基于MRI瘤内联合最佳瘤周影像组学评估乳腺癌Ki⁃67表达状态的研究  

Assessment of Ki⁃67 Expression Status in Breast Cancer Based on MRI Intratumoral Combined with Optimal Peritumoral Radiomics

作　　者：黄静[1,2] 马彦云[2] 武静[1] 李彪 HUANG Jing;MA Yanyun;WU Jing(School of Medical Imaging,Shanxi Medical University,Taiyuan,Shanxi Province 030000,P.R.China)

机构地区：[1]山西医科大学医学影像学院,太原030000 [2]山西医科大学第一医院放射科,太原030000

出　　处：《临床放射学杂志》2024年第8期1317-1324,共8页Journal of Clinical Radiology

基　　金：2022年山西省高等学校教学改革创新项目(J20220416);山西省重点研发计划项目(编号:201803D31100)。

摘　　要：目的探讨基于磁共振成像(MRI)瘤内联合最佳瘤周影像组学模型评估乳腺癌Ki⁃67表达状态的价值。方法回顾性分析符合本研究纳排标准的150例乳腺癌患者资料。在动态对比增强磁共振成像(DCE⁃MRI)第3期图像中勾画肿瘤感兴趣区(ROI)并扩充生成多个瘤周ROI,手动调整勾画范围并分别融合成瘤内及各瘤周肿瘤感兴趣体积(VOI),提取并筛选影像组学特征,利用支持向量机(SVM)分别构建瘤内及各瘤周影像组学模型,确定最佳瘤周区域并与瘤内共同构建瘤内瘤周联合模型;对临床病理及MRI影像特征进行单因素及多因素Logistic统计学分析并构建临床模型。最终,基于瘤内联合最佳瘤周组学特征与筛选出的临床影像特征共同建立综合模型。使用受试者工作特征(ROC)曲线下面积(AUC)评估模型性能,Delong检验比较各模型预测性能的差异。对于术前核心活检与术后病理标本评估Ki⁃67表达水平不一致的病例,利用综合模型进行验证,进一步评估模型效能。结果瘤周3 mm组学模型在各瘤周模型中效能最优,瘤内、瘤周3 mm及瘤内瘤周联合模型在测试集的AUC分别为0.809、0.779及0.867,临床模型及综合模型在测试集的AUC分别为0.716、0.887,综合模型的预测性能优于瘤内瘤周联合模型及临床模型。综合模型在术前核心活检及术后病理标本评估Ki⁃67不一致的病例中准确性达83.3%。结论基于MRI瘤内联合最佳瘤周区域的影像组学模型能够术前有效预测乳腺癌Ki⁃67表达状态,其中纳入临床影像特征构建的综合模型效能更优。Objective To investigate the value of intratumoral combined with optimal peritumoral radiomics based on magnetic resonance imaging(MRI)for evaluating the Ki⁃67 expression status in breast cancer.Methods The data of 150 breast cancer patients who met the inclusion and exclusion criteria of this study were retrospectively analysed.In the DCE⁃MRI phase 3 images,intratumoral regions of interest(ROI)were outlined and expanded to generate multiple peritumoral ROIs,manually adjust the outlining range and fuse them to form intratumour and peritumour volumes of interest(VOI)respectively,Radiomics features were extracted and screened,and construct the intratumour and multiple peritumour image radiomics model using SVM,The optimal peritumoral region was determined and combined with the intratumoral region to construct a joint intratumoral and peritumoral model;clinical pathological and MRI imaging features were analysed by univariate and multivariate logistic statistics,and clinical models were constructed.Ultimately,a comprehensive model was built based on the joint best peritumour radiomics features within the tumour in conjunction with the screened clinical imaging features.The area under the receiver operating characteristic(ROC)curve(AUC)was used to evaluate the performance of models,and Delong's test compared the differences in predictive performance between the models.When Ki⁃67 expression levels were inconsistent between preoperative core biopsies and postoperative pathological specimens,we validated the integrated model using an integrated model to assess its efficacy.Results The peritumour 3mm radiomics model had the best efficacy among multiple peritumour models,and the AUC of the intratumour,peritumour 3mm and combined intratumour and per⁃itumour models were 0.809,0.779 and 0.867 in the test set.respectively,while the AUCs of the clinical model and the comprehensive model were 0.716 and 0.887 in the test set.The combined model had a superior predictive performance than the combined intratumour and peritu

关 键 词：乳腺癌 磁共振成像 影像组学 Ki⁃67 瘤周区域 

分 类 号：R737.9[医药卫生—肿瘤]