基于MRI影像组学预测乳腺癌新辅助疗效的价值  

The value of predicting the efficacy of neoadjuvant therapy in breast cancer based on MRI radiomics

作　　者：谢伶俐 陈基明[1] 张成孟 张爱娟 吴莉莉 李周丽 邵颖 XIE Lingli;CHEN Jiming;ZHANG Chengmeng;ZHANG Aijuan;WU Lili;LI Zhouli;SHAO Ying(Department of Radiology,Yijishan Hospital of Wannan Medical College,Wuhu 241001,Anhui,China)

机构地区：[1]皖南医学院弋矶山医院放射科,安徽芜湖241001

出　　处：《右江民族医学院学报》2024年第5期714-722,共9页Journal of Youjiang Medical University for Nationalities

摘　　要：目的探讨基于MRI影像组学预测乳腺癌新辅助治疗(neoadjuvant therapy,NAT)疗效的价值。方法收集皖南医学院弋矶山医院124例乳腺癌患者影像及临床资料,其中39例患者NAT后病理完全缓解(pathological complete response,pCR),85例患者NAT后病理非完全缓解(non-pathological complete response,non-pCR)。将患者随机分为训练组(88例)和验证组(36例),手动勾画感兴趣区(ROI),用AK软件提取MRI纹理特征,对纹理特征使用最小冗余最大相关(mRMR)和最小绝对收缩和选择算子(LASSO)回归方法进行降维、筛选,建立影像组学模型;采用多因素Logistic回归分析建立包括临床资料、影像组学的个性化预测模型。使用受试者工作特征(ROC)曲线和决策曲线分析(DCA)评价模型的预测效能和临床净收益。结果临床模型预测NAT疗效的曲线下面积(AUC)在训练组和试验组中分别为0.824、0.743,多参数MRI(mpMRI)影像组学模型的AUC分别为0.848、0.752,个性化预测模型的AUC分别为0.911、0.865。DCA显示个性化预测模型的临床净收益优于临床模型和多参数MRI影像组学模型。结论个性化预测模型对NAT疗效的预测具有较高的诊断效能,明显好于临床模型,对早期预测NAT疗效具有一定的临床应用价值。Objective To explore the value of MRI-based radiomics in predicting the efficacy of neoadjuvant therapy(NAT)for breast cancer.Methods The imaging and clinical data from 124 patients with breast cancer at Yijishan Hospital of Wannan Medical College were collected,including 39 patients with pathological complete response(pCR)after NAT and 85 patients with non-pathological complete response(non-pCR)after NAT.The patients were randomly divided into a training group(88 cases)and a validation group(36 cases),regions of interest(ROI)were manually outlined,and MRI texture features were extracted using AK software,Minimal Redundancy Maximal Relevance(mRMR)and Least Absolute Shrinkage and Selection Operator(LASSO)regression methods were applied to reduce dimensionality and screen texture features to establish a radiomics model.A personalized predictive model incorporating clinical data and radiomics was developed using multivariate Logistic regression analysis.Receiver Operating Characteristic(ROC)curves and Decision Curve Analysis(DCA)were used to evaluate the predictive performance and clinical net benefit of the model.Results The Area Under the Curve(AUC)of the clinical model for predicting NAT efficacy was 0.824 in the training group and 0.743 in the validation group.The AUC of the multi-parameter MRI(mpMRI)radiomics model were 0.848 and 0.752,respectively.The AUC of the personalized predictive model were 0.911 and 0.865,respectively.DCA demonstrated that the clinical net benefit of the personalized predictive model was superior to that of the clinical model and the mpMRI radiomics model.Conclusion The personalized prediction model has a high diagnostic efficiency in predicting the curative effect of NAT,which is obviously better than the clinical model,and has a certain clinical application value for early prediction of the curative effect of NAT.

关 键 词：乳腺肿瘤 新辅助治疗 磁共振成像 影像组学 预测模型 

分 类 号：R737.9[医药卫生—肿瘤] R445.2[医药卫生—临床医学]