【摘要】 目的探究基于高分辨MRI形态学特征及原发肿瘤影像组学所建立的联合模型预测T_1～2期直肠癌淋巴结转移的价值。方法回顾性分析119例术前经3.0 T MRI检查且手术病理证实为T_1～2期直肠癌患者的临床及影像资料，所有病例按7∶3的比例随机分为训练集和测试集。通过单因素和多因素Logistic回归分析建立MRI形态学特征模型、影像组学模型以及MRI形态学?影像组学联合模型。采用受试者工作特征（ROC）曲线下面积（AUC）比较三组预测模型的效能。采用自助重抽样法（Bootstrap）重复抽样1000次进行内部验证，从而评估模型的稳定性。最后，选取最优模型创建用于临床应用的列线图，以校准曲线分析评估模型的校准度，采用决策曲线分析（DCA）评估模型的临床价值。结果联合模型具有最高的预测效能，在训练集（AUC=0.888,95%CI:0.824～0.943）和测试集（AUC=0.862,95%CI:0.723～0.972）中的AUC值均最高。联合模型在重抽样内部验证集中的AUC值为（AUC=0.874,95%CI:0.751～0.996）。DCA表明，联合模型的净效益在训练集和测试集中均比其他两种模型更高。结论高分辨MRI形态学?影像组学联合模型在预测T_1～2期直肠癌淋巴结转移方面具有较高的诊断效能，可为个性化临床治疗提供帮助。更多还原

【Abstract】 Objective To evaluate the diagnostic value of a combined model based on high?resolution MRI morpho?logical features and primary tumor radiomics in predicting lymph node metastasis in stage T_1～2 rectal cancer.Methods The clinical and imaging data of 119 patients with stage T_1～2 rectal cancer confirmed by surgical pathology who underwent preoperative 3.0 T MRI examination were retrospectively selected,and all the cases were randomly divided into a training set and a test set in a ratio of 7∶3.Univariate and multivariate Logistic regression analysis were used to establish MRI mor?phological feature model,radiomics model and MRI morphology?radiomics combined model.The performance of the three models was assessed using area under the curve （AUC） of receiver operating characteristic （ROC） curve.The bootstrap resampling method was employed to perform 1000 iterations of internal validation through repeated sampling,thereby as?sessing the stability of the model.Finally,the optimal model was selected to create a radiomics nomogram for clinical appli?cation.The calibration degree of the selected model was evaluated by calibration curve analysis and the clinical value of the model was evaluated by decision curve analysis （DCA）.Results The combined model demonstrated significantly better prediction performance with the highest AUC value in both the training set （AUC=0.888,95%CI:0.824-0.943） and the test set （AUC=0.862,95%CI:0.723-0.972）.The AUC value of the internal validation set of resampling for the combined model was （AUC=0.874,95%CI:0.751-0.996）.DCA demonstrated that the combined model exhibits higher net benefits than the other two models in both the training set and test set.Conclusion The established high?resolution MRI morphology?radiomics combined model in this study has high diagnostic efficacy in predicting lymph node metastasis in stage T1～2 rectal cancer,which would be helpful for individualized clinical treatment.更多还原