上海交通大学学报(医学版) ›› 2019, Vol. 39 ›› Issue (8): 908-.doi: 10.3969/j.issn.1674-8115.2019.08.017

• 论著·公共卫生 • 上一篇    下一篇

基于机器学习的轻度认知功能障碍筛查研究

贾芷莹 1, 2,董旻晔 1, 2,施贞夙 2, 3,金春林 4,李国红 1, 2   

  1. 1.上海交通大学医学院公共卫生学院,上海 200025;2.上海交通大学中国医院发展研究院卫生技术评估研究所,上海 200025;3.上海唯晶信息科技有限公司,上海 200025;4. 上海市卫生与健康发展研究中心,上海 200040
  • 出版日期:2019-08-28 发布日期:2019-09-23
  • 通讯作者: 李国红,电子信箱:guohongli@sjtu.edu.cn。
  • 作者简介:贾芷莹(1993—),女,硕士生;电子信箱: kame@sjtu.edu.cn。
  • 基金资助:
    教育部哲学社科重大公关项目(18JZD040);上海市第四轮公共卫生三年行动计划重点学科建设项目循证公共卫生与卫生经济学(15GWZK0901)

Study of a screening system for mild cognitive impairment based on machine learning model

JIA Zhi-ying1, 2, DONG Min-ye1, 2, SHI Zhen-su2, 3, JIN Chun-lin4, LI Guo-hong1, 2   

  1. 1. Shanghai Jiao Tong University School of Public Health, Shanghai 200025, China; 2. Center for HTA, China Hospital Development Institute, Shanghai Jiao Tong University, Shanghai 200025, China; 3.Winking Entertainment Corporation, Shanghai 200025, China; 4. Shanghai Health and Health Development Research Center, Shanghai 200040, China
  • Online:2019-08-28 Published:2019-09-23
  • Supported by:
    Major Key Research Project in Philosophy and Social Sciences of the Ministry of Education of P. R. China, 18JZD040; Shanghai Fourth Round Public Health Three-Year Action Plan Key Discipline Construction Project of Evidence-based Public Health and Health Economics, 15GWZK0901

摘要: 目的 ·评价项目研制的可用于轻度认知功能障碍筛查的电子化认知评估系统的信度和效度,构建机器学习法判定模型并评估筛查效果。方法 ·采用分层随机的方法在上海和河南农村的社区、老年护理院及专科门诊抽取 55岁以上的符合标准的老年人,由经过严格培训、操作规范的调查员对研究对象进行蒙特利尔认知评估量表( Montreal Cognitive Assessment,MoCA)的现场测试。电子化认知评估系统信度评价采用内部一致性系数,效度评价采用因子分析;以 MoCA评估结果作为标准,使用分类准确率和曲线下面积(area under curve, AUC)比较朴素贝叶斯、随机森林、 Logistic回归和 K-邻近 4种机器学习算法的分类效果。结果 ·研究的 359名对象中,年龄中位数为 63岁,82.80%为中学及以下学历;根据 MoCA评分,可能患有轻度认知功能障碍的有 147名。电子化认知评估系统的 Cronbachs α为 0.84,KMO为 0.78,Bartletts球形检验 P<0.05,共提取 13个公因子,累计方差贡献率为 75.10%。最优朴素贝叶斯分类模型的分类准确率为 88.05%,AUC为 0.941。结论 ·该电子化认知评估系统具有良好的信度、效度及分类效果,利用朴素贝叶斯分类模型分类准确度较高。

关键词: 轻度认知功能障碍, 电子化认知评估系统, 机器学习, 朴素贝叶斯分类模型, 蒙特利尔认知评估量表, 筛查

Abstract: Objective · To evaluate the reliability and validity of a computerized cognitive assessment system designed for screening mild cognitive impairment (MCI), and compare the screening accuracy among constructed different machine learning classification models. Methods · A group of random stratified samples of over 55 years old residents in the communities, nursing homes and memory-clinics Shanghai and Henan were selected to assess their cognitive status using Montreal Cognitive Assessment (MoCA)well-trained investigators. The reliability and validity were assessedintrinsic consistency analysis and factor analysis, respectively. Taking the results of MoCA as standards, four machine learning classification algorithms, i.e., na.ve Bayesian classification model, random forest classifier, Logistic regression classifier, and K-nearest neighbor classifier, were compared in accuracy and area under curve (AUC). Results · A total of 359 participants were included, the median age of whom was 63 years old. And 82.80% of them were secondary school graduates or below. According to the results of MoCA, 147 of them might be MCI. The Cronbachs α and KMO of this system were 0.84 and 0.78, respectively; Bartletts sphericity test was significant (P<0.05); thirteen common factors could explain 75.10% of the system. The best classification model was na.ve Bayesian classification model, and its accuracy andAUC were 88.05% and 0.941, respectively. Conclusion · The new designed computerized cognitive assessment system has been proved to be reliable and valid. The na.ve Bayesian classification model has good classification accuracy.

Key words: mild cognitive impairment (MCI), computerized cognitive assessment system, machine learning, na.ve Bayesian classification model, Montreal Cognitive Assessment (MoCA), screening

中图分类号: