本篇論文提出一套以色彩為基礎之自動化偵測結核菌系統於顯微影像上。整個系統可以分成兩大階段：結核菌區域偵測和分類。結核菌區域偵測的目的為得知結核菌在影像中可能的位置，其中包含了影像前處理、結核菌候選區域偵測和計算特徵參數。在影像前處理的部分，我們設計了一個亮度補償的機制以解決顯微影像上亮度不均勻的問題。由於每張顯微影像中細胞組織與背景的比例不同，我們運用亮度變異程度對所有影像分成三大類，於後續處理中，將針對每張影像所屬類別，運用其所屬參數。接下來對每張影像進行色彩標準化，以降低在同類別中不同顯微影像之間的色彩差異。我們也將影像中佔了大部分的亮背景去除，在剩下的細胞組織中，利用每個像素點的顏色特徵代入高斯混合模型，作為偵測是否為結核菌的依據。經過標記與型態學上的運算，即可將結核菌候選區域框選出來，進而計算特徵參數。
結核菌候選區域分類的階段，我們會根據不同類型影像選用不同的特徵參數來提升結核菌分類的準確率。我們採用模糊邏輯分類器作為分類的機制，並與倒傳遞類神經網路做比較，根據結果顯示在模糊邏輯分類器的敏感性大約八成以上。整體來說，我們的系統可以自動且快速的偵測出結核菌。

This paper presents an automatic color-based Mycobacterium tuberculosis (MTB) detection method on optical microscopic images. The proposed method consists of two phases: detection of MTB candidates and classification. The detection phase is to find the location of MTB candidates and its processing steps include image preprocessing, detection of MTB candidates and calculation of feature parameters. In image preprocessing, we first designed a lightening compensation step to reduce the non-uniform brightness on microscopic images. Then, the microscopic images were classified into three types based on the variance of illumination. Next, a color normalization step is applied to reduce the color variation within the same type of images. The empty background which is light in color and occupies most area is then removed after normalization from the image. In the remaining textures, color features are extracted and evaluated by using Gaussian mixture model (GMM) to detect MTB candidates. By applying labeling and morphological methods, the candidates of MTB can be obtained and the corresponding parameters can be computed.
In classification phase, different sets of parameters are selected to improve the classification accuracy for each type of images. The fuzzy logic classifier and the back-propagation neural network (BPN) were used for the classification of MTB. The experimental results show that the previous one has better performance than the later one. In summary, the proposed system can perform MTB detection automatically, efficiently and with good enough accuracy.

[1] A.S. deBoer, B. Blommerde, P.E.W. deHaas, M.M. G. Sebek, K.S. B. Lambregts-van Weezenbeek, M. Dessens, and D. van Soolingen, “False-Positive Mycobacterium tuberculosis Cultures in 44 Laboratories in Netherlands (1993 to 2000): Incidence, Risk Factors, and Consequences,” Journal of Clinical Microbiology, vol. 40, pp. 4004–4009, 2002.
[2] T. Hanscheid, C.M Ribeiro, H.M Shapiro, and N.G Perlmutter, "Fluorescence microscopy for tuberculosis diagnosis," The Lancet Infectious Diseases, vol. 7, pp. 236–237, 2007.
[3] V. H. Nguyen, N. S. Dinh, R. M. Anthony, F. G. Cobelens and D van Soolingen, "Fluorescence microscopy for tuberculosis diagnosis," The Lancet Infectious Diseases, vol. 7, pp. 238–240, 2007.
[4] 行政院衛生署, “結核病十年減半全民動員計畫,” 行政院衛生署, 2006。
[5] B.L. Stone, W.J. Burman, M.V. Hildred, E.A. Jarboe, R.R. Reves, and M.L. Wilson, “The Diagnostic Yield of Acid-Fast-Bacillus Smear-Positive Sputum Specimens,” Journal of Clinical Microbiology, vol. 35, no. 4, pp. 1030–1031, 1997.
[6] 呂喬洋、林碧芬, “建立標準結核菌實驗室,” 行政院衛生署疾病管制局九十一年度科技研究發展計畫研究報告, 2003。
[7] K. Veropoulos, C. Campbell, and G. Learnmonth, “Image processing and neural computing used in the diagnosis of tuberculosis,” IEE Colloquium on Intelligent Methods in Healthcare and Medical Applications, pp. 1–4, 1998.
[8] M.G. Forero, E.L. Sierra, J. Alvarez, J.L. Pech, G. Cristbal, L. Alcal, and M. Desco, “Automatic sputum color image segmentation for tuberculosis diagnosis,” Proceedings of SPIE, vol. 4471, pp. 251–261, 2001.

[9] M.G. Forero, F. Sroubek, J. Alvarez, N. Malpica, G. Cristbal, A. Santos, L. Alcala, M. Desco, and L. Cohen, “Segmentation, autofocusing and signature extraction of tuberculosis sputum images,” Proceedings of SPIE, vol. 4788, pp. 171–182, 2002.
[10] M.G. Forero, F. Sroubek, and G. Cristbal, “Identification of tuberculosis bacteria based on shape and color,” Real-Time Imaging, vol. 10, no. 4, pp. 251–262, 2004.
[11] M.G. Forero and G. Cristbal, “Automatic identification of Mycobacterium tuberculosis by Gaussian Mixture Models,” Journal of microscopy, vol. 223, no.2, pp. 120–132, 2006.
[12] M.G. Costa, C.F. Marly Filho, J.F. Sena, J. Salem, and M.O. de Lima, “Automatic identification of mycobacterium tuberculosis with conventional light microscopy,” IEEE Engineering in Medicine and Biology Society, pp. 382–385, 2008.
[13] N. OTSU, “A Threshold Selection Method from Gray-level Histogram,” IEEE Transactions on Systems, Man and Cybernetics, vol. 9, no. 1, pp. 62-66, 1979.
[14] E. Reinhard and M. Ashikhmin et al., “Color transfer between images,” IEEE Computer Graphics and Applications, vol. 21, no. 5, pp. 34–41, 2001.
[15] D.L. Ruderman, T.W. Cronin, and C.-C. Chiao, “Statistics of cone responses to natural images: implications for visual coding,” Journal of the Optical Society of America A, vol. 15, no. 8, pp. 2036–2045, 1998.
[16] Y. Linde, A. Buzo, and R.M. Gray, “An Algorithm for Vector Quantizer Design,” IEEE Transactions on Communications, vol. 28, pp. 84-95, 1980.
[17] A.P. Dempster, N.M. Laird, and D.B. Rubin, “Maximum Likelihood from Incomplete Data via the EM Algorithm”, Journal of the Royal Statistical Society. Series B (Methodological), vol. 39, no. 1, pp. 1-38, 1977.
[18] T.Y. Zhang and C.Y. Suen, “A Fast Parallel Algorithm for Thinning Digital Patterns,” Communications of the ACM, vol. 27, pp. 236–239, 1984.
[19] R. C. Gonzalez and R. E. Woods, “Digital Image Processing,” Prentice Hall, 2002
[20] 楊英魁, 孫宗瀛, 鄭魁香, “模糊控制理論與技術,” 全華圖書, 1996.
[21] L. Breiman, J.H. Friedman, R.A. Olshen, and C.J. Stone, “Classification and Regression Tree, ” Belmont: Wadsworth, 1984.
[22] F. Rosenblatt, “The Perceptron: A Probabilistic Model for Information Storage and Organization in Brain,” Psychological Review, vol.65, pp.386-408, 1958.
[23] D.E. Rumelhart and J.L. McClelland, “Parallel Distribute processing: Exploration in the Microstructure of Cognition,” MIT Press Cambridge, MA, USA, vol.1, 1986.