Page 80 - 南京医科大学学报自然科学版
P. 80

南京医科大学学报(自然科学版)                                  第44卷第2期
               ·218 ·                     Journal of Nanjing Medical University(Natural Sciences)   2024年2月


             ·临床研究·

              多参数 MRI 影像组学评估浸润性乳腺癌 HER⁃2 表达状态的临

              床应用价值



              刘婷婷,林佳璐,娄鉴娟,邹启桂,王思奇,蒋燕妮                    *
              南京医科大学第一附属医院放射科,江苏 南京                  210029




             [摘    要] 目的:探讨基于多参数磁共振成像(magnetic resonance imaging,MRI)影像组学特征的模型预测浸润性乳腺癌人类
              表皮生长因子受体⁃2(human epidermal growth factor receptor⁃2,HER⁃2)表达状态的价值。方法:回顾性分析南京医科大学第一
              附属医院2018年1月—2019年12月401例乳腺癌患者的基线期MRI图像及临床资料。使用ITK⁃SNAP软件在快速反转恢复
              序列(turbo inversion recovery magnitude,TIRM)、动态对比增强磁共振成像第2期(dynamic⁃contrast enhanced magnetic resonance
              imaging phase 2,DCE2)、动态对比增强磁共振成像第 4 期(dynamic⁃contrast enhanced magnetic resonance imaging phase 4,
              DCE4)、弥散加权成像(diffusion⁃weighted imaging,DWI)和表观弥散系数(apparent diffusion coefficient,ADC)的最大肿瘤层面手
              动勾画二维感兴趣区域(region of interest,ROI),并对所勾画的ROI区域进行特征提取及降维筛选。应用逻辑回归(logistic re⁃
              gression,LR)算法建立预测HER⁃2表达状态的单参数模型、组合模型和多参数模型。结果:最终筛选出26个最优特征,其中按
              权重排序位居首位的特征为DCE2_original_shape_SurfaceVolumeRatio。单参数模型中预测效能最好的是DCE2模型,训练集及
              测试集的曲线下面积(area under curve,AUC)分别为0.907、0.879;组合模型中联合增强特征的模型比其他未联合增强特征的
              模型预测效能更好(P均≤ 0.001);多参数模型预测效能最佳(训练集及测试集的AUC值分别为0.932、0.906)。结论:基于多参
              数影像特征构建的影像组学模型评估浸润性乳腺癌HER⁃2表达状态有一定的临床价值,其中增强早期特征的预测价值较高。
             [关键词] 乳腺癌;影像组学;HER⁃2;磁共振成像;异质性

             [中图分类号] R737.9                   [文献标志码] A                        [文章编号] 1007⁃4368(2024)02⁃218⁃10
              doi:10.7655/NYDXBNSN230584



              Clinical application value of multi ⁃ parameter MRI radiomics evaluation of HER ⁃ 2
              expression status in invasive breast cancer

              LIU Tingting,LIN Jialu,LOU Jianjuan,ZOU Qigui,WANG Siqi,JIANG Yanni *
              Department of Radiology,the First Affiliated Hospital of Nanjing Medical University,Nanjing 210029,China



             [Abstract] Objective: To investigate the value of a model based on multi⁃parameter magnetic resonance imaging(MRI)radiomics
              features in predicting the expression status of human epidermal growth factor receptor⁃2(HER⁃2)in invasive breast cancer patients.
              Methods:A retrospective analysis was conducted on baseline MRI images and clinical data of 401 breast cancer patients from January
              2018 to December 2019 at the First Affiliated Hospital of Nanjing Medical University. The 2D region of interest(ROI)were segmented
              manually on ITK ⁃ SNAP software at the maximum tumor level of turbo inversion recovery magnitude(TIRM),dynamic ⁃ contrast
              enhanced magnetic resonance imaging phase 2(DCE2),dynamic⁃contrast enhanced magnetic resonance imaging phase 4(DCE4),
              diffusion⁃weighted imaging(DWI)and apparent diffusion coefficient(ADC). Feature extraction and dimensionality reduction screening
              were performed on the delineated ROIs. Logistic regression(LR)algorithm was used to construct single⁃parameter,combination and
              multi⁃parameter models for predicting HER⁃2 expression status. Results:A total of 26 optimal radiomics features were selected,with
              DCE2_original_shape_SurfaceVolumeRatio ranking first in terms of weight. Among the single ⁃ parameter models,the DCE2 model
              showed the best prognostic efficacy(AUC values of 0.907 and 0.879 for the training and test sets,respectively). Among the combined
              models,the combined with DCE model had better predictive performance than the models without DCE(all P values ≤0.001). The


             [基金项目] 国家自然科学基金(81501442)
              ∗
              通信作者(Corresponding author),E⁃mail: jyn_njmu@163.com
   75   76   77   78   79   80   81   82   83   84   85