Computer Engineering and Applications ›› 2021, Vol. 57 ›› Issue (19): 156-163.DOI: 10.3778/j.issn.1002-8331.2006-0205

Previous Articles     Next Articles

Analytical Model for Correlations Between Indicators and Critical Illness

ZHANG Lige, CHEN Yuwen, QIN Xiaolin, YI Bin, LI Yujie   

  1. 1.Chengdu Institute of Computer Applications, Chinese Academy of Sciences, Chengdu 610041, China
    2.University of Chinese Academy of Sciences, Beijing 100049, China
    3.Chongqing Institute of Green and Intelligent Technology, Chinese Academy of Sciences, Chongqing 400714, China
    4.The First Affiliated Hospital, Army Medical University, Chongqing 400038, China
  • Online:2021-10-01 Published:2021-09-29

危重症指标相关性分析模型

张力戈,陈芋文,秦小林,易斌,李雨捷   

  1. 1.中国科学院 成都计算机应用研究所,成都 610041
    2.中国科学院大学,北京 100049
    3.中国科学院 重庆绿色智能技术研究院,重庆 400714
    4.陆军军医大学 第一附属医院,重庆 400038

Abstract:

The selection of preoperative and intraoperative indicators affects the real-time capability and accuracy of prediction for critical illness. At present, there are so many kinds of preoperative and intraoperative indicators that it is difficult to find the potential relationship between them and critical illness. Aiming at the correlations between indicators of patients and critical illness, an analysis model based on machine learning is proposed. The model combines statistical method and Spearman’s rank correlation coefficient to remove redundant indicators. Then, the XGBoost model is used to analyze the contribution of each indicator in predicting critical illness and the contribution is taken as the correlations between each indicator and critical illness. Finally, key indicators of critical illness are selected according to the correlations between each indicator and critical illness. Preoperative and intraoperative data of liver failure and renal failure are used to verify the model. The results show that the model can effectively analyze the correlations between indicators and critical illness, and the key indicators extracted by this model are slightly more effective than all indicators in the prediction of critical illness.

Key words: prediction of critical illness, Spearman’s rank correlation coefficient, XGBoost, analysis of indicators

摘要:

患者术前与术中检测指标的选择影响危重症预测的实时性和准确性。目前患者术前和术中检测指标种类繁多,很难找到它们与危重症之间的潜在联系。针对患者检测指标与危重症之间的关联,提出了基于机器学习的危重症核心指标分析模型。模型通过统计方法与斯皮尔曼等级相关系数去除冗余指标,结合XGBoost模型分析各指标对危重症风险预测的贡献,以此作为各指标与危重症之间的相关性,并提取对应危重症核心指标。选取肝衰与肾衰患者数据对模型进行实验验证,结果表明,该模型能有效分析指标与危重症之间的相关性,提取的核心指标在危重症预测中的效果略高于全部指标。

关键词: 危重症预测, 斯皮尔曼等级相关系数, XGBoost, 指标分析