摘要
目的 探讨动态对比增强MRI(DCE-MRI)和扩散加权成像(DWI)影像组学模型在预测浸润性乳腺癌(IBC)病理分级方面的价值。方法 回顾性分析经病理证实为浸润性乳腺癌的142例患者的影像学资料。按病理分级分为低级别组(n=93)和高级别组(n=49)。利用达尔文科研平台对患者的DCE-MRI和DWI图像病灶进行特征提取,采用logistic回归分析建模,通过对受试者操作特征(ROC)曲线和准确性进行分析,探讨DCE图、DWI图以及两者联合的影像组学标签在预测IBC病理分级方面的价值。结果 单独DCE模型最终提取5个影像组学特征,DCE联合DWI模型最终提取5个影像学特征,单独DWI模型最终提取10个影像组学标签。DCE图与DWI图训练集的曲线下面积(AUC)分别为0.82(0.68,0.96)、0.77(0.62,0.91),准确性分别为0.74、0.68;测试集的AUC分别为0.79(0.62,0.97)、0.66(0.46,0.86),准确性为0.72、0.60;DCE联合DWI图的模型训练集AUC为0.82(0.69,0.95),准确性为0.74;测试集AUC为0.76(0.59,0.96),准确性为0.65。结论 单独DCE图像及DCE联合DWI图像影像组学标签对预测高级别与低级别IBC有一定价值,单独DCE影像组学模型优于DCE联合DWI的影像组学模型。
Objective To investigate the value of dynamic contrast enhanced MRI(DCE-MRI) and diffusion weighted imaging(DWI) radiomics models in predicting the pathological grading of invasive breast cancer(IBC). Methods The imaging data of 142 cases with pathologically confirmed invasive breast cancer were retrospectively analyzed.According to pathological grade, they were divided into low grade group(n=93) and high grade group(n=49).DCE MRI and DWI images of the patients were used to extract the features of the lesions using Darwin scientific research platform.Logistic regression analysis modeling was used to analyze the receiver operating characteristic(ROC) curve and accuracy.The value of DCE,DWI and combined imaging omics labels in predicting the pathological grading of IBC was discussed. Results Five imaging features were finally extracted from DCE model alone, five imaging features were finally extracted from DCE combined with DWI model, and 10 imaging labels were finally extracted from DWI model alone.The areas under the AUC curve of DCE and DWI training sets were 0.82(0.68,0.96) and 0.77(0.62,0.91) respectively, and the accuracy rates were 0.74 and 0.68 respectively.The areas under the AUC curve of the test set were 0.79(0.62,0.97) and 0.66(0.46,0.86),respectively, and the accuracy rates were 0.72 and 0.60.AUC of model training set of DCE combined with DWI map was 0.82(0.69,0.95),and accuracy was 0.74.The AUC of the test set was 0.76(0.59,0.96) and the accuracy rate was 0.65. Conclusion DCE image alone and DCE combined DWI radiomics label have certain value in predicting high and low IBC,and DCE radiomics model alone is slightly better than DCE combined DWI radiomics model.
作者
杜小萌
李艳玲
曹胜男
陈艾琪
李想
唐聪聪
马宜传
DU Xiaomeng;LI Yanling;CAO Shengnan(Department of Radiology,The First Affiliated Hospital of Bengbu Medical College,Bengbu,Anhui Province 233099,P.R.China)
出处
《临床放射学杂志》
北大核心
2022年第9期1641-1644,共4页
Journal of Clinical Radiology
基金
安徽省卫生健康委科研项目(编号:AHWJ2021b147)
蚌埠医学院自然科学重点项目基金资助项目(编号:2020byzd126)。