人工神经网络模型在急性应激障碍预警中的应用  

Application of artificial neural network in early warning system for acute stress disorder screening

作　　者：侯艳红[1] 张林[3] 陈晓菲[1] 张颖[1] 齐秦甲子[1] 徐燕杰[2] 

机构地区：[1]北京解放军第309医院心理科,100091 [2]北京解放军第309医院医务科,100091 [3]北京解放军第309医院消化科,100091

出　　处：《中国急救复苏与灾害医学杂志》2015年第4期340-343,共4页China Journal of Emergency Resuscitation and Disaster Medicine

基　　金：军队心理卫生科研课题（12XLZ331）

摘　　要：目的探讨人工神经网络模型在对急性应激障碍预警中的应用。方法通过现场流行病学整群抽样调查获取研究对象及有关信息；急性应激障碍确诊根据中国精神疾病分类（CCMD-3）诊断标准，并参照国际疾病分类第10版（ICD-10）相关内容。采用个性指标，认知评价，应对方式，社会支持，情绪指标，植物神经功能评定等指标；数据库建立采用SPSSl7．0软件，建立神经网络模型（ANN）。结果积累从2008年1月～2012年12月军队及地方突发事件应对人员，及医院门急诊采集病例近1000人，确诊急性应急障碍患者97人，患病率为9％。从被研究者中抽取急性应激障碍患者和非急性应激障碍者各97名为建模对象，将其情绪因子、性格指标、认知指标、植物神经指标等7个变量作为网络的输入层，进行ANN的拟合。所建模型的预测精度为96％；能够正确预测建模对象中95．3％的非急性应激障碍患者，95．1％的急性应激障碍患者；输入变量敏感性系数排在前四位的依次为焦虑素质、认知功能、应对能力和植物神经功能。结论ANN在急性应激障碍（预测）中精度较高，具有一定的开发应用前景。Objective To explore an artificial neural network model in the application of early warning system for Acute Stress Disorder （ASD） screening in the community. Methods The subjects and related information were obtained by Field Epidemiology cluster sample; ASD was diagnosed by Chinese Classification of Mental Disorders （CCMD-3） diagnostic criteria and International Classification of Diseases 10th edition （ICD-10）; Using indicators of personality, cognitive appraisal, coping styles, social support, sentiment index, evaluating indicators such as plant nerve function; SPSS13.0 software was adopted to establish the database, and the Artificial Neural Network （ANN） was established. Results Accumulation from 2008 to 2012, the army and local emergency responders, and hospital screened the nearly 1000 people, 97 of them were diagnosed acute stress disorder （9%）. such as emotion factor, character index, cognitive index, index of plant nerve, 7 variables as input layer of network, fitting for ANN. The model prediction accuracy is 95%; correctly forecasting modeling objects in 95.3% of patients with acute stress disorder, and 95.1% of the patients with non acute stress disorder. According to the sensitivity coefficient, the top four were anxiety character, cognitive function, response ability and plant nerve function. Conclusion ANN carries a high accuracy in ASD screening （prediction） in community which has wide application prospect.

关 键 词：人工神经网络模型 急性应激障碍 数学模型 预警 

分 类 号：R395.6[哲学宗教—心理学]