多参数MRI放射组学鉴别诊断乳腺癌分子亚型的研究

Evaluation of Molecular Subtypes of Breast Cancer by Multi Parameter MRI

目的评估多参数磁共振成像(MRI)的放射组学和人工智能(AI)在乳腺癌分子亚型评估中的性能。方法对91例乳腺癌患者进行了3T动态增强(DCE)MRI和弥散加权成像(DWI),并结合表观扩散系数(ADC)图进行分析。提取放射学特征,采用人工神经网络(MLP-ANN)对每组>20名患者进行两两比较(70%用于训练,30%用于验证,各5次)。结果以组织病理学为参考标准。MLP-ANN在受试者-工作特征曲线(AUC)下的总体中位数面积(AUC)为0.86(0.77-0.92),用于区分三重阴性(TN)与其他癌症。分离luminal内A和TN癌的总体中位数AUC为0.8(0.75-0.83)。结论来自多参数MRI的放射组学和人工智能可能有助于非侵袭性区分TN和luminal A乳腺癌与其他亚型。

Objective To evaluate the performance of multi parameter magnetic resonance imaging(MRI)and artificial intelligence(AI)in the evaluation of molecular subtypes of breast cancer.Methods 91 patients with breast cancer were studied by using 3T dynamic contrast enhanced(DCE)MRI and diffusion weighted imaging(DWI),and the apparent diffusion coefficient(ADC)images were analyzed.Extract radiological features and use an artificial neural network(MLP-ANN)for pairwise comparison of>20 patients in each group(70%for training,30%for validation,5 times each).Results Using histopathology as the reference standard.The overall median area(AUC)of MLP-ANN under the subject working characteristic curve(AUC)is 0.86(0.77-0.92),used to distinguish triple negative(TN)from other cancers.The overall median AUC for separating intracavitary A and TN cancers is 0.8(0.75-0.83).Conclusion Radioomics and artificial intelligence from multiparameter MRI may be helpful in noninvasive differentiation of TN and CA-A breast cancer from other subtypes.