基于高分辨MRI的影像组学技术识别颈动脉不稳定斑块的研究  

A Study of High Resolution MRI-based Radiomics to Identify Carotid Unstable Plaque

作　　者：刘汉卿 朱世元 王东琳 李林坤[3] 于文文[3] LIU Han-qing;ZHU Shi-yuan;WANG Dong-lin;LI Lin-kun;YU Wen-wen(Affiliated Hosptial of Shandong Second Medical University,Weifang 261035,Shandong Province,China;School of Medical Imaging,Shandong Second Medical University,Weifang 261053,Shandong Province,China;Department of radiology,Weifang People's Hospital,Weifang 261041,Shandong Province,China)

机构地区：[1]山东第二医科大学附属医院,山东潍坊261035 [2]山东第二医科大学医学影像学院,山东潍坊261053 [3]山东省潍坊市人民医院放射科,山东潍坊261041

出　　处：《中国CT和MRI杂志》2025年第2期50-52,59,共4页Chinese Journal of CT and MRI

基　　金：潍坊市卫生健康委员会科研项目(WFWSJK-2023-196)。

摘　　要：目的探讨影像组学在颈动脉高分辨MR诊断不稳定斑块中的可行性。方法回顾性分析由我院临床诊断为颈动脉狭窄的181例患者的临床及影像资料。采用3D slicer软件基于增强扫描T1WI序列勾画患者病灶并生成三维感兴趣体积,对每个感兴趣体积提取影像组学特征,用曼惠特尼U检验及LASSO算法进行特征筛选及降维。采用交叉验证按7:3比例划分数据集,在训练集中用随机森林(RF)、支持向量机(SVM)、逻辑回归(LR)和XGBoost分类器构建预测模型,并在验证集中绘制ROC曲线评价模型效能。结果181例患者中,不稳定斑块90例,稳定斑块91例。最终得到14个特征用于模型构建,LR模型、RF模型、SVM模型、XGBoost模型在验证集AUC分别为0.851(95%CI:0.803~0.899)、0.807(95%CI:0.729~0.885)、0.845(95%CI:0.787~0.903)、0.829(95%CI:0.779~0.879),delong检验显示四种模型的ROC曲线无统计学差异(P>0.05)。结论基于高分辨MR影像组学特征构建的模型可有效识别颈动脉不稳定斑块,LR、RF、SVM、XGBoost四种模型中,LR模型的分类效能最佳,为颈动脉不稳定斑块的识别提供了新的方法。Objective To explore the feasibility of radiomics based on high resolution magnetic resonance in the diagnosis of the carotid unstable plaque.Methods A retrospective analysis was conducted on the clinical and imaging data of 181 patients with carotid artery stenosis diagnosed by our hospital.The 3D slicer software was used to outline the lesions and generate 3D volumes of interest based on the T1WI sequences of the enhanced scans,and the radiomics features were extracted from each volume of interest which underwent feature selection and dimensionality reduction by Mann-Whitney U test and LASSO algorithm.The data set was divided in the proportion 7:3 by Cross-validation,and the prediction models were constructed by four classifiers including random forest(RF),support vector machine(SVM),logistic regression(LR),and XGBoost in the training set,and the ROC curves were drawn in the test set to evaluate the model efficiency.Results Among the 181 patients,90 were unstable plaques and 91 were stable plaques.Finally,fourteen features were used for models'construction,the AUC of the logistic regression model,RF model,SVM model,XGBoost model in the validation set were 0.851(95%CI:0.803-0.899),0.807(95%CI:0.729-0.885),0.845(95%CI:0.787-0.903),0.829(95%CI:0.779-0.879),there was no significant difference in the ROC curves of the four models by the Delong test(P>0.05).Conclusion Models based on high-resolution MR radiomics features can effectively identify unstable plaques in carotid arteries,and the LR model has the best classification performance among the four models of LR,RF,SVM and XGBoost,which provides a new method for the identification of carotid artery unstable plaque.

关 键 词：颈动脉粥样硬化 颈动脉不稳定斑块 磁共振成像 影像组学 

分 类 号：R445.2[医药卫生—影像医学与核医学]