【摘要】 慢性阻塞性肺疾病（COPD）是一种与人体肺部及呼吸道相关的常见慢性疾病,严重威胁着人类健康。氧减状态的准确辨识对于诊断COPD具有重要的临床指导意义。基于受试人员的6分钟步行数据,本文展开了COPD氧减状态的辨识研究,对每个待辨识点进行特征表示,包括血脉氧饱和度指数、脉搏、血脉氧饱和度指数的窗口特征、梯度特征以及耦合特征,在上述特征表示的基础上,使用数据集来训练用于氧减状态辨识的受监督自组织神经网络模型。严格的对比实验结果表明：本文所提出的辨识模型优于现有的其他方法,全局性能指标AUC达到了0.861 1,可以有效用于氧减状态的辨识,对于COPD的诊断具有重要的参考价值。更多还原

【Abstract】 Chronic obstructive pulmonary disease, also known as COPD, is a common chronic disease related to human lungs and respiratory tract, which is a serious threat to human health. The accurate identification of oxygen depletion status has important clinical significance for the diagnosis of COPD. Used to train a supervised selforganizing map model for ODS identification; Thein this study, we perform research on ODS identification based on the six minutes walking data of the recruited subjects. Firstly, five features, including saturation of pulse oxygen（SpO₂）value, pulse rate value, window feature, gradient feature and correlation feature of SpO₂, are extracted and combined to form discriminative feature for each target point; Secondly, on the basis of the feature representation, the training set is experimental results show that the global performance index of the proposed identification model reaches 0.861 1 of AUC, indicating that the proposed method can be effectively used for identifying ODS and could facilitate the diagnosis of COPD.更多还原