基于受监督自组织映射的慢性阻塞性肺疾病氧减状态辨识  

Identification of oxygen depletion status in chronic obstructive pulmonary disease based on supervised self-organized mapping

作　　者：吴月芳[1] 孙培莉 束鑫 於东军[1] WU Yuefang;SUN Peili;SHU Xin;YU Dongjun(Nanjing University of Science and Technology,Nanjing 210094,Jiangsu Province,China;The First Affiliated Hospital of Nanjing Medical University;School of Computer Science,Jiangsu University of Science and Technology)

机构地区：[1]南京理工大学,南京210094 [2]南京医科大学第一附属医院 [3]江苏科技大学计算机学院

出　　处：《中国数字医学》2024年第3期86-91,共6页China Digital Medicine

基　　金：国家自然科学基金(62372234);江苏省自然科学基金(BK20201304)。

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0.8611,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。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(SpO_(2))value,pulse rate value,window feature,gradient feature and correlation feature of SpO_(2),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 0.8611 of AUC,indicating that the proposed method can be effectively used for identifying ODS and could facilitate the diagnosis of COPD.

关 键 词：慢性阻塞性肺疾病 氧减状态辨识 特征表示 受监督自组织映射 

分 类 号：R197[医药卫生—卫生事业管理] R319[医药卫生—公共卫生与预防医学]