基于多模态超声的决策树诊断模型对颈动脉斑块易损性的诊断价值  

The Diagnostic Value of Decision Tree Diagnostic Model Based on Multimodal Ultrasound for the Vulnerability of Carotid Artery Plaques

作　　者：李宁 张树华 唐姗 解婷婷 刘阳 阚艳敏 刘爱东 LI Ning;ZHANG Shuhua;TANG Shan(Department of Ultrasound,Affiliated Hospital of North China University of Technology,Hebei 063000,China)

机构地区：[1]华北理工大学附属医院超声科,唐山063000 [2]天津市天津医院病理科,300000

出　　处：《医学研究杂志》2023年第7期130-135,共6页Journal of Medical Research

基　　金：2020年河北省唐山市科学技术研究与发展计划项目(20150207C)。

摘　　要：目的构建基于多模态超声的缺血性脑卒中决策树诊断模型,为临床对颈动脉易损斑块的早期识别提供新的诊断思路。方法回顾性选取2020年1月~2022年2月于华北理工大学附属医院住院的颈动脉斑块患者113例,其中症状组60例、非症状组53例,所有患者均行常规超声(conventional ultrasound,CUS)、超声造影(contrast-enhanced ultrasound,CEUS)及剪切波弹性(shear wave elasticity,SWE)检查,并记录超声特征。比较两组患者超声特征的差异;分别构建基于CUS、CEUS、SWE及多模态超声的缺血性脑卒中决策树诊断模型,采用10折交叉验证法对诊断模型进行测试,并应用准确性、敏感度、特异性、阳性预测值及阴性预测值对其诊断效能进行评价。结果两组间斑块回声、是否有极低回声区、是否有钙化、斑块中部及远心段增强级别、斑块近心段,中部及远心段杨氏模量值比较,差异均有统计学意义(P均<0.05);分别构建基于CUS、CEUS、SWE及多模态超声的缺血性脑卒中诊断模型,10折交叉验证法显示误判率分别为34.5%、44.2%、26.5%和12.2%;对模型的诊断效能进行评价,其中基于多模态超声所建的缺血性脑卒中决策树诊断模型诊断准确性明显高于其他模型,为92.0%,其敏感度、特异性、阳性预测值和阴性预测值分别为90.7%、94.3%、15.91和0.09。结论应用多模态超声可以更准确地评估颈动脉斑块的易损性;通过构建基于多模态超声的缺血性脑卒中的决策树诊断模型,可以有效提高易损斑块的检出率,斑块中部的超声特征更能体现其稳定性。Objective To construct the decision tree diagnosis model of ischemic stroke based on multimodal ultrasound technology,and to provide a new diagnostic idea for the early identification of vulnerable carotid plaque.Methods A total of 113 patients with carotid plaque hospitalized in Affiliated Hospital of North China University of Technology from January 2020 to February 2022 were retrospectively selected,and they were divided into symptom group(n=60)and non-symptom group(n=53).All the patients underwent conventional ultrasound(CUS),contrast-enhanced ultrasound(CEUS)and shear wave elasticity(SWE),and the ultrasonic parameters were recorded.The differences of ultrasonic characteristics between the two groups were compared.Decision tree diagnosis mode for ischemic stroke based on CUS,CEUS,SWE and multimodal ultrasound were constructed respectively.The diagnostic model was tested by 10-fold cross validation method,and the accuracy,sensitivity,specificity,positive predictive value and negative predictive value were used to evaluate its diagnostic efficacy.Results There were significant differences in plaque echo,very low echo area,calcification,enhancement level of central and distal segments of plaque,Young′s modulus of proximal,central and distal segments of plaque between the two groups(P<0.05).CUS,CEUS,SWE and multimodal ultrasound were used to construct the diagnostic models of ischemic stroke,and 10-fold cross validation method showed that the misjudgment rates were 34.5%,44.2%,26.5% and 12.2%,respectively.The diagnostic efficiency of the model was evaluated,the diagnostic accuracy of the the multimodal ultrasound based decision tree diagnostic model for ischemic stroke was significantly higher than that of other diagnostic models,and its value was 92.0%.The sensitivity,specificity,positive predictive value and negative predictive value were 90.7%,94.3%,15.91 and 0.09,respectively.Conclusion The application of multimodal ultrasound can more accurately evaluate the vulnerability of carotid plaque.By constructing

关 键 词：颈动脉斑块 易损性 决策树 多模态超声 缺血性脑卒中 

分 类 号：R445.1[医药卫生—影像医学与核医学]