基于无创参数辨识急性呼吸窘迫综合征的相关算法研究进展

摘要

介绍了急性呼吸窘迫综合征(acute respiratory distress syndrome,ARDS)的定义及发病机制,回顾了近年来国内外提出的使用无创参数辨识ARDS的方法(包括回归方法、非线性拟合方法、多参数方法),同时总结了各种方法在临床试验中的效果以及各自的优点和不足.指出了今后构建模型时可加入多种无创参数,并尝试使用神经网络、支持向量机和决策树等机器学习算法,进一步提高基于无创参数估测PaO2/FiO2的能力,在无需血气分析的情况下,基于无创参数实现对ARDS的快速诊断和实时监测.%The concept and pathogenesis of acute respiratory distress syndrome (ARDS)were introduced,and the methods for identifying ARDS based on noninvasive parameters in recent years were retrospectively reviewed including regression method, nonlinear fitting method and multi parameter method.The above methods had their advantages and disadvantages summarized. It's suggested that multi noninvasive parameters and machine learning algorithms such as neural network, support vector machine and decision tree be involved in model construction to promote PaO2/FiO2assessment based on noninvasive parameters,so that the rapid diagnosis and real-time monitoring of ARDS can be realized based on noninvasive parameters while there were no need for blood gas analysis.