基于术前MRI特征及定量参数列线图预测乳腺癌患者淋巴血管侵犯

Nomogram for predicting lymphovascular invasion in breast cancer using MRI features and quantitative parameters

目的开发基于术前磁共振成像(magnetic resonance imaging,MRI)特征及表观扩散系数(apparent diffusion coefficient,ADC)的列线图预测浸润性乳腺癌淋巴血管侵犯(lymphovascular invasion,LVI)。材料与方法回顾性分析141例浸润性乳腺癌患者资料,其中2019年3月至2021年7月的66例患者作为训练组,2021年7月至2022年12月的75例患者作为验证组。评估病灶MRI特征(肿块形状、边缘、内部强化形式、瘤周水肿等)及测量ADC。采用卡方检验、独立样本t检验及Mann-Whitney U检验分析LVI阳性组和LVI阴性组间影像学特征和瘤内、瘤周ADC平均值、最大值及最小值的差异。采用多因素逻辑回归分析确定与LVI相关的独立危险因素,建立预测LVI的列线图模型。结果LVI阳性组和LVI阴性组间在肿瘤大小、Ki-67表达以及淋巴结转移率方面的差异具有统计学意义(P<0.05)。多因素逻辑回归分析显示,肿块形态[P=0.014,优势比(odds ratio,OR)=0.142(0.030~0.679)]、内部强化形式[P=0.046,OR=0.157(0.025~0.965)]、肿瘤最大径[P=0.037,OR=4.372(1.093~17.488)]、扩散加权成像(diffusion-weighted image,DWI)边缘征[P=0.024,OR=0.193(0.047~0.803)]及瘤周瘤内ADC比值[P=0.010,OR=1.056(1.013~1.101)]是LVI的独立预测因素。基于MRI特征和瘤周瘤内ADC比值的综合预测模型在训练组及验证组中的受试者工作特征(receiver operating characteristic,ROC)曲线下面积(area under the curve,AUC)分别为0.867、0.872,特异度分别为88.6%、84.6%,精确度分别为74.2%、69.7%。校准曲线表明,列线图在训练组和验证组中预测值与实际值吻合较好。结论基于术前MRI影像特征及ADC的列线图评估乳腺癌LVI状态具有较高的预测效能,有助于临床在指导手术、制订个性化治疗方案及评估预后等方面的决策。

Objective:To develop a nomogram based on preoperative magnetic resonance imaging(MRI)features and apparent diffusion coefficient(ADC)to predict lymphovascular invasion(LVI)in invasive breast cancer.Materials and Methods:A retrospective analysis was conducted on data from 141 patients with invasive breast cancer.Among them,66 patients from March 2019 to July 2021 were assigned to the training group,and 75 patients from July 2021 to December 2022 were assigned to the validation group.The evaluation included assessment of lesion MRI features(such as mass shape,margins,internal enhancement patterns,and peritumoral edema)and measurement of ADC.The differences in imaging features and mean,maximum,and minimum ADC values within the tumor and peritumoral regions between the LVI-positive and LVI-negative groups were analyzed using chi-square tests,independent sample t-tests,and Mann-Whitney U tests.Multivariate logistic regression analysis was employed to identify independent risk factors associated with LVI and to establish a nomogram model for predicting LVI.Results:LVI was significantly associated with tumor size,Ki-67 expression,and lymph node metastasis rate.Multivariable logistic regression analysis showed that tumor shape[P=0.014,odds ratio(OR):0.142(0.030-0.679)],internal enhancement patterns[P=0.046,OR:0.157(0.025-0.965)],maximum tumor diameter[P=0.037,OR:4.372(1.093-17.488)],DWI rim[P=0.024,OR:0.193(0.047-0.803)],and ADCration[P=0.010,OR:1.056(1.013-1.101)]were independent predictors of LVI.Areas under the receiver operating characteristic curve of the comprehensive prediction model based on MRI features and peritumoral intratumoral ADC ratio in the training and validation groups were 0.867 and 0.872,the specificity were 88.6%and 84.6%,and the precision were 74.2%and 69.7%,respectively.Calibration curves demonstrated good agreement between predicted and actual values in the training and validation groups.Conclusions:LVI is correlated with various clinical and pathological prognostic factors and MRI imaging features.The comprehensive model based on preoperative MRI features and apparent diffusion coefficient demonstrates good predictive efficiency for LVI,that contribute to clinical decision-making in guiding surgery,developing individualized treatment plans,and assessing prognosis.