摘要目的 研究一种适合于健康管理系统及基层社区预防保健中使用的无创型发病危险评估模型的建立方法.方法 通过无创地检测人体的系统表型属性,并通过有示教的机器学习优化算法选用基于流行病学的前瞻性发病危险评估模型为示教标准开发出研制成本较低,且易于推广应用的无创型评估模型的通用建立方法;并以有创型国人缺血性心血管病发病危险评估模型作为示教标准对照,研制出相应的无创采样24点人体经络穴位皮肤表面电阻值的评估模型作为例证.结果 例证采用4606个实际体检临床案例构建模型,在合理的综合评价指标下,无创型模型与标准源模型的平均外推符合率始终保持在90%以上.结论 该模型方法可应用于常见病系列的无创型发病危险评估模型开发.

abstractsObjective To develop a method for nou-invasive risk model that can be used in health management system and primary healthcare centers. Methods Based on systematic phenotype measurement and modified superviser-based machine learning, an applicable and low-cost method for non-invasive disease risk prediction model was developed. Invasive risk assessment model for Chinese isehemic cardiovascular diseases (ICVD) was used as a supervised standard reference to provide evidence based non-invasive model with skin resistance on 24 acupeints. Results A total of 4606 cases from a physical examination system were collected to develop the model. For those reasonable evaluation indeies, the coincidence of non-invasive with invasive risk prediction model was > 90% on average level. Conclusions This method could be used to develop non-invasive risk assessment model for the common diseases.