摘要
目的探讨基于T_(2)WI和DCE-MRI的影像组学模型预测浸润性乳腺癌组织学分级的价值。方法回顾性分析浸润性乳腺癌132例患者的MRI增强图像。采用ITK-SNAP逐层勾画肿瘤边界,利用AK软件分别提取T_(2)WI、DCE、T_(2)WI和DCE联合序列三维病灶的影像组学特征,并通过特征选择分别选取了3、2、3个影像组学特征,分别采用Logistic回归、随机森林(RF)、支持向量机(SVM)和决策树(decision tree)不同的机器算法建立模型,采用受试者工作特征(ROC)曲线评估模型的预测效能,通过决策曲线(decision curve analysis,DCA)评估各模型的临床应用价值。结果T_(2)WI和DCE-MRI联合序列采用决策树算法预测效能最佳,均优于Logistic回归、RF和SVM算法模型,且决策树算法模型与Logistic、RF、SVM算法模型的AUC值之间差异具有统计学意义(P<0.001)。在决策曲线分析中,采用决策树算法模型的净收益高于采用Logistic回归、RF和SVM算法模型。结论MRI影像组模型可以无创预测浸润性乳腺癌组织学分级,T_(2)WI和DCE-MRI联合序列采用决策树算法模型预测性能最佳,具有一定的临床价值。
Objective To explore the predictive value of radiomics model based on T_(2)WI and DCE-MRI in evaluation the histological grade of invasive breast cancer.Methods A total of 132 patients with invasive breast cancer pathologically confirmed underwent T_(2)WI and DCE-MRI examination.The ITK-SNAP software was used to delineate tumor boundaries layer by layer on the T_(2)WI,DCE,and T_(2)WI combined with DCE sequences images.The radiomics features were extracted based on the AK software,and 3,2 and 3 effective features were obtained.The models were established by different machine algorithms of Logistic Regression,Random Forest(RF),support vector machine(SVM)and decision tree.Receiver-operating characteristic(ROC)curve was drawn to analyze the predict efficiency and decision curve analysis(DCA)was used to evaluate the clinical application value of different sequence radiomic models.Results The decision tree model based on T_(2)WI combined with DCE had the highest predictive performance,and difference of AUC among the 4 groups had statistical significance(P<0.001).In the analysis of decision curve,the net benefit of decision tree algorithm model was higher than Logistic regression,RF and SVM.Conclusion The radiomics model based on MRI could noninvasively predict the histological grade of invasive breast cancer.The decision tree algorithm model based on the combined sequence of T_(2)WI and DCE-MRI had the best prediction performance and had a certain clinical value.
作者
李明
李薇
吴林桦
平学军
石惠
LI Ming;LI Wei;WU Linhua;PING Xuejun;SHI Hui(Department of Medical Imaging,Affiliated Hospital of Shaanxi University of Traditional Chinese Medicine,Xianyang 712000,China;Ningxia Medical University,Yinchuan 750004,China;Department of Radiology,Baoji Central Hospital,Baoji 721008,China;Department of Radiology,the General Hospital of Ningxia Medical University,First Clinical Medical College of Ningxia Medical University,Yinchuan 750004,China)
出处
《宁夏医科大学学报》
2023年第9期904-908,923,共6页
Journal of Ningxia Medical University
基金
宁夏自然科学基金项目(2021AAC03368)。
关键词
影像组学
浸润性乳腺癌
组织学分级
磁共振成像
机器算法
radiomics
invasive breast cancer
histological grade
magnetic resonance imaging
machine algorithm