摘要
目的 探讨基于动态对比增强磁共振成像(DCE-MRI)和扩散加权成像(DWI)影像组学特征术前预测乳腺癌人表皮生长因子受体-2(HER-2)低表达的临床应用价值.方法 回顾性搜集299例经本院病理证实为乳腺癌患者的首次MRI及临床病理资料,将患者分为HER-2低表达组(n=144)和HER-2非低表达组(n=155);按照8:2比例随机将其分为训练集(n=239)和测试集(n=60).采用ITK-SNAP软件手动逐层勾画DCE-MRI和DWI图像上病灶的感兴趣区(ROI),并提取影像组学特征.采用Mann-WhitneyU检验、Z分数归一化、方差阈值、K最佳、最小绝对收缩和选择算子(LASSO)筛选特征,并建立DCE-MRI、DWI及二者联合模型.应用受试者工作特征曲线(ROC)的曲线下面积(AUC)、敏感度、特异度、准确率评估模型的预测效能.结果 基于DCE-MRI、DWI、二者联合模型术前预测HER-2低表达的AUC值在训练集和测试集中分别为0.754、0.775、0.843和0.774、0.645、0.795.结论 基于DCE-MRI、DWI组学特征模型均可术前无创性预测乳腺癌HER-2低表达状态,且以二者联合模型预测效能最佳,可为临床乳腺癌治疗方案的选择提供参考.
Abstract
Objective The initial MRI and clinicopathological data of 299 patients with breast cancer confirmed by pa-thology in our hospital were retrospectively collected.The patients were divided into low-HER-2 expression group(n=144)and non-low-HER-2 expression group(n=155).They were randomly divided into training set(n=239)and testing set(n=60)according to the ratio of 8∶2.ITK-SNAP software was used to manually delineate the region of interest(ROI)of the lesions on DCE-MRI and DWI images,which were used to extract radiomics features.The methods of Mann-Whitney U test,Z-score normalization,variance threshold,K-best,least absolute shrinkage and selection operator(LASSO)were used to se-lect radiomics features.The models of DCE-MRI,DWI and DCE-MRI combined DWI were established.The area under the curve(AUC)of receiver operating characteristic(ROC),sensitivity,specificity and accuracy were used to evaluate the pre-dictive performance of the models.Results The AUC values of predictive models of HER-2 low expression based on DCE-MRI,DWI,and the combined models in the trainning and testing sets were 0.754,0.775,0.843 and 0.774,0.645,0.795,respectively.Conclusion Both radiomics feature models based on DCE-MRI and DWI could preoperative predict HER-2 low expression status noninvasively in breast cancer,especially for the combined model,which could be helpful for the selec-tion of clinical treatment in breast cancer.
基金项目
河南省自然科学基金面上项目(202300410081)
河南省医学科技攻关计划项目(LHGJ20220055)
NETL
NSTL
万方数据