山东大学耳鼻喉眼学报 ›› 2023, Vol. 37 ›› Issue (6): 126-132.doi: 10.6040/j.issn.1673-3770.0.2023.168

• 临床研究 • 上一篇    下一篇

阻塞性睡眠呼吸暂停儿童睡眠结构研究及睡眠结构判读模型建立

张一彤1,李青香2,石争浩2,尚磊3,袁钰淇1,曹子讷1,麻莉娜1,刘海琴1,任晓勇1,施叶雯1   

  1. 1. 西安交通大学第二附属医院 耳鼻咽喉头颈外科, 陕西 西安 710004;
    2. 西安理工大学 计算机科学与工程学院, 陕西 西安 710048;
    3. 空军军医大学军事预防医学院 卫生统计学教研室, 陕西 西安 710038
  • 发布日期:2023-12-15
  • 通讯作者: 施叶雯. E-mail:shiyewen8813@126.com
  • 基金资助:
    国家自然科学基金项目(62076198)

The sleep structure of Children with obstructive sleep apnea and the development of a sleep structure interpretation model

ZHANG Yitong1, LI Qingxiang2, SHI Zhenghao2, SHANG Lei3, YUAN Yuqi1, CAO Zine1, MA Lina1, LIU Haiqin1, REN Xiaoyong1, SHI Yewen1   

  1. 1. Department of Otorhinolaryngology & Head and Neck Surgery, The Second Affiliated Hospital of Xi'an Jiaotong University, Xi'an 710004, Shaanxi, China2. School of Computer Science and Engineering, Xi'an University of Technology, Xi'an 710048, Shaanxi, China3. Department of Health Statistics, School of Public Health, Air Force Medical University, Xi'an 710038, Shaanxi, China
  • Published:2023-12-15

PDF (PC)

299

摘要： 目的 讨论不同程度阻塞性睡眠呼吸暂停(obstructive sleep apnea, OSA)儿童睡眠结构的改变,并基于深度学习方法构建睡眠结构自动判读模型,提高睡眠结构判读的精准度和效率。 方法 回顾性分析146例学龄期打鼾儿童的多导睡眠监测(polysomnography, PSG)结果,比较不同严重程度OSA儿童呼吸事件、血氧饱和度和睡眠结构等参数的差异。基于深度学习方法,以脑电和眼电联合通道作为输入信号,由卷积神经网络(convolutional neural networks, CNN)、注意力机制和循环神经网络(recurrent neural network, RNN)组成混合神经网络,构建睡眠结构自动判读模型,并于公开数据集和私人数据集进行模型性能评估。 结果 单纯打鼾组、轻度OSA组和中重度OSA组间快动眼(rapid eye movement, REM)期百分比和非快动眼睡眠(non-rapid eye movement, NREM)1期百分比比较差异未见统计学意义(P>0.05)。中重度OSA组N2期百分比明显高于单纯打鼾组、轻度OSA组(P=0.024),中重度OSA组N3期百分比明显低于单纯打鼾组、轻度OSA组(P<0.001),中重度OSA组觉醒指数(total arousal index, ARtotI)和呼吸事件相关觉醒指数(respiratory arousal index, RAI)明显高于单纯打鼾组、轻度OSA组(P<0.001),差异均具有统计学意义。本研究提出的睡眠结构自动判读模型,在公开数据集上总体准确率为79.90%,在N1期、N2期、N3期及R期的F1值分别为45.90%、85.40%、83.70%和79.90%,在私人数据集上总体准确率为71.30%,在N1期、N2期、N3期及R期的F1值分别为40.20%、81.60%、77.70%和67.40%。 结论 OSA儿童随疾病进展,睡眠结构出现N2期延长和N3期缩短,本研究提出的基于深度学习的睡眠结构自动判读模型,具有良好的分类效果和稳定性。

关键词: 阻塞性睡眠呼吸暂停, 儿童, 睡眠结构, 人工智能, 自动判读模型

Abstract: Objective To explore the sleep structure in children with different degrees of obstructive sleep apnea(OSA)and construct an automatic sleep structure interpretation model using deep learning. Methods We retrospectively analyzed the polysomnography(PSG)results of 146 school-aged children, and compared the parameters, such as respiratory events, oxygen saturation, and sleep architecture in children with different severity of OSA. Using deep learning, a hybrid neural network composed of convolutional neural networks(CNN), attention mechanism, and recurrent neural network(RNN)was used to construct an automatic sleep structure interpretation model. A combination of electroencephalogram and electrooculogram were used as input signals. The model’s performance was evaluated using public and private databases. Results There was no significant difference in the percentage of rapid eye movement(REM)and the percentage of non-rapid eye movement(NREM)stage 1 among the primary snoring, mild OSA, and moderate to severe OSA groups(P>0.05). The percentage of the N2 stage was significantly higher in the moderate-severe OSA group than in the primary snoring and the mild OSA groups(P=0.024). The percentage of the N3 stage was significantly lower in the moderate-severe OSA group than in the primary snoring and the mild OSA groups(P<0.001). The total arousal index(ARtotI)and respiratory arousal index(RAI)in the moderate-severe OSA group were significantly lower than those in the primary snoring and the mild OSA groups(P<0.001). Using the public database, the automatic sleep structure interpretation model proposed in this study had an overall accuracy of 79.90%, and F1 values of 45.90%, 85.40%, 83.70%, and 79.90% in N1, N2, N3, and R stages, respectively. Using the private database, the overall accuracy was 71.30%, and the F1 values were 40.20%、81.60%、77.70%, and 67.40% in N1, N2, N3, and R stages, respectively. Conclusion The sleep structure of children with moderate-to-severe OSA showed a prolonged NREM2 and a shortened NREM3. Overall, the deep learning-based automatic sleep structure interpretation model demonstrated stability and a good classification effect.

Key words: Obstructive sleep apnea, Children, Sleep structure, Artificial intelligence, Deep learning model

中图分类号: 

  • R766.4
[1] Spruyt K, Capdevila OS, Serpero LD, et al. Dietary and physical activity patterns in children with obstructive sleep apnea[J]. J Pediatr, 2010, 156(5): 724-730.e3. doi:10.1016/j.jpeds.2009.11.010
[2] 王岩. 儿童睡眠呼吸疾病的分类与治疗进展[J]. 山东大学耳鼻喉眼学报, 2016, 30(5): 5-8. doi: 10.6040/j.issn.1673-3770.0.2016.405 WANG Yan. Progress in the classification and treatment of sleep respiratory diseases in children[J]. China Industrial Economics, 2016, 30(5): 5-8. doi: 10.6040/j.issn.1673-3770.0.2016.405
[3] 刘大波. 重视儿童阻塞性睡眠呼吸暂停低通气综合征睡眠结构紊乱[J]. 山东大学耳鼻喉眼学报, 2018, 32(2): 6-8. doi: 10.6040/j.issn.1673-3770.0.2017.534 LIU Dabo. Evaluation of sleep structure disorder in children with obstructive sleep apnea hypopnea syndrome[J]. Journal of Otolaryngology and Ophthalmology of Shandong University, 2018, 32(2): 6-8. doi: 10.6040/j.issn.1673-3770.0.2017.534
[4] Berry RB, Brooks R, Gamaldo C, et al. AASM scoring manual updates for 2017(version 2.4)[J]. J Clin Sleep Med, 2017, 13(5): 665-666. doi:10.5664/jcsm.6576
[5] 中国儿童OSA诊断与治疗指南制订工作组, 中华医学会耳鼻咽喉头颈外科学分会小儿学组, 中华医学会儿科学分会呼吸学组, 等. 中国儿童阻塞性睡眠呼吸暂停诊断与治疗指南(2020)[J]. 中华耳鼻咽喉头颈外科杂志, 2020, 55(8): 729-747. doi:10.3760/cma.j.cn115330-20200521-00431 Working Group of Chinese Guideline for the Diagnosis and Treatment of Childhood OSA; Subspecialty Group of Pediatrics, Society of Otorhinolaryngology Head and Neck Surgery, Chinese Medical Association; Subspecialty Group of Respiratory Diseases, Society of Pediatrics, Chinese Medical Association, et al. Chinese guideline for the diagnosis and treatment of childhood obstructive sleep apnea(2020)[J]. Chin J Otorhinolaryngol Head Neck Surg, 2020, 55(8): 729-747. doi:10.3760/cma.j.cn115330-20200521-00431
[6] Craik A, He YT, Contreras-Vidal JL. Deep learning for electroencephalogram(EEG)classification tasks: a review[J]. J Neural Eng, 2019, 16(3): 031001. doi:10.1088/1741-2552/ab0ab5
[7] Yu Y, Si XS, Hu CH, et al. A review of recurrent neural networks: LSTM cells and network architectures[J]. Neural Comput, 2019, 31(7): 1235-1270. doi:10.1162/neco_a_01199
[8] Goldberger AL, Amaral LA, Glass L, et al. PhysioBank, PhysioToolkit, and PhysioNet: components of a new research resource for complex physiologic signals[J]. Circulation, 2000, 101(23): E215-E220. doi:10.1161/01.cir.101.23.e215
[9] Qu W, Wang ZY, Hong H, et al. A residual based attention model for EEG based sleep staging[J]. IEEE J Biomed Health Inform, 2020, 24(10): 2833-2843. doi:10.1109/jbhi.2020.2978004
[10] Supratak A, Dong H, Wu C, et al. DeepSleepNet: a model for automatic sleep stage scoring based on raw single-channel EEG[J]. IEEE Trans Neural Syst Rehabil Eng, 2017, 25(11): 1998-2008. doi:10.1109/tnsre.2017.2721116
[11] Humayun AI, Sushmit AS, Hasan T, et al. End-to-end sleep staging with raw single channel EEG using deep residual ConvNets[C] //2019 IEEE EMBS International Conference on Biomedical & Health Informatics(BHI), 2019, 5:1-5. doi:10.1109/BHI.2019.8834483
[12] Andreotti F, Phan H, Cooray N, et al. Multichannel sleep stage classification and transfer learning using convolutional neural networks[J]. Annu Int Conf IEEE Eng Med Biol Soc, 2018: 171-174. doi:10.1109/EMBC.2018.8512214
[13] Phan H, Andreotti F, Cooray N, et al. Joint classification and prediction CNN framework for automatic sleep stage classification[J]. IEEE Trans Biomed Eng, 2019, 66(5): 1285-1296. doi:10.1109/tbme.2018.2872652
[14] Babkoff H, Caspy T, Mikulincer M, et al. Monotonic and rhythmic influences: a challenge for sleep deprivation research[J]. Psychol Bull, 1991, 109(3): 411-428. doi:10.1037/0033-2909.109.3.411
[15] Garpestad E, Ringler J, Parker JA, et al. Sleep stage influences the hemodynamic response to obstructive apneas[J]. Am J Respir Crit Care Med, 1995, 152(1): 199-203. doi:10.1164/ajrccm.152.1.7599824
[16] Durdik P, Sujanska A, Suroviakova S, et al. Sleep architecture in children with common phenotype of obstructive sleep apnea[J]. J Clin Sleep Med, 2018, 14(1): 9-14. doi:10.5664/jcsm.6868
[17] Heubi CH, Knollman P, Wiley S, et al. Sleep architecture in children with down syndrome with and without obstructive sleep apnea[J] Otolaryngol Head Neck Surg, 2021, 164(5): 1108-1115. doi: 10.1177/0194599820960454
[18] 张丰珍, 王桂香, 许志飞, 等. 儿童重度OSAHS睡眠结构及相关因素分析[J]. 临床耳鼻咽喉头颈外科杂志, 2019,(5): 441-446. doi:10.13201/j.issn.1001-1781.2019.05.014 Zhang FZ, Wang GX, Xu ZF, et al. Analysis of sleep structure and related factors in children with severe obstructive sleep apnea-hypopnea syndrome[J]. Journal of Clinical Otorhinolaryngology Head and Neck Surgery, 2019,(5): 441-446. doi:10.13201/j.issn.1001-1781.2019.05.014
[19] Krajca V, Petranek S, Paul K, et al. Automatic detection of sleep stages in neonatal EEG using the structural time profiles[J]. Conf Proc IEEE Eng Med Biol Soc., 2005: 6014-6016. doi:10.1109/iembs.2005.1615862
[20] Tsinalis O, Matthews PM, Guo YK. Automatic sleep stage scoring using time-frequency analysis and stacked sparse autoencoders[J]. Ann Biomed Eng, 2016, 44(5): 1587-1597. doi:10.1007/s10439-015-1444-y
[21] He KM, Zhang XY, Ren SQ, et al. Deep residual learning for image recognition[C] //2016 IEEE Conference on Computer Vision and Pattern Recognition(CVPR). June 27-30, 2016, Las Vegas, NV, USA. IEEE, 2016: 770-778. doi:10.1109/CVPR.2016.90
[22] Begawan IA, Djamal EC, Djajasasmita D, et al. Sleep stage identification based on EEG signals using parallel convolutional neural network and recurrent neural network[C] //2022 International Conference on Advanced Computer Science and Information Systems(ICACSIS). October 1-3, 2022, Depok, Indonesia. IEEE, 2022: 39-44. doi:10.1109/ICACSIS56558.2022.9922962
[1] 朱希倩,王佳,孙祖贤,冯建秀,张梦佳,赵颖,王宏,姜敏敏. 上海市杨浦区2022—2024年6~9岁学龄儿童屈光状态分析[J]. 山东大学耳鼻喉眼学报, 2026, 40(3): 102-109.
[2] 杨冠英,李元彬. 人工智能在干眼管理中的应用进展[J]. 山东大学耳鼻喉眼学报, 2026, 40(3): 115-120.
[3] 朱明琼,李征,刘茹,田涛,彭婧利,吕倩怡,谭华霞. 基于OCT/OCTA的AI筛查系统在抗VEGF治疗糖尿病性黄斑水肿患者效果评价中的应用[J]. 山东大学耳鼻喉眼学报, 2026, 40(1): 68-73.
[4] 程卓, 梁辉, 邢鲁民. 深度学习技术在咽喉内镜应用中的研究进展及前景分析[J]. 山东大学耳鼻喉眼学报, 2026, 40(1): 112-119.
[5] 宋艳玲,司元元,崔彦. 微量玻璃体切除治疗激光笔致儿童全层黄斑裂孔1例并文献复习[J]. 山东大学耳鼻喉眼学报, 2025, 39(6): 144-147.
[6] 熊琴, 张砚, 乌日娜, 李锋, 唐力行. 鼻用糖皮质激素在儿童中的应用[J]. 山东大学耳鼻喉眼学报, 2025, 39(6): 160-167.
[7] 刘南仙,杨泽垠,韩琳,张爱英,赵宇亮,薛静,孙怡君,邵永良. 视频脑电图在儿童复发性眩晕诊断中的意义[J]. 山东大学耳鼻喉眼学报, 2025, 39(5): 20-25.
[8] 黄焕,华红利,邓玉琴,江承洋,王雨薇,杨星海. 儿童过敏性鼻炎、扁桃体腺样体肥大和鼻窦炎之间相关性及其对临床指导价值[J]. 山东大学耳鼻喉眼学报, 2025, 39(5): 34-41.
[9] 王华,张丰珍,龙婷,赵靖,李宏彬,王生才,王桂香. 后颅窝肿瘤术后儿童气管切开原因及预后转归分析[J]. 山东大学耳鼻喉眼学报, 2025, 39(4): 168-173.
[10] 乐冰艳,邹剑,雷蕾,文巧,钱应雪. 儿童扁桃体微生物群与免疫调节及疾病关联[J]. 山东大学耳鼻喉眼学报, 2025, 39(4): 193-200.
[11] 卫志成,彭裕,沈力,李莉琳,沈杭东,李馨仪,许华俊,关建. 犬尿氨酸介导阻塞性睡眠呼吸暂停所致肝功能损伤:一项横断面研究[J]. 山东大学耳鼻喉眼学报, 2025, 39(3): 38-44.
[12] 马馨,董凌康,吴红敏,易红良,邹建银. 外展悬吊缝合技术在阻塞性睡眠呼吸暂停治疗中的改良及应用进展[J]. 山东大学耳鼻喉眼学报, 2025, 39(3): 81-88.
[13] 李莉琳,李馨仪,关建. 阻塞性睡眠呼吸暂停与抑郁症共病机制的研究进展[J]. 山东大学耳鼻喉眼学报, 2025, 39(3): 89-96.
[14] 黄爱萍,王娟,王丽,耿江桥,王亚芳,温鑫. 儿童原发扁桃体Burkitt淋巴瘤累及上颌骨和肺1例并文献复习[J]. 山东大学耳鼻喉眼学报, 2025, 39(3): 148-152.
[15] 张国明,魏文斌,林浩添,迟玮,张少冲,赵培泉,雷柏英,陈有信,王雨生,何明光,梁建宏,卢海,陆方,黄欣,梁小玲,赵欣予,吴桢泉,余震,崔凯璇,刘亚玲,项道满,陈长征,张自峰,林铎儒,于珊珊,孙悦,檀韬,陈燕先,彭婕,董力,程湧,朱雪梅,杨鹏,陈少滨. 人工智能技术辅助早产儿视网膜病变诊疗专家共识(2025)[J]. 山东大学耳鼻喉眼学报, 2025, 39(2): 1-5.
Viewed
Full text


Abstract

Cited
  Shared   
  Discussed   
No Suggested Reading articles found!
网站版权所有 © 2018 《山东大学耳鼻喉眼学报》编辑部　
地址：山东省济南市山大南路27号（250100）电话：+86-0531-88366912 E-mail: ebhxb@sdu.edu.cn
本系统由北京玛格泰克科技发展有限公司设计开发