基于BPNN的手足口病重症化进程中的相关因素变化及预测分析
Application of BPNN in the analysis and prediction of relative factors of the aggravation progress of severehand-foot-mouth disease

摘要：目的 探讨BP神经网络(BPNN)模型在儿童手足口病(HFMD)重症化进程预测中的应用价值，为HFMD重症病例早期识别提供参考依据。方法 采用MATLAB 7.0软件对2013年4～6月河南郑州某医院收治的445例HFMD患儿临床资料构建BPNN模型，得出平均影响值(MIV)排序并归一化；从中挑选32例符合重症标准且自发病到重症的病例作为重症组，对照组纳入60例普通病例，以较大MIV值的因素为变量重新整理数据，统计分析得出单个因素水平和综合因素水平在重症化期间的变化趋势，阐述其与HFMD重症早期识别的关系。结果 在HFMD重症化进程中，精神差、颈强直、易惊的变化是先上升，重症当天及前一天平稳且至最高，随后下降；呼吸频率、心率、嗜睡、热程≥3d及血糖水平的变化是先升后降，重症当天最高；手足抖动的变化是先升后降，重症前一天最高；呕吐呈下降趋势；热峰大致呈下降趋势，重症当天略微回升，之后降至正常体温；白细胞计数变化基本不大，均高于正常值范围，在重症次日恢复至正常值水平；综合因素水平是先上升至重症当天并达到最高，之后下降。结论 BPNN模型可用于分析HFMD重症化进程中的相关因素变化，可为重症病例的早期识别提供参考依据。
ABSTRACT: Objective To explore the value of back propagation neural network (BPNN) model applied to predict the aggravation of child hand-foot-mouth disease (HFMD) to provide reference for early identification of severe HFMD. Methods By using the software MATLAB 7.0 we constructed the BPNN model on 455 children with HFMD who were admitted by a hospital in Zhengzhou during April and June 2013. The mean impact values (MIV) were calculated and normalized. Of theses children, 32 who met the criteria of severe HFMD and the onset of symptoms to severe cases served as the severe group, and 60 mild cases were in the control group. The data were rearranged by the large MIV factors, the variation trend between the level of single factor and comprehensive factors were obtained by statistical analysis, and we further expounded on the relationship between this single factor level and early identification of severe HFMD. Results During the severity progress of HFMD, the changes of poor spirits, stiff neck and easy to panic increased first, reached the peak between the day before severity and the severity day, and then decreased. The changes of respiration frequency, heart rate, somnolence, days of high fever ≥ 3 days and blood sugar level increased first and then decreased, until reached the peak on the severity day. The change of hand and foot trembling increased and reached the peak the day before severity. Vomiting was of a downward trend; heat peak change was of an approximately declining trend, but recovered slightly on the severity day, and then dropped to the normal temperature range. The white cell count had little change but was higher than the normal range, and returned to the normal level on the second day after severity. The comprehensive factors level elevated until the severity day and then dropped. Conclusion The BPNN model can be used for the analysis of related factors influencing the aggravation of HFMD, thus providing reference for early identification of severe HFMD