数字乳腺体层合成影像组学列线图预测乳腺浸润性导管癌Ki-67表达水平  

Radiomics nomogram based on DBT for predicting expression level of Ki-67 in breast invasive ductal carcinoma

作　　者：顾梦瑶 陈琪 刘丽 戚佳佳 李万湖 GU Mengyao;CHEN Qi;LIU Li;QI Jiajia;LI Wanhu(Department of Radiology,Shandong Cancer Hospital and Institute,Shandong First Medical University,Shandong Academy of Medical Sciences,Jinan 250117,China)

机构地区：[1]山东省肿瘤防治研究院(山东省肿瘤医院)山东第一医科大学(山东省医学科学院)影像科,山东济南250117

出　　处：《中国医学影像技术》2025年第3期429-433,共5页Chinese Journal of Medical Imaging Technology

基　　金：吴阶平医学基金会项目(320.6750.2023-18-111)。

摘　　要：目的观察数字乳腺体层合成(DBT)影像组学列线图预测乳腺浸润性导管癌(BIDC)Ki-67表达水平的价值。方法回顾性分析374例BIDC,根据Ki-67表达将其分为高表达组(n=224)与低表达组(n=150);根据7∶3比例划分训练集(n=271,含高表达组162例、低表达组109例)及测试集(n=103,含高表达组62例、低表达组41例)。对比组间临床特征及影像学表现,提取并筛选病灶影像组学特征;分别以8种分类器构建影像学模型及影像组学模型,选择其中最佳者;基于影像学特征及影像组学评分构建列线图模型。绘制受试制工作特征曲线,计算曲线下面积(AUC),评估各模型预测BIDC Ki-67表达水平的效能。结果组间病灶最大径、边缘有无毛刺差异均有统计学意义(P均<0.05),以之构建的多层感知机(MLP)影像学模型在训练集及测试集的AUC分别为0.654、0.715;基于8个影像组学特征构建的MLP影像组学模型在训练集和测试集的AUC分别为0.802、0.806,而列线图模型的AUC分别为0.802、0.806。结论基于DBT影像组学列线图能有效预测BIDC Ki-67表达水平。Objective To observe the value of digital breast tomosynthesis(DBT)-based radiomics nomogram for predicting Ki-67 expression levels in breast invasive ductal carcinoma(BIDC).Methods Data of 374 cases of BIDC were retrospectively analyzed and divided into high-expression group(n=224)and low-expression group(n=150)according to expression of Ki-67 as well as training set(n=271,162 cases in high-expression group and 109 cases in low-expression group)and test set(n=103,62 cases in high-expression group and 41 cases in low-expression group)at the ratio of 7∶3.Clinical characteristics and lesion s image manifestations were compared between groups,and radiomic features were extracted and filtered.Then imaging radiomics models were constructed with 8 classifiers,respectively,and the optimal classifier was selected.A nomogram model was subsequently developed through integrating image features and radiomics scores.Receiver operating characteristic curves were drawn,and the area under the curves(AUC)were calculated to evaluate the performance of the above models.Results Significant differences of the maximum diameter and spiculated margin of the lesions were found between groups(both P<0.05),and AUC of multi-layer perceptron(MLP)image model constructed based on these indexes for predicting BIDC expression level of Ki-67 was 0.654 in training set and 0.715 in test set,of MLP radiomics model constructed based on 8 radiomics features was 0.802 in training set and 0.806 in test set,while of the nomogram model constructed based on image features and radiomics scores was 0.802 in training set and 0.806 in test set,respectively.Conclusion DBT-based radiomics nomogram could be used to effectively predict Ki-67 expression levels in BIDC.

关 键 词：乳腺肿瘤 乳腺X射线摄影 影像组学 KI-67 

分 类 号：R737.9[医药卫生—肿瘤] R817[医药卫生—临床医学]