MRI影像组学术前预测浸润性乳腺癌脉管侵犯的价值

Preoperative Prediction of Lymphovascular Invasion in Invasive Breast Cancer with MRI Radiomics

目的探讨基于MRI影像组学术前预测浸润性乳腺癌脉管侵犯的价值。资料与方法回顾性纳入经手术病理证实的140例浸润性乳腺癌患者,其中乳腺癌淋巴血管浸润(LVI)阳性50例,LVI阴性90例。采用完全随机方法将患者分为训练组和验证组。所有患者均行乳腺常规MRI和动态增强扫描。所有病灶全部层面进行手动勾画ROI,采用AK软件提取磁共振纹理特征;采用最小冗余最大相关(m RMR)和最小绝对收缩和选择算子(LASSO)回归对训练组纹理特征降维、建立影像组学标签。利用临床资料、形态学特征、病理结果及影像组学标签建立影像组学模型。使用受试者工作特征(ROC)曲线评价模型预测效能。结果 AK软件共提取3132个影像特征,LASSO回归降维得到10个价值较大的特征,建立影像组学标签。影像组学标签在训练组和验证组中ROC曲线下面积分别为0.81、0.71,影像组学模型在验证组和训练组曲线下面积分别为0.93、0.91,影像组学模型的预测效能更佳。结论基于MRI影像组学模型术前预测浸润性乳腺癌脉管侵犯具有较高的价值。

Purpose To investigate the value of MRI radiomics in preoperative prediction of lymphovascular invasion(LVI)in invasive breast cancer.Materials and Methods 140 patients with invasive breast cancer confirmed by surgery and pathology were retrospectively analyzed,including 50 LVI positive patients and 90 LVI negative patients.All patients were randomly divided into training group and verification group,underwent routine breast MRI and dynamic enhancement scan.Manual ROI mapping was performed on all levels of lesions on MRI images.AK software was used to extract texture features.Minimum redundancy and maximum relevance(mRMR)and minimum absolute contraction and selection operator(LASSO)regression were used to reduce the dimension of texture features,and established radiomics signature.The radiomics model corporated the clinical data,orphological,characteristics,immunohistochemical results and radiomic signature.ROC curve was used to evaluate the prediction efficiency of the model.Results AK software extracted 3132 image features,mRMR and LASSO regression dimensionality reduction obtained 10 valuable features,and established radiomics signature model.ROC curve of the above two models showed that the AUC of the radiomics signature in training group and verification group were 0.81 and 0.71,respectively;the AUC of radiomics model were 0.93 and 0.91,respectively.The prediction efficiency of radiomics model was better.Conclusion It is of value to predict the LVI of invasive breast cancer before surgery based on the MRI radiomics model.