医学
急性呼吸窘迫
重症监护医学
算法
呼吸窘迫
计算机科学
内科学
麻醉
肺
作者
Berrin Er,Ahmet Görkem Er,Sema Turan
标识
DOI:10.1016/j.jcrc.2024.154581
摘要
Machine learning (ML) approaches were used in many fields in critically ill patients with acute respiratory distress syndrome (ARDS). From diagnosis to outcome prediction, we aimed to review ML methods in ARDS. We conducted a systematic search for English language papers in PubMed Central, Ovid MEDLINE, Ovid Embase databases using the keywords; acute respiratory distress syndrome, machine learning and artificial intelligence. We followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses Guidelines. Two reviewers performed study selection. Reviews, protocols, commentaries, letter to editors, editorials, conference proceedings, animal studies, and the studies not related to ARDS were excluded. A total of 83 of 135 studies between 2002 and March 2023 were analyzed. The studies were categorized based on which supervised and unsupervised methods were used. Classification task was applied as a supervised method in 75 (90%), and unsupervised clustering was applied in 8 (10%) studies. A number of ML tools, including decision trees, logistic regression, support vector machine, gradient boosting and Bayesian networks were used for classification. Of them, 52 were related to diagnosis and ARDS risk prediction, 12 were related to treatment and mechanical ventilation and 11 were related to prognosis and outcome prediction. Clustering was used to classify the patients into different subtypes, compare the response to treatments and determine mechanical ventilation management strategies. 18 studies were using deep learning algorithms which were commonly used for the segmentation of radiological images. In recent years, the number of studies using ML tools for management of ARDS has rapidly increased. However, in real life scenarios, their usage is sparse. Additional studies are needed to identify the ideal algorithms or combination of techniques for ARDS management and to implement state-of-the-art tools in the clinic.
科研通智能强力驱动
Strongly Powered by AbleSci AI