基于药代动力学动态增强磁共振的影像组学特征对三阴型乳腺癌诊断价值的研究

目的 探讨基于药代动力学动态增强MRI(dynamic contrast-enhanced MRI,DCE-MRI)的全肿瘤影像组学特征对三阴型乳腺癌的诊断价值.材料与方法 回顾性分析85例治疗前行DCE-MRI扫描的乳腺癌患者,Luminal型39例、人表皮生长因子受体2(human epidermal growth...

Radiomics features based on pharmacokinetic dynamic contrast-enhanced magnetic resonance imaging for identifying triple negative breast cancer

Objective:To study the evaluation of radiomics features based on pharmacokinetic dynamic contrast-enhanced MRI(DCE-MRI)for differentiating triple negative(TN)breast cancer from other molecular subtype breast cancers.Materials and Methods:This retrospective study included 85 patients with breast cancer who underwent pharmacokinetic DCE-MRI before treatment.Breast cancers were classified into four molecular subtypes by immunohistochemistry,including Luminal(n=39),human epidermal growth factor receptor 2(HER-2)overexpression(n=16)and TN(n=30).Radiomics features of whole breast cancer were extracted from pharmacokinetic quantitative and enhanced images,respectively.Spearman correlation and least absolute shrinkage and selection operator(LASSO)were used for feature selection in R.Logistic model was used for classification of TN vs.luminal,TN vs.HER-2 overexpression,and TN vs.non-TN.Receiver operating characteristics curve and area under curve(AUC)were obtained.Five-fold cross validation was used to verify classification performance.Results:For the TN vs.luminal breast cancer,6 optimal features were selected.Accuracy,and AUC were 0.783 and 0.865,respectively.For the TN vs.HER-2 overexpression breast cancer,14 optimal features were selected.Accuracy and AUC were 0.870 and 0.923,respectively.For the TN vs.non-TN breast cancer,17 optimal features were selected.Accuracy and AUC were 0.847 and 0.913,respectively.Conclusions:The rediomics features can help to differentiate TN from other molecular subtype breast cancer.