乳腺癌术前MRI特征对保乳术中切缘阳性的预测价值分析

The value of preoperative MRI features of breast cancer in predicting the positive margins during breast-conserving surgery

目的:研究乳腺癌术前MRI特征对保乳术中肿瘤切缘阳性的预测诊断价值。方法:回顾分析本院2017年9月-2019年6月有完整术前MRI资料且首次手术为保乳术的176例浸润性乳腺癌进行研究,根据术中病理的切缘情况分为切缘阳性组(56例)和切缘阴性组(120例)。分析每个病例的术前MRI图像,提取包括乳腺纤维腺体类型,同侧和对侧背景实质强化类型,病灶类型,肿块形态和边缘特征,非肿块样强化病灶分布特征,病灶最长径,是否为多中心病灶等MRI特征以及基本的临床、病理特征。通过单因素比较两组间各种MRI特征和基本临床病理特征的差异,再将单因素分析有统计学意义的MRI特征进行多因素Logistic回归分析,采用受试者操作特征(ROC)曲线评价回归模型对保乳术中切缘阳性的预测诊断价值。结果:本组研究的保乳术中切缘阳性率为31.8%(56/176例)。单因素分析结果显示,在MRI特征中,较大的病灶最长径(t=3.162,P=0.002),存在多中心病灶(F=17.853,P=0.000),非肿块样强化的病灶类型(F=11.019,P=0.001)是保乳术中切缘阳性的预测特征;在基本临床病理特征中,浸润性癌伴导管原位癌成分(F=21.475,P=0.000),HER2阳性表达(F=5.663,P=0.017)是保乳术中切缘阳性的预测因素。多因素回归分析结果显示,对浸润性乳腺癌保乳术中切缘阳性预测最有价值的MRI特征是多中心病灶(OR=4.381,P<0.001)和较大的病灶最长径(OR=1.796,P=0.019),病理特征是浸润性乳腺癌伴导管原位癌成分(OR=4.211,P<0.001),回归模型对DCIS-IC预测诊断的ROC曲线下面积为0.75。结论:浸润性乳腺癌部分术前MRI特征和病理特征可以帮助临床医生对保乳术中切缘阳性情况进行预测。

Objective:To study the value of preoperative MRI features of breast cancer in predicting positive margins during breast-conserving surgery(BCS).Methods:A total of 176 patients with invasive breast cancer were retrospectively enrolled from Sep 2017 to July 2019 and divided into positive margins group(56 cases) and negative margins group(120 cases).Preoperative MRI features including fibro-glandular tissue(FGT) type, bilateral background parenchymal enhancement(BPE),the type, shape and margin of mass, the distribution of none mass-like enhancement(NMLE),maximum long diameter of lesion, whether or not with multi-centric lesion were analyzed.The basic clinical and pathological information were also collected.The differences between the two groups among the collected features were compared by univariate analysis.The statistical features in univariate analysis were further analyzed by multivariate logistic-regression model.ROC analyses were applied to determine the performance of the model.Results:The positive margins rate was 31.38%(56/176) in this study.In univariate analysis, several MRI features including largest long diameter of tumor(t=3.162,P=0.002),multi-centric lesion(F=17.853,P=0.000),NMLE type(F=11.019,P=0.001) and the clinical-pathology features, including invasive cancer with intra-ductal carcinoma in situ(DCIS) component, HER2 positive, were the predictors of the positive margins during BCS.The logistic regression analysis showed that the invasive cancer with multi-centric lesion(OR=4.381,P=0.000),larger long diameter(OR=1.796,P=0.019) and DCIS component(OR=4.211,P=0.000) and were prone to have positive margins in BCS,and the AUC of the model was 0.750.Conclusion:Several preoperative MRI features and pathological features of invasive breast cancer could help clinicians to predict the positive margins in BCS.