多模态影像组学预测肿块型乳腺癌术前淋巴结转移的价值

目的探讨基于钼靶、MRI的多模态影像组学预测肿块型乳腺癌术前淋巴结转移的价值。方法收集采用钼靶及MRI检查的111例（114个病灶）肿块型乳腺癌病人的资料。按照8∶2的比例随机将所有病例分为训练组与验证组。对所有图像进行勾画并提取组学特征，利用最小最大值归一化、Select K Best及最小绝对收缩与选择算子回归筛选出有效特征并建立钼靶、MRI、钼靶联合MRI的影像组学预测模型。利用受试者工作特征曲线下面积（AUC）评估模型的效能。结果钼靶、MRI、钼靶联合MRI的影像组学模型在训练组及验证组的AUC值分别为0.76、0.82、0.89及0.74、0.81、0.88。钼靶联合MRI影像组学模型AUC值最大，效能最高。腋窝淋巴结阳性组病灶大小明显大于腋窝淋巴结阴性组（P < 0.01）。结论钼靶联合MRI的多模态影像组学模型可以较好地预测肿块型乳腺癌术前淋巴结转移。

Value of multi-modality radiomics in predicting the preoperative lymph node metastasis of mass type breast cancer

ObjectiveTo explore the value of multi-modality radiomics based on mammography and MRI in predicting the preoperative lymph node metastasis in mass type breast cancer.MethodsThe clinical data of 111 mass type breast cancer patients(114 lesions) detected by mammography and MRI were collected.All patients were randomly divided into the training group and validation group according to the ratio of 8:2.The radiomics features were segmented and extracted from all images.The minimum and maximum normalization, Select K Best and least absolute shrinkage and selection operator were used to obtain the optimal features to establish the mammography, MRI and mammography combined with MRI prediction model.The efficiency of model was evaluated using the area under receiver operating characteristic(ROC) curve(AUC).ResultsThe AUC values of the mammography, MRI and mammography combined with MRI model in the training group and validation group were(0.76, 0.82 and 0.89, respectively) and(0.74, 0.81 and 0.88, respectively).The AUC value and efficiency of mammography combined with MRI model were the highest.The lesion size in positive axillary lymph node group was greater than that in negative axillary lymph node group(P < 0.01).ConclusionsThe multi-modality radiomics model based on mammography combined with MRI can better predict the preoperative lymph node metastasis of mass type breast cancer.