基於心率變異分析之睡眠階段評估指標建立

Abstract

由於身體大部分修復和再生的工作，都發生在睡眠中，若欠缺適量的睡眠，將有可能導致一些後遺症，如嗜睡症、憂鬱症、記憶力下降、性功能障礙、心臟病、中風、糖尿病、癌症…等，因此有效率的監測睡眠情況是必須的；然而，想要完整地監測睡眠情況，目前只能透過部分設立在醫院中的睡眠中心所提供的睡眠多項生理檢查得知，但檢測除了價錢昂貴以及排隊民眾眾多導致等待時間延長之外，最大的缺點是必須監測大量的生理訊號，例如: 腦電圖、心跳、血壓、眼電圖、肌電圖…等，造成受試者因穿戴過多的監測裝置，伴隨而來的不舒適感，進而可能影響睡眠監測結果的正確性，因此，本研究透過提出的睡眠階段評估指標演算法，從心電圖中擷取出心率變異訊號，並且利用心率變異訊號具有代表自主神經系統活性的特性，進而求得睡眠評估指標來觀察睡眠的階段，達到有效率監測睡眠的目的以及扮演提供一個參考指標的輔助角色。 本研究中使用PhysioNet網站所提供的資料庫來進行驗證，彙整總共32筆資料，其中包括患有睡眠呼吸中止症的19位病患以及13位健康者，也將睡眠階段分為三大類，分別為清醒、淺眠以及熟睡，其中淺眠包含睡眠階段一和睡眠階段二；熟睡則為睡眠階段三和睡眠階段四，來檢視本研究所提出的睡眠評估指標與各個睡眠階段的關聯性。研究結果顯示: 睡眠評估指標與三大睡眠階段呈現正相關性，當人從清醒階段進入到熟睡期時，睡眠評估指標也會隨之成長，並且使用無母數統計分析檢測，本研究所提出的睡眠評估指標在三大睡眠階段中，皆具有顯著差異 (p < 0.05)，顯示本研究所提出的睡眠評估指標是可視為檢測清醒階段、淺眠期與熟睡期的一個輔助工具。此外，為了更嚴謹的檢視本研究開發的睡眠評估指標的可信度，將其與高頻心率變異訊號做比對，高頻心率變異訊號已經被許多研究證實為可用於判斷睡眠的指標，結果顯示睡眠評估指標具有與高頻心率變異訊號相同的評估效用 (p < 0.05)，且睡眠評估指標與高頻心率變異訊號在清醒階段呈現中度相關 (r = 0.6)，在淺眠期與熟睡期皆呈現高度相關 (r > 0.9)。 另一方面，本研究針對不同病症的族群做睡眠階段檢視，將睡眠評估指標分別應用於睡眠呼吸中止症病患、嗜睡症病患與健康人中，研究結果顯示: 睡眠階段評估指標能夠清楚地檢測健康人與睡眠呼吸中止症病患的清醒階段、淺眠期與熟睡期 (p < 0.05)，但是對於嗜睡症病患僅能夠檢測清醒與睡著兩階段，沒辦法分辨淺眠期或熟睡期，本研究推測由於愛普渥斯嗜睡度量表屬於病人自主評估表，評估結果會因個人主觀意識而有落差，因此可能影響本研究所提出的睡眠階段評估指標的評估結果。

Sleeping well brings good quality of life, because human body does most of its repair and regeneration work in sleeping. Without sufficient sleep for a long time, many diseases may occur, such as, hypersomnia, melancholia, memory decrease, sexual dysfunction, cardiovascular diseases, stroke, diabetes, and cancers. Thus, it is necessary and helpful to have a better understanding of sleep. Sleep stages are largely monitored and determined by polysomnography (PSG). The PSG is generally administrated by sleep centers in large hospitals, such as the National Taiwan University Hospital. Not only is taking PSG expensive, but also is the waiting line long (patients generally have to wait for two or three months). Another major disadvantage is that subjects have to wear many sensors to collect vital signs, such as electrocardiography (ECG), electroencephalography (EEG), electromyography (EMG), and blood pressure. This may lead to some misleading results caused by uncomfortable factors (e.g. getting nervous in the hospital and wearing many sensors). The ECG is a basic vital sign. Related to sleep stages, heart rate variability (HRV), can represent the parasympathetic activities of the autonomic nervous system (ANS). Therefore, this research creates an algorithm to extract the HRV from the ECG signals to acquiring the sleep stage assessment index (SSAI). Finally, the SSAI is used to determine sleep stages using the relationship between the SSAI and sleep stages. This research also utilizes the data from the PhysioNet database to verify the HRV algorithm and the process of calculating SSAI. The physical data of 32 subjects (19 subjects with sleep apnea and 13 healthy subjects) are drawn from the database. Then, the data are divided into three phases: wake, light sleep (sleep stage 1 and stage 2), and deep sleep (sleep stage 3 and stage 4) according to hypnogram. The relationship between SSAI and sleep stags is explored through analyzing the data from PhysioNet. It is found that the SSAI is positively correlated with the three sleep phases (wake, light sleep, and deep sleep). The SSAI increases when people enter the phase of deep sleep from the phase of wake. Additionally, a Wilcoxon non-parametric statistical test is employed to determine the usefulness of the SSAI. In conclusion, the SSAI is proven to be a good reference index to inspecting sleep stages (p < 0.05). The SSAI is compared with the high frequency of HRV (HF) which has been verified as a sleep stage assessment index to examine the reliability of SSAI. The results show that SSAI could serve as a sleep stage assessment index, like HF. The correlation between SSAI and HF is moderate in wake (r = 0.6) and high in light sleep and deep sleep (r > 0.9). Moreover, SSAI is applied to healthy people, patients with sleep apnea and patients with hypersomnia. It finds that SSAI can successfully determine three phases of sleep for healthy people and patients with sleep apnea (p < 0.05). Moreover, SSAI can determine wake and sleep for the patients with hypersomnia. However, SSAI scores of light sleep and deep sleep are undifferentiated in patients with hypersomnia, because the Epworth Sleepiness Score (ESS) is a self-appraisal questionnaire. Therefore, the results might be affected by personal opinions.