Identification algorithm of disease severity in patients with acute respiratory distress syndrome based on ensemble learning
10.19745/j.1003-8868.2025021
- VernacularTitle:基于集成学习的急性呼吸窘迫综合征患者疾病严重程度辨识算法研究
- Author:
Peng-cheng YANG
1
;
Xin SHAO
;
Chun-chen WANG
;
Kun BAO
;
Yang ZHANG
;
Shi-chen DU
;
Hai-feng XU
Author Information
1. 新疆军区总医院信息科,乌鲁木齐 830001
- Publication Type:Journal Article
- Keywords:
acute respiratory distress syndrome;
ensemble learning;
disease severity;
identification of disease severity;
machine learning
- From:
Chinese Medical Equipment Journal
2025;46(2):1-9
- CountryChina
- Language:Chinese
-
Abstract:
Objective To propose a novel identification algorithm based on ensemble learning for assessing the severity of acute respiratory distress syndrome(ARDS)to achieve continuous monitoring of the disease severity.Methods Firstly,leve-raging the open-source MIMIC-Ⅳ database,a variety of non-invasive physiological parameters of patients were extracted and subjected to preliminary preprocessing.A multivariate feature selection algorithm was employed to rank these parameters and calculate feature importance scores through weighted computation.Secondly,based on the feature importance scores,a subset search algorithm was utilized to identify the subset of features that could yield optimal performance across four machine learning algorithms:neural networks,logistic regression,AdaBoost and XGBoost.Finally,a soft voting ensemble method was designed using a generalized linear regression model to integrate the results of each single machine learning algorithm,and a multivariate ensemble learning algorithm was proposed by combining the optimal feature subsets.The algorithm proposed when used to identify the severity of ADRS was evaluated with MIMIC-Ⅳ database,and compared with the traditional algorithms.Results The sensitivity,specificity,accuracy and AUC of the algorithm were 87.15％,89.23％,88.34％and 0.923 4,respectively,all of which outperformed those of the traditional algorithms.Conclusion The ARDS severity identification algorithm based on ensemble learning is capable of achieving continuous and real-time monitoring of the severity of ARDS,thereby offering robust support for the early identification and warning of ARDS in patients.[Chinese Medical Equipment Journal,2025,46(2):1-9]
