中国组织工程研究 ›› 2023, Vol. 27 ›› Issue (2): 287-292.doi: 10.12307/2022.1016

• 组织构建与生物力学 tissue construction and biomechanics • 上一篇    下一篇

血管性高危因素预测血管性轻度认知障碍:支持向量机模型构建及应用

张  倩,卞敏洁,何  琴,黄东锋   

  1. 中山大学附属第七医院,广东省深圳市  518000
  • 收稿日期:2022-01-05 接受日期:2022-02-11 出版日期:2023-01-18 发布日期:2022-06-20
  • 通讯作者: 黄东锋,硕士,主任医师,教授,中山大学附属第七医院,广东省深圳市 518000
  • 作者简介:张倩,女,1989年生,内蒙古自治区包头市人,汉族,2021年广州医科大学毕业,硕士,医师,主要从事神经康复方面的研究。
  • 基金资助:
    中山大学临床医学研究5010计划(2014001),项目负责人:黄东锋

Predicting vascular mild cognitive impairment based on vascular risk factors: construction and application of a support vector machine model

Zhang Qian, Bian Minjie, He Qin, Huang Dongfeng   

  1. The Seventh Affiliated Hospital, Sun Yat-sen University, Shenzhen 518000, Guangdong Province, China
  • Received:2022-01-05 Accepted:2022-02-11 Online:2023-01-18 Published:2022-06-20
  • Contact: Huang Dongfeng, Master, Chief physician, Professor, The Seventh Affiliated Hospital, Sun Yat-sen University, Shenzhen 518000, Guangdong Province, China
  • About author:Zhang Qian, Master, Physician, The Seventh Affiliated Hospital, Sun Yat-sen University, Shenzhen 518000, Guangdong Province, China
  • Supported by:
    the Clinical Medicine Research Program 5010 of Sun Yat-sen University, No. 2014001 (to HDF)

摘要:

文题释义:
血管性认知障碍:是指由脑血管相关疾病导致的不同程度的认知功能障碍,与血管性高危因素(如高血压、糖尿病、高脂血症等)相关,从主观认知下降、轻度认知障碍到痴呆的一大类综合征。早期仅表现为轻度的记忆力下降或行为异常,很容易被患者及家属忽视。有研究表明,早期针对血管性高危因素或认知障碍进行干预可延缓疾病进展,甚至部分患者可恢复正常。
支持向量机:是一类按监督学习方式对数据进行二元分类的分类器,其决策边界是在不同数据类型之间找到一个最大边距的超平面。支持向量机主要用于解决数据的分类或回归问题,例如图像识别、文本数据的分类,其也是适用于小样本量研究的一种机器学习方法。

背景:随着人口老龄化的不断进展,轻度认知障碍患者逐步增多。越来越多的证据表明,血管性高危因素与血管性认知障碍有显著的相关性。因此,血管性高危因素也可作为识别和预测血管性认知障碍的方法之一。
目的:探索以血管性高危因素构建的支持向量机模型在识别血管性轻度认知障碍中的临床价值,以期作为一种简易的筛查工具应用于基层机构、社区以及居家康复中。
方法:纳入研究的受试者行认知功能评估,根据评估结果分为正常组、血管性轻度认知障碍组和痴呆组;同时采用方差分析筛选出3组间有统计学差异的血管性高危因素,从而构建血管性轻度认知障碍筛查的支持向量机模型,使用敏感度、特异度、阳性预测值、阴性预测值及曲线下面积评估模型的预测性能。
结果与结论:①符合纳入标准的80例受试者中,根据认知功能评估结果分为正常组(39例)、血管性轻度认知障碍组(24例)和痴呆组(17例);②组间单因素方差分析结果发现收缩压、空腹血糖、总胆固醇、低密度脂蛋白、血尿酸5个血管性高危因素组间存在统计学意义(P < 0.05);③故用上述5个因素构建模型,结果显示筛查血管性轻度认知障碍的灵敏度为0.845 3,特异度为0.919 4,阳性预测值为0.818 0,阴性预测值为0.932 7,受试者工作曲线结果显示曲线下面积为0.892 3;④提示基于常规体检项目中血管性高危因素构建支持向量机模型在血管性轻度认知障碍筛查中具有较高的辨别效能,且其简单易行,可作为一种筛查的方法应用于基层机构、社区以及居家康复中。
缩略语:血管性轻度认知障碍:vascular mild cognitive impairment,VaMCI;北京版蒙特利尔认知评估量表:Montreal Cognitive Assessment scale,MoCA;临床痴呆评定量表:clinical dementia rating scale,CDR

https://orcid.org/0000-0002-5228-6851(张倩) 

中国组织工程研究杂志出版内容重点:组织构建;骨细胞;软骨细胞;细胞培养;成纤维细胞;血管内皮细胞;骨质疏松;组织工程

关键词: 血管性高危因素, 血管性轻度认知障碍, 痴呆, 支持向量机, 机器学习

Abstract: BACKGROUND: With the aging of the population, the number of patients with mild cognitive impairment is gradually increased. Increasing evidence has shown that vascular cognitive impairment has a significant correlation with vascular risk factors. Therefore, vascular risk factors can be used to identify and predict vascular cognitive impairment. 
OBJECTIVE: To explore the clinical value of support vector machine model in the recognition of vascular mild cognitive impairment (VaMCI) based on vascular risk factors, which is expected to be used as a screening tool in grassroots institutions, communities and home rehabilitation.
METHODS: Participants enrolled in the study were assessed for cognitive function and then divided into three groups: normal group, VaMCI group, and dementia group. Vascular risk factors with statistically significant differences were selected by analysis of variance among the three groups. The factors mentioned above were used to build the support vector machine model for screening VaMCI. Sensitivity, specificity, positive predictive value, negative predictive value, and area under curve were used to evaluate the predictive performance of the model.
RESULTS AND CONCLUSION: In the study, 80 participants enrolled were assessed for cognitive function and then divided into three groups: normal group (39 cases), VaMCI group (24 cases), and dementia group (17 cases). Systolic blood pressure, fasting blood glucose, total cholesterol, low-density lipoprotein and blood uric acid were observed to be significantly different among the three groups (P < 0.05) and they were therefore used to construct the prediction model. The sensitivity, specificity, positive predictive value, and negative predictive value of screening for VaMCI were 0.845 3, 0.919 4, 0.818 0, and 0.932 7, respectively and the area under the receiver operating curve was 0.892 3. To conclude, the support vector machine model has high diagnostic value in VaMCI screening based on vascular risk factors, and it is simple and easy to operate and can be applied to grassroots institutions, communities and home rehabilitation.

Key words: vascular risk factor, vascular mild cognitive impairment, dementia, support vector machines, machine learning

中图分类号: