有统计学意义(P <0.05),ki-67低表达组及高表达组在均值及熵之间的差异具有统计学意义(P<0.05)。其中,熵在鉴别GST危险度分级及ki-67表达水平的参数中,诊断效能均为最大,其AUC值分别为0.85、0.72,敏感性分别为96.97%、80.00%,特异性分别为59.57%、60.00%。结论 增强CT静脉期纹理参数有助于GST的良恶性分级及ki-67表达水平的预测,可为临床提供更加精准化的治疗策略及评估预后。

Objective To explore the value of vein phase texture analysis parameters based on enhanced CT in predicting the risk grade and Ki-67 expression level of gastric stromal tumors. Methods CT imaging data of 80 GST patients diagnosed by pathology in Weifang people’s hospital were analyzed retrospectively,the patients were divided into very low-risk and low-risk group and middle-risk and high-risk group,And Ki-67 low expression group and high expression group.Image J software is used to measure and record texture analysis parameters, including histogram parameters: mean, standard deviation, skewness and kurtosis;Gray level co-occurrence matrix parameters: energy, contrast, correlation, inverse moment and entropy. Two independent sample t-test or Mann Whitney U-test are used for statistical analysis of texture analysis parameters. The receiver operating characteristic curve(ROC) is used to evaluate the effectiveness of each parameter in identifying GST risk classification and Ki-67 expression level. Results There were statistically significant differences in mean, standard deviation, correlation, deficit moment and entropy between very low-risk and low-risk group and middle-risk and high-risk group(P<0.05),The difference in mean and entropy between the low expression group and the high expression group of Ki-67 was statistically significant(P<0.05).Among them, entropy is the most effective parameter in differentiating GST risk grade and Ki-67 expression level, with AUC values of 0.85 and 0.69, sensitivity of 96.97% and 80.00%, and specificity of 59.57% and 57.14%, respectively. Conclusion Enhanced CT venous phase texture analysis parameters are helpful to the classification of GST and the prediction of Ki-67 expression level, and can provide more precise treatment strategies and evaluate prognosis for the clinic.