應用腦波能量分布於清醒期與睡眠第三期之辨識

Identification of Wake and NREM-3 Stages Using Energy Distributions of Sleep EEG

阮智宇(Zhi-Yu Ruan)

指導教授 ： 劉佩玲

共同指導教授 ： 江秉穎(Ping-Ying Chiang)

國立臺灣大學/工學院/應用力學研究所/碩士(2013年)

https://doi.org/10.6342/NTU.2013.02222

摘要

本研究之目的在發展清醒期與睡眠階段自動辨識之方法，採用最新的AASM規則，並使用時頻分析方法和規劃睡眠階段判讀的決策樹，以符合睡眠技師人工解讀訊號的程序來辨識睡眠階段。針對以往研究常忽略的動作期偵測以及清醒期和睡眠第三期之辨識需計算特徵占整頁時間比例之規則，本研究將整頁時間切割為一秒一個片段，並逐秒計算睡眠階段辨識所需要之頻帶能量與頻帶振幅等參數，進而求出特徵占整頁時間之比例。
本研究以功率頻譜計算睡眠腦波與肌電訊號，偵測雜訊過多之動作期。而清醒期與睡眠第三期則使用時頻分析方法 WVD ( Wigner - Ville Distribution )，由WVD時頻圖之能量分布，將頻率劃分為δ波(0.5∼3.5 Hz)、θ波(4∼7.5 Hz)、α波(8∼12 Hz)、σ:波(12.5~16 Hz)和β波(16.5∼30 Hz)五條頻帶並逐秒計算頻帶能量。最後經由比較每秒各頻帶能量大小之方法來偵測頻率為α波或更高頻的訊號特徵，並計算特徵占整頁時間的比例來辨識睡眠清醒期。以及計算每秒δ頻帶的振幅參數，設定閥值來偵測高幅慢波之特徵，並計算特徵占整頁時間的比例來辨識睡眠第三期。最後依仿照人工判讀之決策樹，將整晚的睡眠階段依動作期和非動作期分類為睡眠清醒期、睡眠第三期與其他睡眠階段三個類別。
本研究的受試者為二十位健康成年人，分別為七位男性與十三位女性，年齡在20到55歲之間。將所有受試者自動分類的睡眠階段與睡眠技師判讀的睡眠階段做比較，得到睡眠清醒期的敏感度為81.08%，睡眠第三期的敏感度為80.19%，其他睡眠階段的特異度為89.90%，整體精確率達86.34%，都有相當高的準確率。此外，將本研究之結果進行信度檢驗得到之Kappa值達0.749，介於0.6~0.8之間，代表本研究的自動判讀結果與睡眠技師人工判讀結果有高度的吻合度。

關鍵字

睡眠腦波 ； 睡眠階段 ； 自動判讀 ； 時頻分析 ； 維格納分佈

並列摘要

This study develops an automatic identification of Wake and NREM-3 stages based on AASM rules. In order to follow the procedure of the artificial sleep stage scoring of sleep technicians, this study uses the time-frequency analysis and the decision tree to identify the sleep stages. Since the formerly studies often ignored the detection of MT and the rules that the identification of Wake and NREM-3 needs to count what percentage of the marks account for one epoch, this study uses an one-second window to get the feature coefficients to find what percentage of the marks account for one epoch.
At first, this study uses the EEG and EMG channel to detect the MT using the method of power spectral density. And then using the EEG signal’s time-frequency density function of Wigner - Ville Distribution to compute the band- power of five frequency bands (delta: 0.5-3.5 Hz, theta: 4-7.5Hz, alpha: 8-12Hz, sigma: 12.5-16Hz, and beta: 16.5-30Hz ) and the summation of amplitude of delta band by one-second window to find the marks of Wake and NREM-3. At last, using the decision tree form the artificial process of sleep stage scoring to classify the sleep stages into Wake, NREM-3 and the others.
In this study, there are twenty healthy adults (seven males and thirteen females, age: 34.1±11.7 years). The results of auto-classification compared to those of human expert’s are sensitivity of 81.08% for Wake, sensitivity of 80.19% for N3, and specificity of 89.90% for the other sleep stages. Besides the accuracy is very high up to 86.34%. And the Kappa coefficient of agreement is 0.749 which means that this study has a substantial agreement with experts.

並列關鍵字

sleep EEG ； sleep stage ； automatic identification ； time-frequency analysis ； Wigner-Ville distribution

參考文獻

Acharya RU, Faust O, Kannathal N, Chua T, Laxminarayan S., “Non-linear analysis of EEG signals at various sleep stages,” Comput Meth Programs Biomed, 2005; 80: p. 37-45.

Chapotot, F., Becq, G., “Automated sleep-wake staging combining robust feature extraction, artificial neural network classification, and flexible decision rules,” Int J Adapt Control Signal Process 2010; 24: p. 409-23.

Costa, M., Goldberger, A.L., Peng, C.K., “Multiscale entropy analysis of complex physiological time series,” Phys. Rev. Lett., 2002.

Fava, C., Montagnana, M., Favaloro, E.J., Guidi, G.C., Lippi, G., “Obstructive Sleep Apnea Syndrome and Cardiovascular Diseases,” Semin. Thromb. Hemost., 2011; 37: p. 280-297.

Fraiwan L, Lweesy K, Khasawneh N, Wenz H, Dickhaus H., 2011, Automated sleep stage identification system based on time-frequency analysis of a single EEG channel and random forest classifier. Comput Methods Programs Biomed., 2012, 108, 10–19.