- 學位論文
運用生理特徵於睡眠呼吸中止症嚴重性之診斷
Sleep Apnea Syndrome Detection Based on Multiple Physiological Signals
摘要
現代人因工作與生活的壓力大導致睡眠品質每況愈下，而在睡眠品質不良的症狀中，以睡眠呼吸中止症影響生活作息最為嚴重，睡眠呼吸中止症不僅影響自身的生理調節功能失常，也常常導致工作專注力及效率不佳，還有可能在白天工作環境中發生嗜睡的情形，可能因為自身的疏忽，導致不可挽回的悲劇。 此篇研究主要探討如何早期發現睡眠呼吸中止症，並協助病患能夠提早預防，以免病情日益惡化，而傳統的睡眠呼吸中止症病患大多使用多通道睡眠圖(Polysomnogram, PSG)來進行檢測，然而使用PSG檢測需要專業的睡眠治療師來進行睡眠呼吸中止症評估及判斷，且檢測需要花費較長的時間，設備相當昂貴，加上醫院內睡眠中心的床位有限，通常需要等待安排檢測時間較久無法快速進行檢查。因此，本研究發展一快速、客觀且低成本的睡眠呼吸中止症篩檢與預測模式，預測病症發生的可能性與嚴重程度，及早發現及早治療，降低發生的機率並避免病症惡化。本研究將採用多種生理特徵結合機器學習方法進行睡眠呼吸中止症嚴重程度的篩檢模型建立，並評估不同機器學習方法檢測睡眠呼吸中止症病況的成效，採用模糊C均值(Fuzzy C mean, FCM)演算法建構睡眠呼吸中止症的預測模型，輔助醫師達到臨床及早發現病徵的效果。此外，在實驗過程中，研究發現睡眠呼吸中止症的嚴重患者會有呼吸中止導致清醒的自救情形，而只仰賴SpO2與睡眠中每小時無呼吸與呼吸減弱次數(AHI)會導致誤判產生，誤判睡眠呼吸中止導致清醒的患者為輕微的患者，本研究腦波可以正確反映當中的睡眠呼吸中止症特徵並不會因此消失，並且本研究發現與睡眠呼吸中止症相關的新型腦波特徵。此外，在實驗過程中，研究發現睡眠呼吸中止症的嚴重患者會有因呼吸中止而導致清醒的自救情形，而只仰賴SpO2與睡眠中每小時無呼吸與呼吸減弱次數(AHI)會導致誤判產生，誤判睡眠呼吸中止導致清醒的患者為輕微的患者，而本研究發現新型態的腦波特徵可以正確反映嚴重患者因呼吸中止而清醒的情況，解決因呼吸中止造成的誤判情形，協助醫師臨床診斷。
並列摘要
The pressures of modern working life have caused deterioration in the sleep quality of people. Among sleep disorders, sleep apnea affects our everyday life the most. It not only disturbs physiological regulatory functions, but also often leads to poor focus and lack of efficiency at work. It may even lead to drowsiness during the working day, resulting in irreversible tragedy caused by inattention. This study focused mainly on the early detection of sleep apnea and assisting patients in early prevention, to avoid worsening of their condition. Polysomnogram (PSG) is commonly used as an assessment tool of sleep apnea patients. However, using PSG requires sleep study specialists to conduct the assessment and diagnosis. PSG is also time-consuming and the equipment is fairly expensive. Furthermore, due to limited hospital beds in sleep centers, patients usually need to wait a long time to be scheduled for a sleep study, and consequently cannot undergo the testing promptly. Therefore, this study developed a fast, objective, and cost-effective sleep apnea screening and prediction model to predict the possibility of the occurrence and severity of illness. This model also provides early assessment and treatment to the patients, thereby reducing the incidence and preventing the deterioration of the disease. The study used a variety of physiological characteristics in combination with machine learning to establish a screening model for the severity of sleep apnea, and to evaluate the effectiveness of different machine learning methods in detecting this disorder. Fuzzy C mean (FCM) algorithm was used to construct the prevention model for sleep apnea, which would assist doctors in achieving the goal of early identification of symptoms. In addition, during the experiment, the researchers found that patients with severe sleep apnea would have apnea that resulted in self-rescuing awakening. Thus, relying solely on peripheral capillary oxygen saturation (SpO2) and apnea-hypopnea index (AHI) would result in false diagnosis of patients with apnea-induced awakening as having mild symptoms. This study showed that altered brainwave activity can accurately reflect the features of sleep apnea and is not affected by apnea-induced awakening. It also found new brainwave features associated with sleep apnea.
參考文獻
延伸閱讀
- 鄭景元、林嘉謨（2014）。睡眠呼吸中止症的內科治療。台灣醫學，18(2)，204-208。https://doi.org/10.6320/FJM.2014.18(2).10
- 徐逸民（2010）。探討睡眠障礙患者睡眠呼吸中止嚴重度及嗜睡狀態〔碩士論文，亞洲大學〕。華藝線上圖書館。https://www.airitilibrary.com/Article/Detail?DocID=U0118-1511201215464920
- 楊聰財（2022）。了解睡眠檢測如何診斷呼吸中止症和失眠。消費者報導雜誌，(500)，42-44。https://www.airitilibrary.com/Article/Detail?DocID=P20180921002-202212-202212050007-202212050007-42-44
- 侯人文（2019）。使用深度神經網路之阻塞型睡眠呼吸中止症病人之自動化睡眠分期判讀〔碩士論文，國立臺灣大學〕。華藝線上圖書館。https://doi.org/10.6342/NTU201903265
- 馬素華（2011）。A Brief Introduction to the Obstructive Sleep Apnea Syndrome。臺東大學人文學報，1(2)，119-136。https://doi.org/10.29778/NTUCH.201112.0004