乳腺磁共振BI-RADS 4类病变恶性风险预测列线图诊断模型的建立及应用

Building and application of breast magnetic resonance imaging BI-RADS 4 lesions for predicting the risk of malignancy using a nomogram diagnostic model

目的:回顾性分析乳腺磁共振BI-RADS 4类病例临床及影像资料,筛选乳腺恶性病变的影响因素,建立乳腺恶性风险预测列线图诊断模型并评估其诊断价值。方法:收集2017年1月至2020年12月武汉大学中南医院乳腺MRI诊断为BI-RADS 4类的病例资料,分为肿块型及非肿块样强化(NME)型两类并按3∶1比例随机各分为实验组及验证组,分析比较临床及影像特征,采用Logistic回归分析筛选恶性病变的独立危险因素,使用ROC曲线分析其价值,构建列线图诊断模型并对病例再分类。结果:418例病例符合纳入标准。肿块型病变中年龄大于39.5岁、边缘毛刺、最大信号强度投影(MIP)阳性、时间信号强度曲线(TIC)流出型及表观扩散系数(ADC)<1.063×10^(-3)mm^(2)/s是恶性病变的独立危险因素,ROC曲线下面积(AUC)值为0.968(95%CI:0.943~0.994);恶性病变列线图诊断模型实验组、验证组的C-index分别为0.962(95%CI:0.930~0.994)、0.922(95%CI:0.846~0.998),22.1%(52/235)病变再分类降级为3类,11.9%(28/235)病变升级为5类。NME型病变中病灶钙化、ADC<1.036×10^(-3)mm^(2)/s、MIP阳性及病灶分布类型(局灶、多区域、弥漫、节段样)是恶性独立危险因素,皮肤红肿为保护因素,AUC值为0.964(95%CI:0.934~0.995)。恶性病变列线图诊断模型实验组、验证组的C-index分别为0.956(95%CI:0.918~0.994)、0.905(95%CI:0.816~0.994),16.4%(30/183)NME病变再分类降级为3类,18.0%(33/183)NME病变升级为5类。结论:基于乳腺磁共振BI-RADS 4类病变的临床及影像特征建立的恶性风险预测列线图模型诊断价值高,可用于4类病变的再分类。

Objective:To retrospectively analyze the clinical and imaging data of the cases diagnosed as BIRADS category 4 lesions by MRI,screen out the influencing factors for breast malignancy,establish a nomogram diagnostic model for predicting breast malignancy risk,and evaluate its diagnostic value.Methods:We collected case data of breast MRI diagnosis as BI-RADS category 4 lesions in Zhongnan Hospital of Wuhan University from January 2017 to December 2020.The data was divided into mass type and non-mass enhancement(NME)type,then randomly divided into experimental group and validation group in a ratio of 3∶1 in each type.The clinical and imaging characteristics of these cases were analyzed and compared,using logistic regression analysis to screen for independent risk factors of malignant lesions,and the ROC curve was drawn to analyze the value,construct a nomogram diagnostic model,and reclassify cases.Results:In this study,a total of 418 cases met the inclusion criteria.Age greater than 39.5 years old,marginal spiculation,maximum intensity projection(MIP)positivity,efflux type in time-intensity curve(TIC),and apparent diffusion coefficient(ADC)<1.063×10^(-3) mm^(2)/s were independent risk factors for malignancy in mass type lesions.The area under the ROC curve(AUC)value was 0.968(95%CI:0.943-0.994);The C-index of the malignant column chart diagnostic model in experimental group and validation group were 0.962(95%CI:0.930-0.994)and 0.922(95%CI:0.846~0.998),respectively.22.1%(52/235)of lesions were reclassified and downgraded to category 3,and 11.9%(28/235)of lesions were upgraded to category 5.In NME-type lesions,calcification,ADC<1.036×10^(-3) mm^(2)/s,MIP positivity,and lesion distribution type(focal,multi-regional,diffuse,segment)were independent risk factors for malignancy,while skin redness and swelling were protective factors.The AUC value was 0.964(95%CI:0.934-0.995).The C-index of the malignant nomogram diagnostic model experimental group and validation group were 0.956(95%CI:0.918-0.994)and 0.905(95%CI:0.816-0.994),respectively.16.4%(30/183)of NME lesions were reclassified and downgraded to category 3,and 18.0%(33/183)of NME lesions were upgraded to category 5.Conclusion:The malignant risk prediction nomogram diagnostic model based on clinical and MRI imaging features of breast BI-RADS category 4 lesions has high diagnostic value and can be used for reclassification.