摘要
目前,机器学习在重症医学科发展迅速,脓毒症诊疗中的应用更是目前研究的热点。集成算法、决策树、神经网络等监督学习法在患者紧急入院时脓毒症早期诊断与风险评估中具有重要意义,但监督学习算法要求所有数据进行标记,具有困难性;聚类算法和数据降维一类半监督学算法在脓毒症预测、脓毒症预后因素分析等方面应用较为常见,但算法相对简单,在多分类任务的处理上得出结果不理想;半监督学习算法结合了前两者,在现实世界更为实用,鉴于半监督学习算法的数据特征,在脓毒症诊疗的决策支持应用方面有待统筹。作者从脓毒症预测及诊断中常见模型的角度总结不同机器学习模型在脓毒症预测及诊断中的应用并进行综述,以期为国内学者提供参考。
At present,machine learning has developed rapidly in the department of critical care medicine,and the application of sepsis diagnosis and treatment is the focus of current research.Supervised learning methods such as ensemble algorithm,decision tree and neural network are of great significance in providing early diagnosis and risk assessment of sepsis at the time of emergency admission,but supervised learning algorithms require all data to be marked,which is difficult;semi-supervised algorithms such as clustering algorithm and data dimensionality reduction are commonly used in sepsis prediction and sepsis prognostic factor analysis,but the algorithm is relatively simple and the results are not ideal in the processing of multi-classification tasks.The semi-supervised learning algorithm combines the former two and is more practical in the real world.In view of the data characteristics of the semi-supervised learning algorithm,it needs to be coordinated in the decision support application of sepsis diagnosis and treatment.From the perspective of common models in sepsis prediction and diagnosis,this paper summarizes and reviews the application of different machine learning models in sepsis prediction and diagnosis,in order to provide reference for domestic scholars.
作者
付晨菲
梁群
潘郭海容
丛迪迪
赵佳瑶
王龙
FU Chenfei;LIANG Qun;PAN Guohairong;CONG Didi;ZHAO Jiayao;WANG Long(Heilongjiang University of Chinese Medicine,Harbin 150040,Heilongjiang,China;The First Hospital Affiliated to Heilongjiang University of Chinese Medicine,Harbin 150040,Heilongjiang,China)
出处
《辽宁中医药大学学报》
CAS
2024年第2期163-171,共9页
Journal of Liaoning University of Traditional Chinese Medicine
基金
国家中医药管理局中医药应急专项课题(2020ZYLCYJ06-2)
黑龙江省重点研发计划项目(GY2021ZB0198)
黑龙江中医药大学校级科技创新研究平台项目(2018pt06)。
关键词
机器学习
脓毒症
模型
machine learning
sepsis
models