世界衛生組織(WHO)於今年發布的全球糖尿病報告指出，去年全球糖尿病的醫療花費約為22兆台幣，罹病人數從1980年至2014年已成長三倍，達4.22億人。而衛生服務部在2016年提供的資料顯示，糖尿病在2014年為國人的第五大死因，共奪走9845條寶貴的生命，平均每小時就造成1人死亡。
而在罹患糖尿病之前，通常都會先在微循環訊號上產生異常，之後才會在人體上有明顯的變化。故本研究的目標就是找出可以明顯區分出有無糖尿病的微循環指標，並利用這些指標達到診斷糖尿病或監控糖尿病的功能。
為了實現此目標，我們找來了三個不同族群來當我們的受測者，分別是健康年輕人21位、高齡糖尿病人22位以及高齡的健康老人16位，總共59人。透過血管分析儀器量到的脈波傳遞速度以及自製的量測系統所測得的脈波傳遞時間，我們已經發展了一些特徵指標。接著我們以二次判別配合留一交叉驗證和留二交叉驗證來進行概括性的測試，驗證這些指標的效能。
在區分糖尿病和健康老人方面，以腳部PTT搭配右邊haPWV，經留二交叉驗證後的分類的準確率有78%、sensitivity為86%、specificity為69%、PPV為79%。而在區分年輕和老人兩族群時，以左邊haPWV為特徵指標所得到的準確率達98%、sensitivity為97%、specificity和PPV則是1。

According to a WHO report released in 2016, the global annual medical cost of diabetes in 2015 is about NT$ 22 trillion. In the period of 1980s to 2014, the number of diabetic patients has tripled and increased to 422 million. The data provided by the Taiwan’s Ministry of Health and Welfare shows that diabetes was the fifth leading cause of death in Taiwan in 2014 and accounted for 9845 deaths. On average, a Taiwanese died of diabetes every hour.
It is likely that abnormal microcirculation occurs before diabetes onset. Therefore, the goal of this study is to find out microcirculation signal features that can effectively identify diabetes patients so that these features can be used to diagnose diabetes or monitor the progress of diabetes.
Toward this goal, we recruit three groups of test subjects which include 21 young people, 22 diabetic patients and 16 healthy elder people. Via the pulse wave velocity (PWV) measure by a vascular profiler the pulse transit time (PTT) obtained by an in-house measurement device, we have developed a number of features. This work tests the efficacy of these features by using quadratic discriminant analysis (QDA) that employs leave-one-out and leave-two-out generalization test strategies.
In distinguishing the diabetic patients and healthy old people, the feature combination of foot PTT and right hand PWV yields the leave-two-out classification results of accuracy 0.78, sensitivity 0.86, specificity 0.69, and positive-predictive-value (PPV) 0.78. In classifying young and old age groups, the left had PWV feature obtained 98% accuracy, sensitivity 0.97, specificity 1 and PPV 1.

論文審定書 i
誌謝 ii
摘要 iii
Abstract iv
目錄 v
圖次 ix
表次 xii
第一章 緒論 1
1.1 前言 1
1.2 研究動機與目的 2
1.3 論文架構 2
第二章 實驗原理與方法 3
2.1 實驗對象 3
2.2 實驗原理 4
2.2.1 心電圖ECG 4
2.2.2 光體積描述儀 5
2.2.3 脈波傳遞時間PTT 5
2.2.4 脈波傳遞時間PTT身高校正 6
2.2.5 脈波傳遞速度PWV 7
2.3 實驗方法 8
2.3.1 實驗設備 8
2.3.2 訊號量測方法 11
第三章 訊號分析流程與方法 14
3.1 訊號前處理 14
3.1.1 程式處理流程 14
3.1.2 人工處理流程 14
3.2 門檻值法則 15
3.3 二次判別分析(Quadratic Discriminant Analysis) 17
3.4 分類器 19
3.4.1 留一交叉驗證(Leave One Out) 19
3.4.2 留二交叉驗證(Leave Two Out) 20
第四章 實驗結果與討論 22
4.1 樂齡老人和健康年輕人分類結果 22
4.1.1相關指標介紹 22
4.1.2 雙特徵二次判別直接分類結果 23
4.1.3 單特徵門檻值Leave One Out分類結果 24
4.1.4 雙特徵二次判別Leave One Out分類結果 24
4.1.5 單特徵門檻值Leave Two Out分類結果 25
4.1.6 雙特徵二次判別Leave Two Out分類結果 26
4.1.7 二次判別直接分類和Leave Two Out比較 27
4.1.8單特徵和雙特徵比較 28
4.1.9 PTT和PWV的比較 28
4.2 樂齡老人加糖尿病老人和健康年輕人分類結果 29
4.2.1 雙特徵二次判別直接分類結果 29
4.2.2 單特徵門檻值Leave One Out分類結果 30
4.2.3 雙特徵二次判別Leave One Out分類結果 31
4.2.4 單特徵門檻值Leave Two Out分類結果 32
4.2.5 雙特徵二次判別Leave Two Out分類結果 32
4.2.6 二次判別直接分類和Leave Two Out比較 33
4.2.7 單特徵和雙特徵比較 34
4.2.8 PTT和PWV的比較 35
4.3 糖尿病老人和樂齡老人分類結果 35
4.3.1 雙特徵二次判別直接分類結果 35
4.3.2 單特徵門檻值Leave One Out分類結果 36
4.3.3 雙特徵二次判別Leave One Out分類結果 37
4.3.4 單特徵門檻值Leave Two Out分類結果 38
4.3.5 雙特徵二次判別Leave Two Out分類結果 38
4.3.6 二次判別直接分類和Leave Two Out比較 39
4.3.7 單特徵和雙特徵比較 40
4.3.8 PTT和PWV的比較 40
4.3.9 年紀因素考量 41
4.3.10 消除年紀因素後的PTT標準差之分類 44
4.3.11 PTT越零點之分類 45
第五章 結論與未來展望 47
參考文獻 48
附錄 50
附錄I 樂齡老人和健康年輕人QDA直接分類結果 50
附錄II 樂齡老人加糖尿病人和健康年輕人QDA直接分類結果 54
附錄III糖尿病人和樂齡老人QDA直接分類結果 57
附錄IIII 動脈硬化檢測儀結果 61
附錄V 各族群基本資料 63
附錄VI Leave One Out & Leave Two Out比較 68

[1] 衛服部國民健康署。2016年6月取自http://www.hpa.gov.tw/BHPNet/Web/News/News.aspx?No=201604070001.
[2] O'Rourke, M.F., et al., Clinical applications of arterial stiffness; definitions and reference values. American Journal of Hypertension, 2002. 15(5): 426-444.
[3] Allen J., Overbeck K., Nath A. F., Murray A., and Stansby G., “A prospective comparison of bilateral photoplethysmography versus the ankle-brachial pressure index for detecting and quantifying lower limb peripheral arterial disease,” Journal of Vascular Surgery, vol. 47, no. 4, 2008:794-802, 4//.
[4] Nicolaides A. N., “Investigation of Chronic Venous Insufficiency: A Consensus Statement,” Circulation, vol. 102, no. 20, November 14, 2000, e126-e163.
[5] Rogers J. H. and Laird J. R., “Overview of New Technologies for Lower Extremity Revascularization,” Circulation, vol. 116, October 30, 2007, no. 18, 2072-2085.
[6] R. Bezooijen, H. C. M. van den Bosch, A. V. Tielbeek, G. R. P. Thelissen, K. Visser, M. G. M. Hunink, L. E. M. Duijm, J. Wondergem, J. Buth, and P. W. M. Cuypers, “Peripheral Arterial Disease: Sensitivity-encoded Multiposition MR Angiography Compared with Intraarterial Angiography and Conventional Multiposition MR Angiography1,” Radiology, April 1, 2004, vol. 231, no. 1,: 263-271.
[7] Allen J., “Photoplethysmography and its application in clinical physiological measurement,” Physiological Measurement, 2007, vol. 28, no. 3, : R1.
[8] Omron VP-1000__Manual_BP203RPE-II.
[9] Rajan V., Varghese B., Leeuwen T., and Steenbergen W., “Review of methodological developments in laser Doppler flowmetry,” Lasers in Medical Science, 2009, vol. 24, no. 2, : 269-283, 2009/03/01.
[10] Hertzman, A. B. and Spielman, C. R, “Observations on the finger volume pluse recorded photoelectrically”, Am. J. Physiol , 1937,vol.119,334-5.
[11] 馮盈綺，抬腿動作對微循環生理訊號的影響，2014，國立中山大學機械與機電工程學系研究所碩士學位論文。
[12] Fung P., Dumont G., Ries C., Mott C. and Ansermino M., “Continuous noninvasive blood pressure measurement by pulse transit time”, in Ann. Int.Conf. IEEE ENG. Med. Biol. Soc., 2004,vol.1,738-741.
[13] Blacher, J., et al., Aortic pulse wave velocity index and mortality in end-stage renal disease. Kidney Int, 2003. 63(5): 1852-60.
[14] Kanda, T., et al., Arterial pulse wave velocity and risk factors for peripheral vascular disease. Eur J Appl Physiol, 2000. 82(1-2): 1-7.
[15] 高中軍訓第三冊光碟，幼獅出版社。
[16] Blacher, J., et al., Aortic pulse wave velocity as a marker of cardiovascular risk in hypertensive patients. Hypertension, 1999. 33(5): 1111-7.
[17] Zheng, Z.J., et al., Associations of ankle-brachial index with clinical coronary heart disease, stroke and preclinical carotid and popliteal atherosclerosis: the Atherosclerosis Risk in Communities (ARIC) Study. Atherosclerosis, 1997. 131(1): 115-25.
[18] 示孟玉，使用心電圖及光體積變化描述儀辨識情緒的技術探討，2013，國立中正大學電機工程研究所碩士論文。
[19] Santosh S., Maya R.G., Bela A.F., Bayesian Quadratic Discriminant Analysis.
Journal of Machine Learning Research 8, 2007; 1277-1305
[20] Villarroya A., Ríos M., Oller J.M., Discriminant analysis algorithm based on a distance function and on a Bayesian decision. Biometrics. 1995Sep;51(3):908-19.