微循環檢測由於操作方便、非侵入式、價格便宜、對病人沒有痛苦，能直接動態觀察人體微循環與微血流，對於疾病診斷、治療與預後等有很大幫助。微循環主要觀察部位為手、足甲襞、眼結膜、舌、唇等部位，而甲襞微循環通常觀看左右手無名指甲襞。量測活體血液流速時，最常遭遇問題為活體晃動無法穩固樣本，晃動將隨放大率增大而加劇。雖可設計夾具固定觀測部位，惟夾具將影響血流速度，導致量測誤差。而以雷射都卜勒血流測速法，則價錢昂貴、僅能測量較大血管或大區域之平均流速、缺乏多樣性血管外形靜態特徵分析、精確度較顯微鏡為差，且因由於細小動脈規則性收縮和舒張造成微血管血流速度不平穩，僅能測量局部血流速度，無法完整全面描述整條微血管血流速度，易造成微循環障礙重要訊息之遺漏。
本論文透過電腦視覺技術對顯微微循環影像序列進行因活體晃動之影像位移補償自動穩定影像序列便不需手指夾具完全固定。再擷取血管區域以量測及分析其靜態及動態病理特徵，進而診斷人體健康狀況。將三維電腦圖學建模技術加以融合，重建擷取之血管區域及靜態特徵，並偵測血流流動變化，推算血液流速、流向等動態資訊，藉此診斷人體末梢血管循環狀態。將動態血流資訊整合至三維血管模組，呈現三維血管動態血流視覺，透徹觀察微循環靜態及動態特徵。
本論文擷取之多樣化微循環血管特徵，可區分為動態與靜態兩部分。靜態特徵包括血管管袢之數量、長度、寬度、密度、動脈輸入支/靜脈輸出支/袢頂 管徑、血色、袢頂寬度、彎曲度，配合醫學臨床資訊進行各項特徵之量測；而動態特徵則以流速及流向為主要量測指標，分析微血管血液流動情形優劣並整合所有特徵資訊分析病患當前健康情況，給予醫師診斷時參考，不再流於主觀獨斷，整體系統具有操作方便、不需採集血液，成本低廉等優點，臨床診治上將有極大發展潛力。

Nailfold capillary microscopy is simple, non-invasive, no injuries and easy to observe human`s microcirculation and micro blood stream directly. Due to these advantages, it plays a significant role in diseases diagnoses, treatments and prognosis. The observation of microcirculation focuses on hand, foot naildfold, conjunctival, lingual surface and lips. Nailfold microcirculation is usually performed on the ring finger. However, when measuring the speed of blood flow, difficulty to stabilize the region of interest (ROI) is often encountered. This problem becomes more serious when the magnification of microscope increases. Fixture to stabilize finger will inevitably affect the speed of blood flow under observation. The Laser Doppler blood flow velocimetry method, is expensive, only can be used in bigger capillary or to measure the average flow velocity of lager observed area, lacking of diversified morphological features of capillary, it’s precision is worse than microscopy image capture method, and because of the regular contraction and relaxation of arterioles it can only measure the local blood flow velocity, cannot describe whole details of capillary blood flow velocity, some important information of microcirculation will be ignored easily.
This thesis employs computer vision technique to operate displacement compensation of microscopy image sequence to stabilize observed area and extract area of capillary. Then the morphological and hemodynamic pathology features will be derived and analyzed to evaluate the status of a person’s health. Not only morphological features, e.g., length, density and color, but also hemodynamic features, e.g., blood flow velocity will be measured to assess the microcirculation in end capillary. The most significant characteristic of this project is to combine three-dimensional models reconstruction technology of computer graphic to reconstruct three-dimensional capillary models and perform the three-dimensional dynamic blood flow visualization. Thus, the capillary blood flow can be adjusted and observed in the desired orientation, magnification and viewpoint.
A variety of pathologically significant features of nailfold microcirculation will be extracted in the project proposed. These features can be classified into morphological and hemodynamic features. The morphological features extracted include the number, width/height, density, arteriolar limb caliber, curved segment caliber, venular limb caliber, blood color, tortuosity, and width of the curved segment of capillaries. On the other hand, hemodynamic features including velocity, direction of blood flow will also be extracted. By integrating both morphological and hemodynamic features, the status of a person’s health can be evaluated by the doctor. The novel system proposed is not only easy to operate, low-cost but also has the great potential to be utilized clinically.

論文審定書………………………………………………………………………………i
誌謝……………………………………………………………………………………ii
中文摘要………………………………………………………………………………iv
英文摘要………………………………………………………………………………vi
圖次……………………………………………………………………………………x
表次……………………………………………………………………………………xii
第一章 簡介 1
1.1 背景及目的 1
1.2 臨床常使用之微循環檢測疾病 3
第二章 相關研究 5
2.1 甲襞微血管分類 5
2.2 微循環主要觀察指標 6
2.3 微循環特徵分析之相關研究 10
2.4 研究器材 15
第三章 研究方法 17
3.1 研究方法流程 17
3.2 影像序列位移補償 18
3.2.1 拉普拉斯高斯濾波器(Laplacian of Gaussian Filter) 18
3.2.2 區塊比對位移推導 20
3.3 影像序列強化及血管區域分離 22
3.3.1 Green Channel 22
3.3.2 直方圖等化 23
3.3.3 Otsu’s影像二值化 24
3.3.4 二值化影像序列之時間域統計擷取 27
3.4 血管中央線提取 29
3.4.1 中軸轉換(Medial-axis transform) 29
3.4.2 血管管徑提取 31
3.5 三維血管模組重建 32
3.6 三維動態血流視覺 36
第四章 微循環特徵擷取及分析 38
4.1 靜態特徵擷取 38
4.1.1 管袢數量 40
4.1.2 管袢長寬 40
4.1.3 管袢密度 41
4.1.4 管徑寬度 41
4.1.5 血色 42
4.1.6 袢頂寬度 43
4.1.7 彎曲度 44
4.2 動態特徵擷取 45
4.2.1 流速 45
4.2.2 流向 47
第五章 實驗結果 48
第六章 結論 59
參考文獻 60

[1]陳文傑、田牛、趙國忠，“微循環的理論和應用”，人民衛生出版社，1987。
[2]田牛, "微循環方法學", 原子能出版社, 1985.
[3]Fai Yau, Jenq Nan Yih, Hsiao Yi Lee, Yiu-Keung Lai, and Hen-Hong Chang, "In vivo observation of the blood flow in human capillaries under finger nailfold," Journal of Medical and Biological Engineering, Vol. 22, no. 2, pp. 91-96, Jun. 2002.
[4]Riaño-Rojas J. C., Prieto-Ortiz F. A., "Segmentation and Extraction of Morphologic Features from Capillary Images," IEEE Conf. Sixth Mexican International, pp.148-159. Nov. 2007.
[5]Verdant C, De Backer D., "How monitoring of the microcirculation may help us at the bedside, " Curr Opin Crit Care, 11(3): pp. 240-244, 2005.
[6]Dinrenfass: Blood viscosity, hyperviscosity and hyper-cerebrovascular disorder. Edit by Dinrenfass,MTP Press; Egland, pp.35-191, 1986.
[7]Copley AL. Diabetes mellitus – a biorhelogical disease of the vessel-blood organ Clinical Hemorheology, 11(3/1), pp.149, 1991.
[8]W. C. Gibson, H. J. Bosley, R. S. Griffiths, “Photomicrographic studies of the nailbed capillary networks in human control subjects,” J. Nerv. Ment. Dis., Vol. 219, pp. 219–231, 1956.
[9]A. S. Norris, J. R. Chowning, “Capillary morphology of the nailfold in the mentally ill,” J. Neuropsychology,Vol. 5, pp. 225-234, 1963.
[10]P. M. Houtman, C. G. M. Kallenberg, V. Fidler, A. A. Wouda, “Diagnostic significance of nailfold capillary patients with Raynaud’s phenomenon,” J. Rheumatol., Vol. 13, no. 3, p. 556, 1986.
[11]B. F. Jones, M. Oral, C. W. Morris, E. F. J. Ring, "A Proposed Taxonomy for Nailfold Capillaries Based on Their Morphology," IEEE Conf. Engineering in Medicine and Biology Society, pp.333-341. Apr. 2001.
[12]劉育英，“微循環圖譜”，人民軍醫出版社，2005。
[13]F.Langeder, B.G. Zagar, " Image Processing Strategies To Accurately Measure Red Blood Cell Motion In Superficial Capillaries," IEEE Conf. Systems signals and Devices 2009, pp. 1-5, Mar. 2009.
[14]Tzu-Ching Shih, Geoffrey Zhang, Chih-Chieh Wu, Hung-Da Hsiao, Tung-Hsin Wu, Kang-Ping Lin, Tzung-Chi Huang, “Hemodynamic analysis of capillary in finger nail-fold using computational fluid dynamics and image estimation,” Mircrovasc. Res. 81, 68–72, 2011.
[15]T. Eiju, K. Matsuda, J. Ohtsubo, K. Honma, K. Shimizu, "Frequency sihfting of LDV for blood velocity measurement by a moving wedgedglass," Applied Optics, Vol. 20 no. 22, pp.3833-3837, Nov. 1981.
[16]S. Einav and H. J. Berman, "Fringe mode transmittance laser Doppler microscope anemometer: Its adaptation for measurement in themicrocirculation, " Journal of Biomedical Engineering, vol. 10, no. 5, pp. 393-399, Oct. 1988.
[17]L. Alparone, S. Baronti, A. Casini, “A novel approach to the suppression of false contours originated from Laplacian-of Gaussian zero-crossings,” Proc. IEEE Int. Conf. on Image Processing ICIP96, Vol. 3, pp. 825-828, 1996.
[18]Ricci E., Perfetti R., “Retinal Blood Vessel Segmentation Using Line Operators and Support Vector Classification,” Medical Imaging, IEEE Transactions on, Vol. 26, pp. 1357, Oct. 2007.
[19]Gonzalez, R.C., and Woods, R.E., “Digital Image Processing 2nd edition,” (Prentice Hall ISBN: 0201180758), 2002.
[20]Otsu, N.: ‘A threshold selection method from gray level histogram’, IEEE Trans. Syst. Man Cybern., 20, pp. 62–66, 1979.
[21]Ward, A.D., Hamarneh, G., “The Groupwise Medial Axis Transform for Fuzzy Skeletonization and Pruning,” PAMI(32), No. 6, pp. 1084-1096. 2010.
[22]Songtao Han, Zhiwen Xu. “Three Dimensional Simulation for Blood Vessel of Ocular Funds,” IEEE Conf. Mechatronics and Automation 2009, pp. 893-898, Aug. 2009.
[23]Pazos-Moura CC, Moura EG, Bouskela E, Torres-Filho IP, Breitenbach MM, “Nailfold capillaroscopy in diabetes mellitus: morphological abnormalities and relationship with microangiopathy,” Braz J Med Biol Res, 20, pp. 777-780, 1987.
[24]Enrico Grisan, Marco Foracchia, Alfredo Ruggeri. “A novel method for automatic evaluation of retinal vessel tortuosity,” Proc. 25th Annual International Conference of the IEEE EMBS, Cancun, Mexico, 2003.