◎模式识别与人工智能⑬危重症指标相关性分析模型张力戈微,陈芋文3,秦小林口,易斌李雨捷4中国科学院成都计算机应用研究所,成都6100411.中国科学院大学,北京100049中国科学院重庆绿色智能技术研究院,重庆4007142.陆军军医大学第一附属医院,.重庆400038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XGBoosi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:指标分析文献标志码:A中图分类号:TP391doi:10.3778/j.issn.1002-8331.2006-0205AnalyticalModelforCorrelationsBetweenIndicatorsandCriticalIllnessZHANGLige'2,CHENYuwen'.QINXiaolin'\YIBin4,LIYujie41.ChengduInstituteofComputerApplications,ChineseAcademyofSciences,Chengdu61(X)41,ChinaUniversityofChineseAcademyofSciences.Beijing100049.China2.ChongqingInstituteofGreenandIntelligentTechnology,ChineseAcademyofSciences,Chongqing400714,ChinaThcFirstAffiliatedHospital.ArmyMedicalUniversity.Chongqing400038.ChinaAbstract:Theselectionofpreoperativeandintraoperativeindicatorsaffectsthereal-timecapabilityandaccuracyofpredictionforcriticalillness.Atpresent,therearesomanykindsofpreoperativeandintraoperativeindicatorsthatitisdifficulttofindthepotentialrelationshipbetweenthemandcriticalillness.Aimingatthecorrelationsbetweenindicatorsofpatientsandcriticalillness,ananalysismodelbasedonmachinelearningisproposed.ThemodelcombinesstatislicalmethodandSpearman'srankcorrelationcoefficienttoremoveredundantindicators.Then,theXGBoostmodelisusedtoanalyzethecontributionofeachindicatorinpredictingcriticalillnessandthecontributionistakenasthecorrelationsbetweeneachindicatorandcriticalillness.Finally,keyindicatorsofcriticalillnessareselectedaccordingtothecorrelationsbetweeneachindicatorandcriticalillness.Preoperativeandintraoperativedataofliverfailureandrenalfailureareusedtoverifythemodel.Theresultsshowthatthemodelcaneffectivelyanalyzethecorrelationsbetweenindicatorsandcriticalillness,and(hekeyindicatorsextractedbythismodelarcslightlymoreeffectivethanallindicatorsinthepredictionofcriticalillness.Keywords:predictionofcriticalillness;SpearmanJsrankcorrelationcoefficient;XGBoost;analysisofindicators围术期患者出现危重症,不仅会增加患者的医疗费Khuri等5研究表明术后30天内发生严重不良事件的患者用,影响患者的康复结果心),甚至会导致患者死亡。中位生存时间减少69%。短期手术并发症的长期后基金项目:国家重点研发计划(2()1XYFC01167()4):重庆市科技创新与应用研发项目(cstc2OI9jscx-msxmXO237):四川省科技计划(20197DZX0005,2()19ZDZX0006,2020JDR0006):陆军军医大学第一附届医院伦理委员会批准项目(KY201936)。作者简介:张力戈(1995-),男,博」研究生CCF学生会员,研究领域为昭特习、优化算法陈判G副研究M,E-mail:chenyuwen@cigit.ac.cn;制林研物,CCF会员:易斌,概;李颂,博士。收稿日期:2020-06-15修回日期:2020-10-23文章编号:1(X)2-8331(2021)19-0156-08果对患者生命健康和生活质量有深远的影响气有效预测危重症风险有助于医生及时诊断和治疗患者,避免药物过度使用,有利于医院资源的合理配置,同时降低患者的痛苦和死亡率。患者的各种监测指标在危重症预测中起着重要作...