DCE-MRI影像联合临床信息预测乳腺癌复发风险评分

Prediction of Oncotype DX RS in breast cancer by integrating of DCE-MRI radiomics and clinicopathologic data

探索基于动态增强磁共振成像(dynamic contrast enhanced magnetic resonance imaging,DCE-MRI)的影像组学特征联合临床病理信息对乳腺癌21基因检测复发风险评分(Recurrence Score,RS)的预测作用。采集130例雌激素受体(Estrogen Receptor,ER)阳性且无淋巴结转移的早期乳腺癌患者数据;分割影像病灶区域,提取形态、统计、纹理特征;对特征进行单变量预测分析,并运用网格搜索和十折交叉验证相结合的方法选择最佳参数组合和最优特征子集建立弹性网络回归模型进行多变量预测分析。在单变量预测分析中,有8维影像特征和1维临床病理特征与RS显著相关(P<0.05)。在多变量预测分析中,基于影像特征建立的模型R^(2)为0.264(P=0.038),联合临床病理信息后,R^(2)提高到0.295(P=0.033)。结果表明,DCE-MRI影像组学特征和临床病理信息与RS存在关联。

In this study,we established the statistical model by integrating the dynamic enhanced magnetic resonance imaging(DCE-MRI)and clinicopathological features for prediction of breast cancer Oncotype DX recurrence score(RS).The early breast cancer patients were enrolled with estrogen receptors(ER)positive and lymph node metastasis negative.The lesion areas were obtained,where the radiomic features were extracted,including morphological,statistical and texture features.Afterwards,univariate analysis of radiomics features were examined.The predictive model was established based on the selected features and the model parameters selected by a grid search method under the ten-fold cross-validation.Thereafter,an elastic network regression model was built for multivariate regression analysis.In the univariate feature analysis,8 imaging features and one clinicopathological features were significantly correlated with RS(P<0.05).The multivariate regression model was established using the selected image features,which yield the R square of 0.264(P=0.038).The prediction model was improved in terms of R square of 0.295(P=0.033)after combing clinicopathological factors.The results showed that DCE-MRI radiomics and clinicopathologic information can be used as biomarkers for association with the recurrence of breast cancer.