臺灣博碩士論文加值系統

精準醫療時代的來臨，顯示著疾病診斷趨向於個人化、客製化的發展，而現今醫學與資訊的結合已是時代所趨，而猝睡症是嗜睡症的一種，患者常有日間嗜睡、猝倒、入睡前幻覺等症狀，臨床診斷上須透過睡眠多項檢驗等輔以其他多項工具進行診斷，而目前大多數與猝睡症有關之研究皆只採用部分或是特定的檢測工具進行分析，而本研究共收集了約十種與猝睡症相關之量測、問卷資料，並以集成學習為基礎進行分類猝睡症I型與猝睡症II型的模型建構，在每一種不同的資料集皆進行了支援向量機、決策樹、類神經網路、最近鄰居法、樸素貝葉斯等五種分類器的訓練與參數調校，並以最佳的模型參數來訓練個別資料集的分類器，並以集成學習為基礎整合個別資料的分類器並建立混合模型，單一分類器的準確度約為57.38%~71.64%，而混合模型的準確度則為80.88%，其結果表明在有多種不同資料集下以集成學習為基礎建構混合比以單一分類器表現來的更好，而在建構過程中也透過決策樹挖掘了各資料集的特徵重要度與參考規則，如可以用MSLT與PSG中的部分參數或是Comorbidity中的hallucination等對於猝睡症類別的做進一步的分類等，這些參考規則可供未來作為臨床診斷猝睡症類別之參考指標．而臨床上亦可優先安排模型具有較高鑑別度的測驗進行，如除了必要的MSLT與PSG檢查外可優先安排PET等檢查，期能縮短臨床診床診斷流程。

The advent of the era of precision medicine shows that the diagnosis of diseases tends to be personalized and customized. Nowadays, the combination of medicine and information is the trend of the times, and Narcolepsy is a kind of Hypersomnia. Patients often have symptoms such as excessive daytime sleepiness, cataplexy, hypnagogic hallucination, Narcolepsy must be diagnosed by multiple tests of sleep, multi-stage sleep test, etc.
Most of the studies related to narcolepsy use only partial or specific tests. In this study, about ten kinds of measurement and questionnaire data related to narcolepsy were collected, and build a classifier based on ensemble learning to classify the narcolepsy type I and narcolepsy type II.
All kind of dataset will be training and selecting parameters by five kinds of classifiers, such as support vector machine, decision tree, neural network, nearest neighbor method, and naive Bayes, and training the classifiers of individual datasets with the best model parameters, and integrating the individual classifiers based on ensemble learning and establishing a hybrid model, the accuracy of the individual classifier is about 57.38%~71.64%, and the accuracy of the hybrid model is 80.88%. The result shows that the model based on ensemble learning is better than individual classifier. In the process of construction, the feature importance and reference rules of each data set are also mined through the decision tree. For example, we can use some parameters in the PSG and MSLT or the hallucination in the Comorbidity to further classify the narcolepsy category. These reference rules are available as a reference for future clinical diagnosis of narcolepsy.
In clinical practice, it is also possible to prioritize tests with high discrimination in the model. For example, in addition to the necessary MSLT and PSG tests, PET and other tests can be prioritized. The above period can shorten the clinical diagnosis process.

碩士學位考試委員會審定書 I
致謝 II
論文摘要 III
英文摘要 IV
目錄 V
圖目錄 VII
表目錄 VIII
附錄表目錄 IX
第一章 緒論 1
第一節 研究背景 1
第二節 研究動機 2
第三節 研究目的 2
第四節 研究流程 3
第二章 文獻探討 4
第一節 猝睡症 4
2.1.1 猝睡症之相關症狀 4
3.1.2 猝睡症之診斷 5
3.1.3 與猝睡症相關之量測、問卷 7
第二節 機械學習 10
2.2.1　支援向量機 10
2.2.2　決策樹 10
2.2.3　樸素貝葉斯分類器 12
2.2.4　最近鄰居法 13
2.2.5　類神經網路 13
第三節 集成學習 15
第四節 與猝睡症研究相關之文獻回顧 18
第三章 研究方法 20
第一節 研究架構 20
第二節 研究對象 22
第三節 研究工具 23
第四節 資料預處理 23
第五節 分析工具與模型參數選擇 25
第六節 模型調參過程 26
第七節 決策樹挖掘特徵流程 27
第八節 集成學習分類器之組合 27
第九節 模型評估標準 28
第四章 研究結果與討論 30
第一節 資料集模型調參結果 30
第二節 決策樹挖掘之指標 40
第三節 集成學習之強分類器評估 42
第四節 討論與實務應用 43
第五章 結論 44
第一節 研究結論與建議 44
第二節 研究限制與未來展望 45
參考文獻 46
 
圖目錄
圖1研究流程圖 3
圖2 SVM分類示意圖[30] 10
圖3決策樹示意圖[32] 12
圖4基本神經元之構成[34] 14
圖5 基本的反向傳播類神經網路[34] 15
圖6 Bagging演算法流程圖[21] 16
圖7 Boosting示意圖[33] 17
圖8 研究架構圖 21
圖 9資料預處理流程圖 24
圖10模型調參過程圖 26
圖 11決策樹挖掘特徵之流程圖 27
圖 12集成學習分類器組合之流程圖 28

表目錄
表1 ICSD-3 猝睡症類別診斷標準[4] 6
表 2 各模型參數表 25
表 3 混淆矩陣 28
表 4 MSLT資料集各模型調參最佳結果 30
表 5 PSG資料集各模型調參最佳結果 31
表 6 ESS資料集各模型調參最佳結果 32
表 7 PDSS資料集各模型調參最佳結果 33
表 8 SF_36資料集各模型調參最佳結果 34
表 9 WCST資料集各模型調參最佳結果 35
表 10 CPT_II資料集各模型調參最佳結果 36
表 11 Comorbidity資料集各模型調參最佳結果 37
表 12 PET資料集各模型調參最佳結果 38
表 13 PET/MRI資料集各模型調參最佳結果 39
表 14 各資料集之重要指標 40
表 15 各資料集最佳分類器與混合模型比較表 42

附錄表目錄
附錄1人體試驗委員會試驗同意書 50
附錄2資料集輸入參數表 51
附錄3 MSLT資料集決策樹調參結果 65
附錄4 MSLT資料集支援向量機調參結果 66
附錄5 MSLT資料集類神經網路調參結果 67
附錄6 MSLT資料集樸素貝葉斯與最近鄰居法調參結果 68
附錄7 PSG資料集決策樹調參結果 69
附錄8 PSG資料集支援向量機調參結果 70
附錄9 PSG資料集類神經網路調參結果 71
附錄10 PSG資料集樸素貝葉斯與最近鄰居法調參結果 72
附錄11 ESS資料集決策樹調參結果 73
附錄12 ESS資料集支援向量機調參結果 74
附錄13 ESS資料集類神經網路調參結果 75
附錄14 ESS資料集樸素貝葉斯與最近鄰居法調參結果 76
附錄15 PDSS資料集決策樹調參結果 77
附錄16 PDSS資料集支援向量機調參結果 78
附錄17 PDSS資料集類神經網路調參結果 79
附錄18 PDSS資料集樸素貝葉斯與最近鄰居法調參結果 80
附錄19 SF_36資料集決策樹調參結果 81
附錄20 SF_36資料集支援向量機調參結果 82
附錄21 SF_36資料集類神經網路調參結果 83
附錄22 SF_36資料集樸素貝葉斯與最近鄰居法調參結果 84
附錄23 WCST資料集決策樹調參結果 85
附錄24 WCST資料集支援向量機調參結果 86
附錄25 WCST資料集類神經網路調參結果 87
附錄26 WCST資料集樸素貝葉斯與最近鄰居法調參結果 88
附錄27 CPT_II資料集決策樹調參結果 89
附錄28 CPT_II資料集支援向量機調參結果 90
附錄29 CPT_II資料集類神經網路調參結果 91
附錄30 CPT_II資料集樸素貝葉斯與最近鄰居法調參結果 92
附錄31 Comorbidity資料集決策樹調參結果 93
附錄32 Comorbidity資料集支援向量機調參結果 94
附錄33 Comorbidity資料集類神經網路調參結果 95
附錄34 Comorbidity資料集樸素貝葉斯與最近鄰居法調參結果 96
附錄35 PET資料集決策樹調參結果 97
附錄36 PET資料集支援向量機調參結果 98
附錄37 PET資料集類神經網路調參結果 99
附錄38 PET資料集樸素貝葉斯與最近鄰居法調參結果 100
附錄39 PET/MRI資料集決策樹調參結果 101
附錄40 PET/MRI資料集支援向量機調參結果 102
附錄41 PET/MRI資料集類神經網路調參結果 103
附錄42 PET/MRI資料集樸素貝葉斯與最近鄰居法調參結果 104

[1]t. W. H. P. B. OBAMA. (2015). THE PRECISION MEDICINE INITIATIVE. Available: https://obamawhitehouse.archives.gov/precision-medicine
[2]M. Thorpy, "Therapeutic advances in narcolepsy," Sleep Medicine, vol. 8, no. 4, pp. 427-440, 2007/06/01/ 2007.
[3]K. Maski and J. A. Owens, "Insomnia, parasomnias, and narcolepsy in children: clinical features, diagnosis, and management," The Lancet Neurology, vol. 15, no. 11, pp. 1170-1181, 2016/10/01/ 2016.
[4]A. A. o. S. Medicine, "International classification of sleep disorders—third edition (ICSD-3)," AASM Resource Library, 2014.
[5]I. Kavakiotis, O. Tsave, A. Salifoglou, N. Maglaveras, I. Vlahavas, and I. Chouvarda, "Machine Learning and Data Mining Methods in Diabetes Research," Computational and Structural Biotechnology Journal, vol. 15, pp. 104-116, 2017/01/01/ 2017.
[6]D. Bansal, R. Chhikara, K. Khanna, and P. Gupta, "Comparative Analysis of Various Machine Learning Algorithms for Detecting Dementia," Procedia Computer Science, vol. 132, pp. 1497-1502, 2018/01/01/ 2018.
[7]E. Gélineau, De la narcolepsie. impr. Tessier frères, 1881.
[8]R. Press, "Narcolepsy 1985," Sleep, vol. 9, no. 1, pp. 99-101, 1986/03/01 1986.
[9]S. Nishino, "Clinical and neurobiological aspects of narcolepsy," Sleep Medicine, vol. 8, no. 4, pp. 373-399, 2007/06/01/ 2007.
[10]M. J. Thorpy and Y. Dauvilliers, "Clinical and practical considerations in the pharmacologic management of narcolepsy," Sleep Medicine, vol. 16, no. 1, pp. 9-18, 2015/01/01/ 2015.
[11]R. Pelayo, "Narcolepsy," in Encyclopedia of the Neurological Sciences (Second Edition)Oxford: Academic Press, 2014, pp. 307-310.
[12]J. Fernandez-Mendoza and S. L. Calhoun, "Chapter 21 - Excessive Daytime Sleepiness: Age, Sleep, Mood, and Metabolic Modulation," in Modulation of Sleep by Obesity, Diabetes, Age, and Diet, R. R. Watson, Ed. San Diego: Academic Press, 2015, pp. 193-202.
[13]陳建安, "台灣族群猝睡症之分子基因研究," 碩士, 分子生物暨人類遺傳學研究所, 慈濟大學, 花蓮縣.
[14]M. J. Sateia, "International classification of sleep disorders," Chest, vol. 146, no. 5, pp. 1387-1394, 2014.
[15]E. Mignot, X. Lin, J. Arrigoni, C. Macaubas, F. Olive, J. Hallmayer, P. Underhill, C. Guilleminault, W. Dement, and F. Grumet, "DQB1* 0602 and DQA1* 0102 (DQ1) are better markers than DR2 for narcolepsy in Caucasian and black Americans," Sleep, vol. 17, no. suppl_8, pp. S60-S67, 1994.
[16]F. M. S. Coelho, M. Pradella-Hallinan, M. Predazzoli Neto, L. R. A. Bittencourt, and S. Tufik, "Prevalence of the HLA-DQB1* 0602 allele in narcolepsy and idiopathic hypersomnia patients seen at a sleep disorders outpatient unit in São Paulo," Revista Brasileira de Psiquiatria, vol. 31, no. 1, pp. 10-14, 2009.
[17]Y.-H. Chen, Y.-S. Huang, W.-H. Chien, and C.-H. Chen, "Association analysis of the major histocompatibility complex, class II, DQ β1 gene, HLA-DQB1, with narcolepsy in Han Chinese patients from Taiwan," Sleep Medicine, vol. 14, no. 12, pp. 1393-1397, 2013/12/01/ 2013.
[18]W. Chen and E. Mignot, "CHAPTER 6 - Narcolepsy and Hypersomnia of Central Origin: Diagnosis, Differential Pearls, and Management," in Review of Sleep Medicine (Second Edition), T. J. Barkoukis and A. Y. Avidan, Eds. Philadelphia: Butterworth-Heinemann, 2007, pp. 75-94.
[19]王志寰, "猝睡症患者疾病嚴重度、神經認知功能對生活品質關聯之縱貫研究：階層線性模型分析," 博士, 心理學研究所, 國立政治大學, 台北市.
[20]S. Anic-Labat, C. Guilleminault, H. C. Kraemer, J. Meehan, J. Arrigoni, and E. Mignot, "Validation of a cataplexy questionnaire in 983 sleep-disorders patients," Sleep, vol. 22, no. 1, pp. 77-87, 1999.
[21]M. Hirshkowitz, A. Sarwar, and A. Sharafkhaneh, "Chapter 143 - Evaluating Sleepiness," in Principles and Practice of Sleep Medicine (Fifth Edition), M. H. Kryger, T. Roth, and W. C. Dement, Eds. Philadelphia: W.B. Saunders, 2011, pp. 1624-1631.
[22]A. Shimura, S. Hideo, Y. Takaesu, R. Nomura, Y. Komada, and T. Inoue, "Comprehensive assessment of the impact of life habits on sleep disturbance, chronotype, and daytime sleepiness among high-school students," Sleep Medicine, vol. 44, pp. 12-18, 2018/04/01/ 2018.
[23]林小蕓, "社會經濟狀況相關健康不均度探討--以SF-36與疾病嚴重度指標為例," 碩士, 醫務管理學研究所, 長庚大學, 桃園縣, 2008.
[24]林青慧, "臺灣簡短36（SF-36）健康量表工具信效度及常模之建立," 碩士, 醫務管理研究所, 中國醫藥學院, 台中市, 2003.
[25]R. J. Ogg, P. Zou, D. N. Allen, S. B. Hutchins, R. M. Dutkiewicz, and R. K. Mulhern, "Neural correlates of a clinical continuous performance test," Magnetic Resonance Imaging, vol. 26, no. 4, pp. 504-512, 2008/05/01/ 2008.
[26]C. Mesquita, B. P. Nazar, C. M. S. Pinna, B. Rabelo, M. A. Serra-Pinheiro, J. Sergeant, and P. Mattos, "How can Continuous Performance Test help to assess inattention when mood and ADHD symptoms coexist?," Psychiatry Research, vol. 243, pp. 326-330, 2016/09/30/ 2016.
[27]J. R. Kirby, H.-J. Kim, and R. Silvestri, "Chapter 11 - Cognitive Constructs and Individual Differences Underlying ADHD and Dyslexia: A Cognitive Mosaic Approach," in Cognition, Intelligence, and Achievement, T. C. Papadopoulos, R. K. Parrila, and J. R. Kirby, Eds. San Diego: Academic Press, 2015, pp. 197-223.
[28]M. Hoffmann, "CHAPTER 18 - Cognitive Testing," in Clinical Neurotoxicology, M. R. Dobbs, Ed. Philadelphia: W.B. Saunders, 2009, pp. 222-234.
[29]I. Rausch, H. H. Quick, J. Cal-Gonzalez, B. Sattler, R. Boellaard, and T. Beyer, "Technical and instrumentational foundations of PET/MRI," European Journal of Radiology, vol. 94, pp. A3-A13, 2017/09/01/ 2017.
[30]林宗勳, Support Vector Machines簡介. 台灣大學資訊工程研究所, 2000.
[31]鄭捷, 今天不學機器學習，明天就被機器取代：從Python入手+演算法. 台灣: 佳魁資訊, 2016.
[32]S. W. Jarvis, C. Kovacs, T. Badriyah, J. Briggs, M. A. Mohammed, P. Meredith, P. E. Schmidt, P. I. Featherstone, D. R. Prytherch, and G. B. Smith, "Development and validation of a decision tree early warning score based on routine laboratory test results for the discrimination of hospital mortality in emergency medical admissions," Resuscitation, vol. 84, no. 11, pp. 1494-1499, 2013/11/01/ 2013.
[33]G. Wang, J. Sun, J. Ma, K. Xu, and J. Gu, "Sentiment classification: The contribution of ensemble learning," Decision Support Systems, vol. 57, pp. 77-93, 2014/01/01/ 2014.
[34]P. Dande and P. Samant, "Acquaintance to Artificial Neural Networks and use of artificial intelligence as a diagnostic tool for tuberculosis: A review," Tuberculosis, vol. 108, no. Supplement C, pp. 1-9, 2018/01/01/ 2018.
[35]G. Litjens, T. Kooi, B. E. Bejnordi, A. A. A. Setio, F. Ciompi, M. Ghafoorian, J. A. W. M. van der Laak, B. van Ginneken, and C. I. Sánchez, "A survey on deep learning in medical image analysis," Medical Image Analysis, vol. 42, pp. 60-88, 2017/12/01/ 2017.
[36]A. Awad, M. Bader-El-Den, J. McNicholas, and J. Briggs, "Early hospital mortality prediction of intensive care unit patients using an ensemble learning approach," International Journal of Medical Informatics, vol. 108, pp. 185-195, 2017/12/01/ 2017.
[37]F. N. Koutanaei, H. Sajedi, and M. Khanbabaei, "A hybrid data mining model of feature selection algorithms and ensemble learning classifiers for credit scoring," Journal of Retailing and Consumer Services, vol. 27, pp. 11-23, 2015/11/01/ 2015.
[38]Y. Kazemi and S. A. Mirroshandel, "A novel method for predicting kidney stone type using ensemble learning," Artificial Intelligence in Medicine, vol. 84, pp. 117-126, 2018/01/01/ 2018.
[39]C. O. Inocente, M.-P. Gustin, S. Lavault, A. Guignard-Perret, A. Raoux, N. Christol, D. Gerard, Y. Dauvilliers, R. Reimão, F. Bat-Pitault, J.-S. Lin, I. Arnulf, M. Lecendreux, and P. Franco, "Depressive feelings in children with narcolepsy," Sleep Medicine, vol. 15, no. 3, pp. 309-314, 2014/03/01/ 2014.
[40]J. Black, N. L. Reaven, S. E. Funk, K. McGaughey, M. M. Ohayon, C. Guilleminault, and C. Ruoff, "Medical comorbidity in narcolepsy: findings from the Burden of Narcolepsy Disease (BOND) study," Sleep Medicine, vol. 33, no. Supplement C, pp. 13-18, 2017/05/01/ 2017.
[41]Z. Wang, H. Wu, W. S. Stone, J. Zhuang, L. Qiu, X. Xu, Y. Wang, Z. Zhao, F. Han, and Z. Zhao, "Body weight and basal metabolic rate in childhood narcolepsy: a longitudinal study," Sleep Medicine, vol. 25, pp. 139-144, 2016/09/01/ 2016.
[42]M.-J. Lee, S.-Y. Lee, S.-S. Yuan, C.-J. Yang, K.-C. Yang, T.-L. Lee, C.-C. Sun, Y.-C. Shyu, and L.-J. Wang, "Comorbidity of narcolepsy and depressive disorders: a nationwide population-based study in Taiwan," Sleep Medicine, vol. 39, pp. 95-100, 2017/11/01/ 2017.
[43]A. V. Olsen, J. Stephansen, E. Leary, P. E. Peppard, H. Sheungshul, P. J. Jennum, H. Sorensen, and E. Mignot, "Diagnostic value of sleep stage dissociation as visualized on a 2-dimensional sleep state space in human narcolepsy," Journal of Neuroscience Methods, vol. 282, pp. 9-19, 2017/04/15/ 2017.
[44]R. Leenings, C. Glatz, A. Heidbreder, M. Boentert, G. Pipa, and P. Young, "P 120 A machine learning approach to detecting sleep and sleep disorders in acceleration sensor data," Clinical Neurophysiology, vol. 128, no. 10, p. e388, 2017/10/01/ 2017.
[45]F. Pedregosa, G. Varoquaux, A. Gramfort, V. Michel, B. Thirion, O. Grisel, M. Blondel, P. Prettenhofer, R. Weiss, and V. Dubourg, "Scikit-learn: Machine learning in Python," Journal of machine learning research, vol. 12, no. Oct, pp. 2825-2830, 2011.
[46]C. Wang, A. Elazab, J. Wu, and Q. Hu, "Lung nodule classification using deep feature fusion in chest radiography," Computerized Medical Imaging and Graphics, vol. 57, pp. 10-18, 2017/04/01/ 2017.