臺灣博碩士論文加值系統

在這爆炸資訊的年代，我們隨時隨地都在接收外部給予的資訊及刺激，睡眠症候群儼然擠上文明病的行列，藉由睡眠階段分析睡眠品質的優劣。睡眠階段是在睡眠醫學研究中的關鍵技術用於判斷睡眠結構，並依據此結構作為其他相關睡眠研究的相關基礎數據，其經常會因睡眠障礙及身心疾病而產生異常。其中以快速動眼期最為重要，若能準確的判斷快速動眼期，則可以推斷各項睡眠階段的相關分析，然而過去睡眠階段皆為人工尋找較為曠日廢時，藉由本文的研究減少尋找需求時間，協助醫護人員在觀察睡眠階段的分布相關變化並且能夠提供有效的資訊。
本研究的資料來自於勝美醫院附設睡眠醫學中心，由513人次進行整夜睡眠量測，選用沒佩戴呼吸器進行睡眠量測的患者作為研究資料，相較於傳統使用腦電圖做為研究樣本的研究方法，本研究使用心率訊號作為研究樣本。
本研究目的在於借重深度學習技術，使用卷積神經網路作為主要架構，輔以Inception等網路結構，搭配心率訊號產生的心跳間距(R-R Interval, RRI)進行快速動眼期的判斷，首先以卷積神經網路進行初步的心率訊號的分析，後續將分類結果經由鄰居法則搭配Inception網路架構以及補償法則針對誤判的結果進行修正，希望能使用此技術搭配較輕易量測到的心率訊號，如穿戴裝置就能夠輕易地量測睡眠狀況給予醫護人員更為專業的分析結果。
目前本文所使用的整夜心率訊號進行睡眠階段判讀結果，可得到快速動眼期與非快速動眼期兩睡眠期的分類效能達92.8%的訓練精度，最佳的kappa值高達73.6%，可以表示有較高的準確性。

Considering the increasing popularity of wearable devices and the importance of rapid eye movement (REM) sleep, this goal of this study is to develop an automatic REM sleep detection method by using heart rate signals.
The proposed approach was developed based on 513 polysomnography (PSG) results provided by a hospital sleep center. To characterize the heart rate signal, the first step of the proposed approach is to extract the RR-intervals from the electrocardiogram (ECG) signals recorded by PSG. By characterizing each epoch of PSG with 128 RR-intervals, this study used a convolutional neural network (CNN) to classify REM sleep epochs and epochs of the remaining sleep stages. After this epoch-based learning stage, with the results of the first CNN for the entire night’s epochs as the inputs, a second CNN was used to improve the REM sleep detection results. By using the results of the second CNN for a number of consecutive epochs as inputs, a machine learning method and an empirical technique were used to try to further improve the REM sleep detection results.
The results are summarized as follows. The sensitivity, specificity, accuracy and the Kappa value of the proposed approach are 75.8%, 96.2%, 92.8%, and73.6%, respectively.

目錄
論文審定書.............................................................................................................i
致謝........................................................................................................................ii
摘要...................................................................................................................... iii
Abstract.................................................................................................................iv
目錄........................................................................................................................v
圖目錄...................................................................................................................ix
表目錄...................................................................................................................xi
第一章 緒論...........................................................................................................1
1.1 前言..........................................................................................................1
1.2 研究目的與方法......................................................................................2
1.3 文獻回顧..................................................................................................3
1.4 小結..........................................................................................................4
1.5 論文架構..................................................................................................4
第二章 睡眠生理訊號及心率訊號.......................................................................5
2.1 睡眠生理狀態..........................................................................................5
2.2 睡眠生理檢查..........................................................................................6
2.3 睡眠分期規則..........................................................................................8
2.4 不同睡眠階段的型態..............................................................................9
2.4.1 清醒期(Stage W) ..........................................................................9
2.4.2 睡眠第一期(Stage 1)..................................................................10
2.4.3 睡眠第二期(Stage 2)..................................................................11
2.4.4 睡眠第三期與第四期(Stage 3、4)............................................12
2.4.5 快速動眼期(Stage REM) ...........................................................13
2.5 心電圖原理............................................................................................14
2.5.1 心電圖 QRS 波形.......................................................................14
2.5.2 心電圖 RR Interval.....................................................................16
第三章 實驗流程與架構.....................................................................................17
3.1 實驗流程設計........................................................................................17
3.2 挑選 ECG 訊號峰值..............................................................................18
3.3 消除心電訊號基線飄移........................................................................18
3.4 以小波轉換偵測 QRS 波......................................................................20
3.5 視窗擷取................................................................................................20
3.6 RRI 訊號擷取........................................................................................21
第四章 資料庫與分類器架構介紹.....................................................................22
4.1 簡述資料庫............................................................................................22
4.1.1 資料庫的篩選.............................................................................22
4.1.2 資料的說明.................................................................................23
4.2 深度學習的硬體及軟體選用................................................................24
4.2.1 電腦硬體使用.............................................................................24
4.2.2 電腦軟體使用.............................................................................24
4.3 機器學習架構........................................................................................25
4.3.1 XGBoost(Extreme Gradient Boosting).......................................25
4.3.2 一位有效編碼(One Hot Encoding)............................................27
4.3.3 K 折交叉驗證(K-fold Cross Validation)....................................28
4.3.4 超參數設定.................................................................................29
4.3.5 缺失值的填補.............................................................................32
4.4 深度學習架構........................................................................................33
4.4.1 卷積神經網路(CNN) .................................................................33
4.4.2 MLP 網路架構 ...........................................................................36
4.4.3 GoogLeNet..................................................................................37
4.4.4 ResNet.........................................................................................39
4.5 改善分類效能方法................................................................................40
4.5.1 整夜睡眠週期全域分析.............................................................40
4.5.2 鄰居法則修正分類效能.............................................................42
4.5.3 補償法則改善分類效能.............................................................44
第五章 快速動眼期自動分類實驗結果.............................................................45
5.1 分類效能指標........................................................................................45
5.2 訓練及預測資料集的建立....................................................................47
5.3 深度學習網路架構設定........................................................................47
5.4 CNN 架構偵測快速動眼期 ..................................................................50
5.5 改善分類效能方法................................................................................51
5.5.1 全域週期訓練修正結果.............................................................51
5.5.2 鄰居法則分類器訓練修正結果.................................................53
5.5.3 補償法則修正結果.....................................................................54
5.6 小結........................................................................................................55
第六章 引申預測及分類器補償修正.................................................................56
6.1 心率訊號的缺失值................................................................................56
6.2 填補缺失值............................................................................................57
6.3 填補缺失值後比較................................................................................58
6.4 比較結果................................................................................................59
6.4.1 CNN 架構偵測快速動眼期 .......................................................59
6.4.2 全域週期訓練修正結果.............................................................60
6.4.3 鄰居法則分類器訓練修正結果.................................................61
6.4.4 補償法則修正結果.....................................................................62
6.5 填補前後數據比對................................................................................62
6.6 小結........................................................................................................64
第七章 討論與未來展望.....................................................................................65
7.1 討論........................................................................................................65
7.2 未來展望................................................................................................66
參考文獻..............................................................................................................67

[1]Steiger, A., & Kimura, M. (2010). Wake and sleep EEG provide biomarkers in depression. Journal of Psychiatric Research, 44(4), 242-252.
[2]Kluge, M., Schüssler, P., Dresler, M., Yassouridis, A., & Steiger, A. (2007). Sleep onset REM periods in obsessive compulsive disorder. Psychiatry research, 152(1), 29–35.
[3]Tandon, R., Lewis, C., Taylor, S. F., Shipley, J. E., DeQuardo, J. R., Jibson, M., & Goldman, M. (1996). Relationship between DST nonsuppression and shortened REM latency in schizophrenia. Biological psychiatry, 40(7), 660–663.
[4]Mellman, T. A., Pigeon, W. R., Nowell, P. D., & Nolan, B. (2007). Relationships between REM sleep findings and PTSD symptoms during the early aftermath of trauma. Journal of traumatic stress, 20(5), 893–901.
[5]Dykierek, P., Stadtmüller, G., Schramm, P., Bahro, M., van Calker, D., Braus, D. F., Steigleider, P., Löw, H., Hohagen, F., Gattaz, W. F., Berger, M., & Riemann, D. (1998). The value of REM sleep parameters in differentiating Alzheimer''s disease from old-age depression and normal aging. Journal of psychiatric research, 32(1), 1–9.
[6]Schaltenbrand, N., Lengelle, R., & Macher, J. P. (1993). Neural network model: application to automatic analysis of human sleep. Computers and biomedical research, an international journal, 26(2), 157–171.
[7]Schaltenbrand, N., Lengelle, R., Toussaint, M., Luthringer, R., Carelli, G., Jacqmin, A., Lainey, E., Muzet, A., & Macher, J. P. (1996). Sleep stage scoring using the neural network model: comparison between visual and automatic analysis in normal subjects and patients. Sleep, 19(1), 26–35.
[8]Kubat, M., Pfurtscheller, G., & Flotzinger, D. (1994). AI-based approach to automatic sleep classification. Biological cybernetics, 70(5), 443–448.
[9]Robert, C., Guilpin, C., & Limoge, A. (1998). Review of neural network applications in sleep research. Journal of neuroscience methods, 79(2), 187–193.
[10]Park, H. J., Oh, J. S., Jeong, D. U., & Park, K. S. (2000). Automated sleep stage scoring using hybrid rule- and case-based reasoning. Computers and biomedical research, an international journal, 33(5), 330–349.
[11]Agarwal, R., & Gotman, J. (2001). Computer-assisted sleep staging. IEEE transactions on bio-medical engineering, 48(12), 1412–1423.
[12]Louis, R. P., Lee, J., & Stephenson, R. (2004). Design and validation of a computer-based sleep-scoring algorithm. Journal of neuroscience methods, 133(1-2), 71–80.
[13]Hanaoka, M., Kobayashi, M. and Yamazaki, H. (2002), Automatic sleep stage scoring based on waveform recognition method and decision-tree learning. Syst. Comp. Jpn., 33: 1-13.
[14]Caffarel, J., Gibson, G. J., Harrison, J. P., Griffiths, C. J., & Drinnan, M. J. (2006). Comparison of manual sleep staging with automated neural network-based analysis in clinical practice. Medical & biological engineering & computing, 44(1-2), 105–110.
[15]Porée, F., Kachenoura, A., Gauvrit, H., Morvan, C., Carrault, G., & Senhadji, L. (2006). Blind source separation for ambulatory sleep recording. IEEE transactions on information technology in biomedicine : a publication of the IEEE Engineering in Medicine and Biology Society, 10(2), 293–301.
[16]Baumgart-Schmitt, R., Herrmann, W. M., & Eilers, R. (1998). On the use of neural network techniques to analyze sleep EEG data. Third communication: robustification of the classificator by applying an algorithm obtained from 9 different networks. Neuropsychobiology, 37(1), 49–58.
[17]Flexer, A., Gruber, G., & Dorffner, G. (2005). A reliable probabilistic sleep stager based on a single EEG signal. Artificial intelligence in medicine, 33(3), 199–207.
[18]Malafeev, A., Omlin, X., Wierzbicka, A., Wichniak, A., Jernajczyk, W., Riener, R., & Achermann, P. (2019). Automatic artefact detection in single-channel sleep EEG recordings. Journal of sleep research, 28(2), e12679.
[19]Sinha R. K. (2008). Artificial neural network and wavelet based automated detection of sleep spindles, REM sleep and wake states. Journal of medical systems, 32(4), 291–299.
[20]Magosso, E., Ursino, M., Zaniboni, A., Provini, F., & Montagna, P. (2007). Visual and computer-based detection of slow eye movements in overnight and 24-h EOG recordings. Clinical neurophysiology : official journal of the International Federation of Clinical Neurophysiology, 118(5), 1122–1133.
[21]Riazy, S., Wendler, T., & Pilz, J. (2018). Automatic two-channel sleep staging using a predictor-corrector method. Physiological measurement, 39(1), 014006.
[22]Li, Q., Li, Q., Liu, C., Shashikumar, S. P., Nemati, S., & Clifford, G. D. (2018). Deep learning in the cross-time frequency domain for sleep staging from a single-lead electrocardiogram. Physiological measurement, 39(12), 124005.
[23]陳世昌（2010）。使用不同組合的腦電圖、眼動圖及肌電圖訊號自動偵測慢波睡眠。國立中山大學機械與機電工程學系研究所碩士論文，高雄市。
[24]劉勝義 （2004）。 臨床睡眠檢查學。 臺北市：合記圖書。
[25]劉昌賢（2002）。以數位訊號處理器為基礎的即時心率分析器設計。長庚大學電機工程研究所碩士論文，桃園縣。
[26]黃明智（2006）。以數位訊號處理方式建立心率變異度與呼吸障礙指數的相關性。國立中山大學機械與機電工程學系研究所碩士論文，高雄市。
[27]Urtnasan, E., Park, J. U., Joo, E. Y., & Lee, K. J. (2018). Automated Detection of Obstructive Sleep Apnea Events from a Single-Lead Electrocardiogram Using a Convolutional Neural Network. Journal of medical systems, 42(6), 104.
[28]楊正榮（2004）。以小波轉換為基礎的QRS波偵測。國立中山大學機械與機電工程學系研究所碩士論文，高雄市。
[29]Chen, T., & Guestrin, C. (2016). XGBoost: A Scalable Tree Boosting System. Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 785-794.
[30]Krizhevsky, A., Sutskever, I., & Hinton, G.E. (2012, December). Imagenet classification with deep convolutional neural networks. In Proceedings of NIPS, pages 1106–1114, 2012.
[31]Srivastava, N., Hinton, G., Krizhevsky, A., Sutskever, I. & Salakhutdinov, R. (2014). Dropout: a simple way to prevent neural networks from overfitting. Journal of Machine Learning Research, 15(1), 1929–1958
[32]陳政緯（2019）。使用深度學習開發基於腦電訊號之老鼠睡眠階段自動判讀方法。國立中山大學機械與機電工程學系研究所碩士論文，高雄市。
[33]Szegedy, C., Liu, W., Jia, Y., Sermanet, P., Reed, S., Anguelov, D., Erhan, D., & Rabinovich, A. (2014). Going deeper with convolutions. In Proceedings of CVPR, 1–9.
[34]McHugh M. L. (2012). Interrater reliability: the kappa statistic. Biochemia medica, 22(3), 276–282.
[35]Malinowska, U., Klekowicz, H., Wakarow, A., Niemcewicz, S., & Durka, P. J. (2009). Fully parametric sleep staging compatible with the classical criteria. Neuroinformatics, 7(4), 245–253.
[36]洪志遠（2018）。使用機器學習的技術進行腦電圖訊號之自動偵測快速動眼睡眠期。國立中山大學機械與機電工程學系研究所博士論文，高雄市。