摘要
联合动态增强磁共振成像(DCE-MRI)、T2加权成像(T2WI)以及弥散加权成像(DWI)的影像特征,建立基于多参数影像组学的预测模型,分别对乳腺癌分子分型、组织学分级和Ki-67表达进行预测。采集150例术前、化疗前的浸润性导管癌患者乳腺MRI数据,获取DCE-MRI、T2WI和DWI影像。分割各参数影像的病灶区域,并提取多参数影像特征。在训练集采用支持向量机递归特征消除(SVM-RFE)算法,获得影像组学最优特征子集并构建基于SVM的预测模型,在测试集中测试模型性能。采用概率平均法、概率投票法和概率模型优化法,分别将基于不同参数影像构建的预测模型进行融合,得到多参数影像联合预测结果,并计算ROC曲线下的面积(AUC)评估模型的分类性能。单参数影像模型预测Luminal A、Luminal B、HER2和Basal-like等4种分子分型的最佳AUC分别为0.672 1、0.694 0、0.677 7和0.708 6,多参数影像模型的预测结果提高到AUC分别为0.799 5、0.727 9、0.737 5和0.792 5。单参数影像模型预测分级的最佳AUC为0.753 3,多参数影像模型的预测结果提高到0.801 7。单参数影像模型预测Ki-67表达的最佳AUC为0.664 7,多参数影像模型预测结果提高到0.771 8。相比于单参数影像模型的预测结果,多参数影像模型的预测结果有所提升,且差异具有显著性(P<0.05)。实验结果表明,采用多参数磁共振影像(DCE-MRI、T2WI以及DWI)组学的联合,可以显著提高单一参数影像模型预测乳腺癌病理信息的性能,对乳腺癌的诊断和个性化治疗方案的选择具有重要意义。
To create a prediction model based on multiparametric magnetic resonance imaging( MRI) radiomics features extracted from dynamic enhanced magnetic resonance imaging( DCE-MRI),T2 weighted imaging( T2 WI) and diffusion weighted imaging( DWI) to predict molecular subtypes,histological grade and Ki-67 expression of breast cancer. In this study,150 cases of breast invasive ductal carcinoma before surgery and chemotherapy were collected,and multiparametric images of DCE-MRI,T2 WI and DWI were obtained. Breast tumor areas in the different parametric images were segmented and multiparametric imaging features were extracted. The best imaging feature subset was obtained using support vector machine recursive feature elimination( SVM-RFE) algorithm,and a prediction model based on SVM was created using the training set of each parameter imaging series. The performance of prediction model was tested in the test set. The prediction models for all parameter imaging series were fused using the probabilistic averaging method,the probabilistic voting method,and the probabilistic model optimization method. The prediction performance was evaluated by calculating the area under the ROC curve( AUC). The single-parametric imaging models discriminated among the Luminal A,Luminal B,HER2,and Basal-like subtypes with the best AUC values of 0. 672 1,0. 694 0,0. 677 7,and0. 708 6,respectively,and the prediction performance of multiparametric imaging models was increased to AUC of0. 799 5,0. 727 9,0. 737 5 and 0. 792 5,respectively. The single-parametric imaging models discriminated among histological grades with the best AUC values of 0. 753 3,and the prediction performance of multiparametric imaging model was increased to AUC of 0. 801 7. The single-parametric imaging models discriminated among Ki-67 expression with the best AUC values of 0. 664 7,and the prediction performance of multiparametric imaging model was increased to AUC of 0. 771 8. The prediction accuracy of multiparametric imaging models was increased significantly compared to
作者
娄潇方
范明
许茂盛
王世威
厉力华
Lou Xiaofang;Fan Ming;Xu Maosheng;Wang Shiwei;Li Lihua(Institute of Biomedical Engineering and Instrument,Hangzhou Dianzi University,Hangzhou 310018,China;Department of Radiology,Zhejiang Hospital of Traditional Chinese Medicine,Hangzhou 310006,China)
出处
《中国生物医学工程学报》
CAS
CSCD
北大核心
2020年第5期513-523,共11页
Chinese Journal of Biomedical Engineering
基金
国家自然科学基金(61871428,61731008)
浙江省自然科学基金(J19H180004)。
