MRI影像组学模型术前预测乳腺癌人表皮生长因子受体2表达状态的价值

Value of MRI image omics model in preoperative prediction of human epidermal growth factor receptor 2 expression in breast cancer

目的探讨常规及功能MRI影像组学在预测乳腺癌人表皮生长因子受体2 (humanepidermalgrowthfactor receptor-2, HER-2)表达状态中的价值。材料与方法 回顾性收集我院2016年1月至2020年5月142例经手术病理证实的乳腺癌患者的完整资料,其中HER-2阳性57例、阴性85例;将患者随机分为训练组(100例,HER-2阳性60例、阴性40例)、验证组(42例,HER-2阳性25例、阴性17例)。所有患者均行乳腺常规和动态对比增强磁共振成像扫描,手动勾画感兴趣区并用AK软件提取纹理特征,利用最小冗余最大相关和最小绝对收缩和选择算子回归方法对纹理特征降维,建立影像组学标签;采用多因素logistic回归构建包含影像组学标签和临床因素的个性化预测模型。通过受试者工作特征(receiver operating characteristic, ROC)曲线和决策曲线分析(decision curve analysis, DCA)评价诊断效能和临床应用价值。结果 临床预测模型在训练组和验证组中预测乳腺癌HER-2阳性的ROC曲线下面积(area under the curve, AUC)分别为0.81和0.69,联合序列影像组学标签的AUC分别为0.89和0.81,个性化预测模型的AUC分别为0.94和0.87。DCA表明个性化预测模型临床应用价值高于临床预测模型及联合序列影像组学标签。结论 个性化预测模型诊断效能优于联合序列影像组学标签及临床预测模型,对预测HER-2表达状态具有一定的价值。

Objective:To explore the value of conventional and functional MRI radiomics in prediction of human epidermal growth factor receptor 2(HER-2)status in breast cancel.Materials and Methods:In this retrospective study,a total of 100 patients with breast cancer confirmed by surgery and pathology were enrolled from January 2016 to May 2020 in our hospital,including 57 cases of HER-2 positive and 85 cases of HER-2 negative.The patients were randomly divided into training group[100 cases,HER-2(+)60 cases,HER-2(-)40 cases],testing group[42 cases,HER-2(+)25 cases,HER-2(-)17 cases].All patients underwent routine and dynamic contrast enhanced magnetic resonance imaging scans of the breast.A region of interest(ROI)of the primary breast tumor in each patient was delineated,and then the texture features of the ROI were extracted by AK.The minimum redundancy maximum redundancy and the least absolute shrinkage and selection operator methods were used to reduce the dimensionality of texture features and establish radiomics signature.Multivariate logistic regression was used to establish individualized prediction model(including clinical factors and radiomics signature).The performance of the model was assessed by area under the receiver operating characteristic curve(AUC).Decision curve analysis(DCA)were used to evaluate the clinical usefulness.Results:The area under the curve(AUC)of the clinical prediction model for positive HER-2 in the training group and the testing group was 0.81 and 0.69,respectively.The AUC of the combined sequentomics label was 0.89 and 0.81,respectively.The AUC of personalized prediction models was 0.94 and 0.87,respectively.DCA indicated that the value of individualized prediction model was higher than clinical prediction model and joint radiomics signature in clinical practice.Conclusions:The individualized prediction model is superior to clinical prediction model and joint radiomics signature,and it has high value in predicting of HER-2 status.