基于MRI影像组学构建PD-1/PD-L1抑制剂治疗dMMR/MSI-H直肠癌疗效的预测模型  

Construction of a Prediction Model of dMMR/MSI-H Rectal Cancer with PD-1/PD-L1 Inhibitor Immunotherapy Based on MRI Radiomics

作　　者：张岚[1] 周彦汝 韩鼎盛 张嘉诚 何旭 刘鹏[2] ZHANG Lan;ZHOU Yanru;HAN Dingsheng;ZHANG Jiacheng;HE Xu;LIU Peng(MRI Department,First Affiliated Hospital of Henan University of TCM,Zhengzhou Key Laboratory of Intelligent Analysis and Utilization of Chinese Medicine Information,Henan University of Chinese Medicine,Zhengzhou 450000,China;Hunan Provincial People's Hospital(The First Affiliated Hospital of Hunan Normal University))

机构地区：[1]河南中医药大学第一附属医院MRI科,河南中医药大学中医药信息智能分析与利用郑州市重点实验室,河南省郑州市450000 [2]湖南省人民医院(湖南师范大学附属第一医院)

出　　处：《中国医学计算机成像杂志》2024年第3期343-348,共6页Chinese Computed Medical Imaging

基　　金：河南省自然科学基金面上项目(232300421187);河南省医学科技攻关计划项目(LHGJ20230679);长沙市自然科学基金(kq2014201);2021SKY影像科研基金(Z-2014-07-2101)。

摘　　要：目的:探讨MRI影像组学模型在程序性细胞死亡蛋白-1(PD-1)/程序性细胞死亡-配体1(PD-L1)抑制剂联合全程新辅助治疗(TNT)局部进展期直肠癌(LARC)的疗效预测价值。方法:收集河南中医药大学第一附属医院PD-1/PD-L1抑制剂联合TNT治疗的80例错配修复基因缺陷(dMMR)/微卫星高度不稳定(MSI-H)基因型中低位LARC患者的临床和影像资料。将入组患者按7∶3比例分为训练集和测试集,提取影像组学特征,从中筛选并构建影像组学模型。描绘影像组学模型的Rad-score与病理金标准之间的受试者工作特征(ROC)曲线,计算曲线下面积(AUC),并评价模型的诊断效能。采用决策曲线分析(DCA)计算风险阈值的范围,并评估临床获益情况。收集湖南省人民医院25例dMMR/MSI-H基因型LARC患者的影像资料作为外部验证集。结果:训练集、测试集及外部验证集三者之间的临床特征无统计学差异(P>0.05)。经过降维处理、t检验及一致性检验以及LASSO交叉验证后,筛选出一阶偏度特征和体积2个特征构建影像组学模型。训练集、测试集和外部验证集的影像组学预测模型ROC曲线的AUC、灵敏度、特异度、阳性预测值和阴性预测值分别为0.920、97.1%、85.7%、91.9%、94.7%;0.885、80.0%、88.9%、92.3%、72.7%;0.875、87.5%、88.9%、93.3%、80.0%。DCA曲线显示,当风险阈值范围为0%~82%时,采用影像组学模型预测LARC患者为病理完全缓解(pCR)的获益大于将所有患者都视为pCR或者无病理完全缓解(npCR)。结论:基于MRI影像组学构建的dMMR/MSI-H型局部进展期直肠癌PD-1/PD-L1抑制剂联合全程新辅助放化疗疗效预测模型,有较大潜力为不同基因分型的直肠癌患者制定个体化治疗策略提供量化依据。Purpose:To investigate the predictive value of MRI radiomics model for the efficacy in locally advanced rectal cancer(LARC)treated with programmed cell death protein 1(PD-1)/programmed cell death-ligand 1(PD-L1)inhibitor and total neoadjuvant therapy(TNT).Methods:The data of 80 LARC patients with deficient mismatch repair gene(dMMR)/microsatellite instability-high(MSI-H)who received PD-1/PD-L1 inhibitor combined with TNT in The First Affiliated Hospital of Henan University of TCM were collected.The enrolled patients were divided into a training set and a test set at a ratio of 7∶3.The features were extracted and screened,and the radiomics model was constructed.Receiver operating characteristic(ROC)curve between Rad-score of radiomics model and the pathological result was depicted and area under curve(AUC)was calculated to evaluate the diagnostic performance.Decision curve analysis(DCA)was used to calculate the range of risk thresholds and assess the clinical benefit.The imaging data of 25 dMMR/MSI-H LARC patients from Hunan Provincial People's Hospital were collected as the external validation set.Results:There were no significant differences in the clinical characteristics among training set,test set and validation set(all P>0.05).After dimensionality reduction,t test,consistency test,and LASSO cross validation,the first-order skewness feature and volume feature were screened to construct a radiomics prediction model.The AUC,sensitivity,specificity,positive predictive value and negative predictive value of radiomics model in the training test,test set and external validation set were 0.920,97.1%,85.7%,91.9%,94.7%;0.885,80.0%,88.9%,92.3%,72.7%;0.875,87.5%,88.9%,93.3%,80.0%,respectively.The DCA curve showed that the strategy of radiomics model in predicting pathologic complete response(pCR)was better than in treating all the patients as pCR or as no pathologic complete response(npCR)when the risk threshold was 0%-82%.Conclusion:The prediction model based on MRI radiomics of dMMR/MSI-H LARC patients treated by PD

关 键 词：磁共振成像 影像组学 直肠肿瘤 局部进展期 程序性细胞死亡蛋白-1/程序性细胞死亡-配体1 全程新辅助放化疗 

分 类 号：R735.37[医药卫生—肿瘤] R445.2[医药卫生—临床医学]