临床特征及量表在阻塞性睡眠呼吸暂停低通气综合征患者初筛中的价值  

Value of clinical characteristics and scale in preliminary screening of obstructive sleep apnea hypopnea syndrome patients

作　　者：张莹 翟振强 吴东亚 王德奎 冯筠[2] Zhang Ying;Zhai Zhenqiang;Wu Dongya;Wang Dekui;Feng Jun(Department of Neurology,Xi′an International Medical Center Hospital,Xi′an 710117,China;School of Information Science and Technology,Northwest University,Xi′an 710127,China)

机构地区：[1]西安国际医学中心医院神经内科,西安710117 [2]西北大学信息科学与技术学院,西安710127

出　　处：《国际呼吸杂志》2023年第10期1180-1186,共7页International Journal of Respiration

基　　金：国家自然科学基金(62073260)。

摘　　要：目的探讨常用临床特征及量表在阻塞性睡眠呼吸暂停低通气综合征(OSAHS)患者初筛中的价值,寻找提升OSAHS初筛效果的人工智能算法,优化临床诊断思路及辅助检查决策。方法本研究为病例对照研究。采用非随机抽样法,选取2019年10月至2022年10月在西安国际医学中心医院进行多导睡眠监测(PSG)的465例患者为研究对象,根据PSG结论分为OSAHS组(349例)和非OSAHS睡眠障碍组(116例),根据OSAHS严重程度分为OSAHS亚组(轻度、中度和重度)。比较OSAHS组以及OSAHS亚组患者的临床特征;分析不同人工智能分类算法在OSAHS初筛中的性能。结果在OSAHS组、非OSAHS组和OSAHS亚组中,OSAHS组与非OSAHS组相比,年龄、BMI、颈围以及Epworth嗜睡评分均升高,差异均有统计学意义(均P<0.05),OSAHS亚组三组之间相比,性别、BMI、颈围以及Epworth嗜睡评分差异均有统计学意义(均P<0.05)。在OSAHS亚组中,女性BMI和颈围之间具有中等程度相关性(r=0.67,P<0.01),而在男性中具有弱相关性(r=0.36,P<0.01)。多特征联合在OSAHS初筛任务中准确率达到81%,优于使用单特征预测。同时,使用人工智能算法支持向量机、线性判别分析、逻辑回归、最近邻结点算法进行OSAHS初筛的准确率分别为77.08%、79.17%、81.25%和89.58%。结论性别、BMI、颈围对OSAHS初筛具有较好的预测效果,Epworth嗜睡量表评分及年龄预测效果较低。几种人工智能算法中,最近邻结点算法相比支持向量机、线性判别分析、逻辑回归算法具有更好的性能,同时优于传统基于统计的OSAHS方法。中国西北地区人群在更低的BMI值有罹患OSAHS的可能,建议对女性患者采用更低的颈围测量标准,综合评估OSAHS的可能诊断,进而选择简易睡眠监测,提高诊断效率,节约医疗成本。Objective To explore the value of common clinical features and scales in the preliminary screening of obstructive sleep apnea hypopnea syndrome(OSAHS)patients,to find artificial intelligence algorithms to improve the preliminary screening effect of OSAHS,and to optimize clinical diagnosis thinking and to assist examination and decision making.Methods This was a case-control study.Using non-random sampling,a total of 465 patients who underwent polysomography(PSG)in Xi′an International Medical Center Hospital from October 2019 to October 2022 were studied.According to PSG results,they were divided into OSAHS group(349 cases)and non-OSAHS sleep disorder group(116 cases).Based on OSAHS severity,the OSAHS group were further divided into OSAHS subgroups(mild,moderate,and severe).The clinical characteristics of patients in OSAHS group and in OSAHS subgroup were compared.The performance of different AI classification algorithms in OSAHS preliminary screening was analyzed.Results Among OSAHS group,non-OSAHS group,and OSAHS subgroups,compared with the non-OSAHS group,the OSAHS group showed a significant increase in age,BMI,neck circumference,and Epworth drowsiness score,with statistical significance(all P<0.05).There were statistically significant differences in gender,BMI,neck circumference,and Epworth score among the three groups of OSAHS subgroups(all P<0.05).In the OSAHS subgroup,there was a moderate correlation between BMI and neck circumference in females(r=0.67,P<0.01),while there was a weak correlation in males(r=0.36,P<0.01).The accuracy of multi feature combination in the initial screening task of OSAHS reached 81%,which was better than using single feature prediction.Meanwhile,the accuracy of using artificial intelligence algorithms such as Support Vector Machine,Linear Discriminant Analysis,Logistic Regression,and Nearest Neighbor Node Algorithm for initial screening of OSAHS was 77.08%,79.17%,81.25%,and 89.58%,respectively.Conclusions Gender,BMI,and neck circumference show good predictive effect on OSAHS pre

关 键 词：呼吸暂停 阻塞性睡眠呼吸暂停低通气综合征 EPWORTH嗜睡量表 Stop-bang量表 多导睡眠监测 

分 类 号：R766[医药卫生—耳鼻咽喉科]