目的 探讨多参数磁共振成像(multi-parametric magnetic resonance imaging,mp-MRI)的影像组学模型术前对子宫内膜癌(endometrial carcinoma,EC)微卫星不稳定(microsatellite instability,MSI)状态的预测价值.材料与方法 回顾性分析171名病理证实为EC患者的临床、病理及影像资料,以7∶3的比例随机划分为测试集与验证集,使用3D slicer软件对所有患者横轴位T2WI、扩散加权成像(diffusion weighted imaging,DWI)及矢状位对比增强T1WI(contrast-enhanced T1WI,CE-T1WI)延迟期序列进行感兴趣区(region of interest,ROI)的勾画并提取影像组学特征,采用组内相关系数(intra-class correlation coefficient,ICC)、最小绝对收缩和选择算子(least absolute shrinkage and selection operator,LASSO)算法和十折交叉验证及Person相关性检验对特征进行筛选并计算影像组学评分(radiomics score,Rad-score),使用Rad-score构建各单一序列模型(T2WI模型、DWI模型、CE-T1WI模型)及联合模型(Combine模型),绘制各模型受试者工作特征(receiver operating characteristic,ROC)曲线,使用曲线下面积(area under the curve,AUC)、敏感度、特异度等指标评价模型的效能,并在验证集中进行验证.DeLong检验比较各模型AUC值的差异.结果 最终从T2WI、DWI及CE-T1WI序列中分别保留了 6、3、3个特征.在测试集中T2WI模型、DWI模型、CE-T1WI模型和Combine模型的AUC值分别为0.869[95％置信区间(confidence interval,CD:0.772～0.938]、0.768(95％CI:0.645～0.865)、0.912(95％CI:0.830～0.966)、0.927(95％CI:0.865～0.966);在验证集中 T2WI 模型、DWI 模型、CE-T1W 模型和 Combine模型的 AUC 值分别为0.736(95％CI:0.573～0.896)、0.714(95％CI:0.560～0.872)、0.856(95％CI:0.675～0.990)、0.907(95％CI:0.813～0.977).DeLong检验显示DWI模型与Combine模型、CE-T1WI模型的AUC值差异有统计学意义(P＜0.05),余各模型两两之间的AUC值差异没有统计学意义(P＞0.05).结论 基于mp-MRI的影像组学模型可以在术前较好地预测EC的MSI表达状态,且与单一序列相比,联合模型的预测效能更高,这有助于临床制订个体化治疗方案,从而改善患者的预后.
Application research of imaging genomics in preoperative prediction of microsatellite stability of endometrial cancer using mp-MRI
Objective:To explore the predictive value of multi-parametric magnetic resonance imaging(mp-MRI)radiomics models for preoperative microsatellite instability(MSI)status in endometrial carcinoma(EC).Materials and Methods:A retrospective analysis was conducted on clinical,pathological,and imaging data of 171 patients with pathologically confirmed EC.The patients were randomly divided into a training set and a validation set in a 7∶3 ratio.Using the 3D Slicer software,regions of interest(ROIs)were delineated on axial T2WI,diffusion-weighted imaging(DWI),and sagittal contrast-enhanced T1WI(CE-T1WI)delayed phase sequences,and radiomic features were extracted.Feature selection and calculation of radiomics scores(Rad-scores)were performed using intra-class correlation coefficient(ICC),least absolute shrinkage and selection operator(LASSO)algorithm,ten-fold cross-validation,and Pearson correlation test.Models were constructed using Rad-scores for individual sequences(T2WI model,DWI model,CE-T1WI model)and a combined model.Receiver operating characteristic(ROC)curves were plotted for each model,and model performance was evaluated using area under the curve(AUC),sensitivity,specificity,and other metrics.The models were validated on the test set.The DeLong test was used to compare the differences in AUC values among the models.Results:Among the 171 EC patients,35 had MSI and 136 had microsatellite stability(MSS).From the T2WI,DWI,and CE-T1WI sequences,6,3,and 3 features were retained,respectively.In the training set,the area under the curve(AUC)values for the T2WI model,DWI model,CE-T1WI model,and combine model were 0.869[95％confidence interval,(CI):0.772-0.938],0.768(95％CI:0.645-0.865),0.912(95％CI:0.830-0.966),and 0.927(95％CI:0.865-0.966),respectively.In the validation set,the AUC values were 0.736(95％CI:0.573-0.896),0.714(95％CI:0.560-0.872),0.856(95％CI:0.675-0.990),and 0.907(95％CI:0.813-0.977)for the T2WI model,DWI model,CE-T1WI model,and combine model,respectively.The DeLong test indicated that there were statistically significant differences in AUC values between the DWI model and both the combine model and the CE-T1WI model(P＜0.05).No statistically significant differences were found between the AUC values of the other model pairs(P＞0.05).Conclusions:The radiomic model based on mp-MRI can effectively predict the MSI status of EC preoperatively.The combined model shows higher predictive performance compared to individual sequences,which helps in formulating personalized treatment plans and improving patient outcomes.
国家科技期刊平台
NSTL
万方数据
