基于多序列MRI影像组学及临床病理术前预测三阴性乳腺癌淋巴结转移

Preoperative Prediction of Lymph Node Metastasis in Triple-negative Breast Cancer Based on Multi-sequence MRI Radiomics and Clinicopathology

目的 探讨基于多序列MRI影像组学及临床病理的诺莫图术前预测三阴性乳腺癌(TNBC)腋窝淋巴结转移的可行性。方法回顾性分析120例TNBC患者的临床病理及MRI资料,按7:3比例将其随机分为训练集及验证集,在DCE-MRI第一期和DWI图像上勾画ROI并提取影像组学特征。依次采用一致性相关系数及最小绝对收缩和选择算法进行特征降维并构建影像组学标签。(1)采用多因素逻辑回归分析探寻临床病理特征中的独立影响因素;(2)采用多因素逻辑回归分析分别建立基于DCE-MRI、DWI及多序列(DCE-MRI联合DWI)的影像组学预测模型;(3)采用最优影像组学模型的组学特征构建组学评分,并联合临床病理独立影响因素构建诺莫图模型。统计分析:采用受试者工作特征曲线评价其诊断效能,使用Delong检验评价不同模型间曲线下面积(AUC)的差异,应用校准曲线及决策曲线分析(DCA)评估模型的效能及临床效益。结果(1)最大肿瘤直径和血管浸润状态为临床病理独立影响因素;(2)相比单序列模型,基于多序列的影像组学模型效能更优(AUC为0.79);(3)联合影像组学和临床病理独立影响因素进一步提高了诊断效能,诺莫图模型的AUC达0.89,且DCA显示其具有更高的临床价值。结论联合多序列MRI影像组学及临床病理特征构建的诺莫图可准确术前预测TNBC腋窝淋巴结转移状态,具有较高的临床应用价值。

Objective To investigate the feasibility of preoperative prediction of axillary lymph node metastasis in triple-negative breast cancer(TNBC) based on nomogram constructed by multi-sequence radiomics and clinicopathology.Methods The clinicopathological and MRI data of 120 TNBC patients were retrospectively collected and randomly divided into training and validation sets in a 7:3 ratio.Regions of interest were sketched on the first phase of DCE-MRI and DWI images and the corresponding radiomic features were extracted.The dimensionality of radiomic features was reduced by sequentially applying the consistent correlation coefficient and minimum absolute shrinkage and selection algorithms.(1) Independent factors of the clinicopathological features were also identified by multivariate logistic regression analysis.(2) Multivariate logistic regression analysis was used to build radiomics prediction models with DCE-MRI,DWI,and multi-sequence(DCE-MRI combined with DWI),respectively.(3) A radiomics score was constructed with the radiomic feature of the optimal radiomics model,and then a nomogram model was constructed by combining the radiomics score and the independent clinicopathological factors.Statistical analysis:The diagnostic efficacy was evaluated using the receiver operating characteristic curve,and the differences of the area under the curve(AUC) between the models were evaluated using the Delong's test.Clinical benefits were evaluated by calibration curve and the decision curve analysis(DCA).Results(1) Maximum tumor diameter and vascular invasion status were clinicopathologically independent factors;(2) Compared with the radiomics model based on DCE-MRI or DWI,the multi-sequence model was more accurate(AUC of 0.79);(3) The AUC of the nomogram model was 0.89,and the DCA showed a higher clinical value,demonstrating that the combination of radiomics and independent clinicopathological factors further improved diagnostic efficacy.Conclusion The nomogram model constructed by the combination of multi-sequence MRI radiomics and clinicopathological feature can accurately predict the status of axillary lymph node metastasis in TNBC,and is of high clinical value.