多序列3D-MRI影像组学对乳腺小肿块的鉴别价值  

The Value of Multi-Sequence 3D-MRI Radiomics in the Differential Diagnosis of Small Breast Masses

作　　者：赵以惠 陈艾琪 杜小萌 李想 唐聪聪 卢楚鸣 吴明明 钱宝鑫 马宜传[1,3] ZHAO Yihui;CHEN Aiqi;DU Xiaomeng(Department of Radiology,The First Affiliated Hospital of Bengbu Medical College,Bengbu Anhui 233000,P.R.China)

机构地区：[1]蚌埠医学院第一附属医院,233000 [2]蚌埠医学院研究生院,233000 [3]数字医学与智慧健康安徽省重点实验室 [4]汇医慧影医疗科技有限公司,北京100000

出　　处：《临床放射学杂志》2024年第3期352-356,共5页Journal of Clinical Radiology

基　　金：安徽省卫健委科研项目(编号:AHWJ2021b147);蚌埠医学院自然科学重点项目(编号:2020byzd126)。

摘　　要：目的 基于动态增强磁共振(DCE-MRI)序列及扩散加权成像(DWI)序列构建影像组学模型,探讨其对直径≤2 cm的乳腺肿块良恶性的鉴别价值。方法 选取2019年1月至2022年8月就诊于本院122例患者,均接受MRI检查,且经测量肿块直径≤2 cm。将所有患者图像以DICOM格式上传至慧影大数据平台,使用双盲法在DWI及DCE第三期图像上逐层勾画感兴趣区(ROI),后将该病灶勾画的所有ROI融合成三维容积感兴趣区(3D-VOI)进行组学分析。按照4∶1将数据集随机分为训练集与测试集,采用逻辑回归(LR)分类器,构建DCE、DWI及DCE与DWI联合鉴别模型,以病理检查为金标准,评价三种影像组学模型的鉴别效能,并比较三种模型的曲线下面积(AUC)、准确率、特异度及敏感度。结果 根据病理结果将122例患者分为良性42例,恶性80例,以DCE构建组学模型鉴别乳腺小肿块的AUC值为0.83(0.65~1.00)、准确率67%、特异度81%、敏感度67%;以DWI构建组学模型鉴别乳腺小肿块的AUC值0.81(0.67~0.98)、准确率64%,特异度78%、敏感度75%;以DCE与DWI联合模型鉴别乳腺小肿块AUC值0.93(0.80~1.00)、准确率80%、特异度88%、敏感度89%。结论 DCE-MRI与DWI序列联合所建立的模型无创性鉴别乳腺小肿块良恶性的价值更高。Objective Based on dynamic contrast enhancement-magnetic resonance imaging(DCE-MRI) sequence and diffusion weighted imaging(DWI),DWI sequence to construct a radiomics model to explore its value in the differential diagnosis of benign and malignant breast masses less than 2 cm in diameter.Methods A total of 122 patients admitted to our hospital from January 2019 to August 2022 were selected.All patients underwent MRI examination,and the diameter of the tumor was less than 2 cm.The images of all patients were uploaded to Huiyin big data platform with digital imaging and communications in medicine(DICOM),and the region of interest(ROI) was delineated on DWI and DCE phase 3 images layer by layer using the double-blind method,and then all rois delineated by the lesion were fused into Three dimensional volume region of interest(3D-VOI) for omics analysis.The dataset was randomly divided into training set and test set according to 4∶1.Logistic regression(LR) classifier was used to construct DCE,DWI and DCE-DWI combined discrimination models.The area under the curve(AUC),accuracy,specificity and sensitivity of the three models were compared.Results According to the pathological results,the 122 patients were divided into benign group(n=42) and malignant group(n=80).The AUC value of the radiomics model constructed by DCE was 0.83(0.65-1.00),the accuracy was 67%,the specificity was 81%,and the sensitivity was 67%.The AUC,accuracy,specificity and sensitivity of the radiomics model constructed by DWI in identifying small breast masses were 0.81(0.67-0.98),64%,78% and 75%,respectively.The AUC value of the combined model of DCE and DWI was 0.93(0.80-1.00),the accuracy was 80%,the specificity was 88%,and the sensitivity was 89%.Conclusion The combination of DCE-MRI and DWI has a higher value in noninvasive differential diagnosis of small breast masses.

关 键 词：乳腺小肿块 影像组学 3D 

分 类 号：R445.2[医药卫生—影像医学与核医学] R737.9[医药卫生—诊断学]