基于T_(2)WI和扩散加权成像的磁共振影像组学预测直肠癌的壁外血管侵犯  

T_(2)WI and DWI-based radiomics model for preoperative prediction of extramural venous invasion of rectal adenocarcinoma

作　　者：游其奎 林祺[1] 邱少华 李齐英 You Qikui;Lin Qi;Qiu Shaohua;Li Qiying(Department of Radiology,Longyan First Affiliated Hospital of Fujian Medical University,Fujian 364000,China)

机构地区：[1]福建医科大学附属龙岩市第一医院放射科,364000

出　　处：《实用医学影像杂志》2025年第2期85-90,共6页Journal of Practical Medical Imaging

基　　金：福建省自然科学基金项目(2020J0112)。

摘　　要：目的建立基于T_(2)WI和扩散加权成像(DWI)的磁共振影像组学模型预测直肠癌的壁外血管侵犯(EMVI)并验证。方法本研究纳入了142例直肠癌患者(85例EMVI阳性,57例EMVI阴性),99例为训练集,43例为验证集。从术前的DWI和T_(2)WI图像中分别提取了398个影像组学特征,总共796个。采用最大相关最小冗余(mRMR)和最小绝对收缩与选择算子(LASSO)选择最终的影像组学特征。建立影像组学评分(Rad-score)模型以评估直肠癌EMVI的状态。采用受试者工作特征曲线(ROC)、准确度、灵敏度、特异度、阳性预测值(PPV)和阴性预测值(NPV)评估模型性预测性能。结果最终我们选择了的23个影像组学特征来构建Rad-score模型。该模型在训练集和验证集中均表现良好,训练集和验证集的曲线下面积(AUC)95%CI分别为0.91(0.88,0.95)和0.72(0.55,0.89),训练集的准确度、灵敏度、特异度、阳性预测值(PPV)、阴性预测值(NPV)值分别为0.86、0.83、0.88、0.83、0.88和0.75,验证集分别为0.69、0.59、0.76、0.63、0.73。结论Rad-core模型在术前预测直肠癌EMVI的方面表现出良好的诊断性能。因此,它可以作为直肠癌EMVI预测的关键工具。Objective To establish and validate a magnetic resonance imaging(MRI)radiomics model to predict extramural venous invasion(EMVI)in rectal cancer.Methods A total of 142 patients with rectal cancer were included in this study(85 were EMVI positive and 57 were EMVI negative),99 patients were on the training set,and 43 patients were on the testing set.T_(2)-weighted imaging(T_(2)WI)and diffusion-weighted imaging(DWI)images were acquired from the patients before surgery.A total of 398 radiomics features were extracted from the DWI and T_(2)WI images,respectively,and 796 in total.The maximum relevance minimum redundancy(mRMR)and the least absolute shrinkage and selection operator(LASSO)were used to select the final radiomics features.The rad-score model was established to evaluate the EMVI status of rectal cancer.The predictive performance of the model was evaluated using the?receiver operating characteristic curve(ROC),accuracy,sensitivity,specificity,positive predictive value(PPV),and negative predictive value(NPV).Results A total of 23 radiomics features were selected to construct the Rad-score model.The model performed well in both the training set and testing set,with an area under the curve(AUC)95%confidence interval(95%CI)of 0.91(0.88,0.95)in the training set and 0.72(0.55,0.89)in the validation set.The accuracy,sensitivity,specificity,PPV,and NPV in the training set were 0.86,0.83,0.88,0.83,and 0.88,respectively.In the testing set,the results were 0.69,0.59,0.76,0.63,and 0.73.Conclusion The rad-score model demonstrates good diagnostic performance in predicting the EMVI of rectal cancer before surgery.Therefore,it can be used as a key tool for predicting EMVI in rectal cancer.

关 键 词：磁共振成像 直肠肿瘤 壁外血管侵犯 

分 类 号：R445.2[医药卫生—影像医学与核医学]