CT影像组学预测肝内胆管细胞癌术后生存状态的价值

Value of CT radiomics in predicting the postoperative survival of intrahepatic cholangiocarcinoma

目的:探讨CT影像组学在肝内胆管细胞癌(ICC)患者术后生存状态预测中的价值。方法:回顾性分析经术后病理证实的127例ICC患者的临床与影像资料。127例患者分为训练集(68例)、测试集(28例)及外部验证集(31例)。从训练集CT图像中提取特征,建立影像组学模型。采用多因素Cox回归分析模型构建临床模型。将影像组学模型的影像组学评分和临床模型的预测因子用于构建联合模型,并绘制列线图及校准曲线。以C指数评估3种模型的预测性能。绘制3种模型的决策曲线。基于训练集的影像组学评分的截断值,将患者分为ICC高危组与低危组。对高危组和低危组患者在测试集和外部验证集中的生存时间及其差异的分析采用Kaplan-Meier生存分析和对数秩检验。结果:K-M曲线显示高危组和低危组3年总体生存期差异有统计学意义。联合模型校准曲线显示,1年总生存期与2、3年总生存期相比,预测与实际生存可能性更趋于一致。联合模型在训练集、测试集和外部验证集中的C指数分别为0.912、0.696、0.798,预测效能最高;决策曲线分析显示联合模型鉴别效能最佳。结论:利用影像组学评分结合临床危险因素构建定量化模型能够预测ICC患者的术后生存状态,可为临床决策提供新方法。

Objective:To analyze the value of CT radiomics in the postoperative survival of intrahepatic cholangiocarcinoma(ICC).Methods:A retrospective analysis of the clinical and imaging data of 127 ICC patients confirmed by surgical pathology was performed.127 patients were divided into a training set(68 cases),a validation set(28 cases),and an external validation set(31 cases).The radiomics signatures were extracted from the training set,which were used to build the radiomics model.The clinical predictors were developed using a multivariate Cox regression model.The combined model incorporated the radiomics scores and clinical predictors,and a nomogram and a calibration curve were drawn.C index was used to evaluate the predictive effectiveness of the three models,and the decision curve of 3 models was drawn.Based on the cut-off value of the radiomics score of the training set,68 patients were divided into a high-risk group and a low-risk group.The survival time and its difference between the high-risk group and the low-risk group in the validation set and the external validation set were determined by Kaplan-Meier analysis and log-rank test.Results:K-M curve showed the 3-year overall survival between the high-risk group and the low-risk group was significantly different.And the calibration curve of the combined model showed that 1-year overall survival was more consistent with the actual survival probability than with the 2-and 3-year overall survival.The C-index of the combined model in the training set,the validation set and the external validation set was 0.912,0.696 and 0.798,respectively,the combined model had the highest effienciency.The combined model also had the best performance in decision curve analysis.Conclusion:Radiomics can predict postoperative survival for ICC,which can provide an alternative approach for clinical decision-making.