臺灣博碩士論文加值系統

根據行政院衛生署之統計，乳癌高居台灣女性癌症死亡原因第四名，對女性生命造成相當大的威脅，因此乳癌之早期診斷與適當治療具有相當大之重要性。乳癌最重要的臨床表徵為可觸摸到的乳房腫瘤硬塊，當女性自行發現乳房硬塊時，應趕緊就醫，若發現疑似癌症之病症時，則必須安排更進一步之檢查，以粗針抽吸切片檢查或細針抽吸抹片檢查的方式採集病灶組織之樣本，進行組織學病理檢查與細胞學細胞檢查，盡早確定病變之發生，盡早展開治療以提高治癒之機會。
本研究之主要目的在利用細胞學以及影像處理技術，針對乳房乳癌細胞影像進行分析，分割出細胞核，並針對細胞核部分取出與乳癌判別有關之相關形態與特徵參數，以判別細胞核是否異常，儘早偵測出乳癌之發生。
研究過程中直接從顯微鏡擷取乳房細針抹片或粗針切片影像進行分析，利用色彩模型之轉換、對比強化分閥值演算法、數學形態學運算、多重主動式輪廓模型等影像處理技術，分割出欲分析之細胞核部分，並分離重疊之細胞核，再分別計算細胞核之面積、曲率、緊密度、灰階均勻度等特徵參數，經由與病理科醫師討論之後取得所有特徵正常與異常之間之臨界值，以辨識正常與異常細胞核，並將辨識結果與病理科醫師以肉眼觀察之結果互相比對，計算準確率、敏感度與有效性等數據評估系統之辨識結果，最後將特徵參數與辨識結果建立訓練資料與測試資料，導入支援向量機以評估系統之效能。本系統提供病理科醫師在細胞影像診斷判讀上之輔助功能，減輕病理科醫師在人為判讀上的負擔。實驗結果證實本系統具有不錯的準確率。

According to the statistic data from the Department of Health, Executive Yuan, R.O.C, breast cancer is the fourth leading cancer for Taiwan women. It causes great threat to women’s life in Taiwan. There how to identify the breast cancer as early as possible and remedy efficient is very important to women. The most important clinical feature of breast cancer is the touchable knot in women’s breast. When a woman finds the knot of breast by herself, she should go to accept examination as soon as possible, and if the doctor discovers possible symptoms of breast cancer, a further inspection should be arranged. The doctor could use fine needle aspiration or core needle aspiration to obtain specimens of breast tissue, identify the disease based on cytology and histology and arrange treatment as early as possible.
The main goal of our study is to analyze the cell images of breast cancer by using cytology and image processing techniques. First, nuclei were isolated from the sampled image. Features related to the diagnosis of breast carcinoma were extracted to identify the sampled image whether contains normal or abnormal cells. Therefore, it may assist physicians evaluate the breast carcinoma as early as possible.
A sequences of cell images were acquired from the fine needle smears or core needle biopsy on microscope. We utilize many image processing techniques such as color space transformation, thresholding, contrast enhancement, mathematical morphology, multiple active contour model in order to detect the nucleus. Then we separate the overlapped nuclei, measure some nucleus features such as area, curvature, compactness and texture respectively, and determine the thresholds for normal or abnormal cell, in terms of these features according to pathologist’s discussions. Then we compare the identification results with the macrography result evaluated by pathologist, as the accuracy, sensitivity and specificity in order to evaluate the identification performance of the proposed system. Finally we selected all of the feature parameters as training data and test data, and feed these datas into support vector machine (SVM) to evaluate the performance of the proposed system.
This system provides diagnostic and screening information to the pathologist to decrease the load of the pathologists. The experimental results also reveal that the proposed system can obtain good accuracy.

中文摘要……………………………………………………………i
Abstract……………………………………………………………iii
誌謝……………………………………………………………………v
目錄……………………………………………………………………vi
圖表目錄………………………………………………………………xi
表格目錄……………………………………………………………xiv
第一章 導論……………………………………………………………1
1.1 研究動機……………………………………………………1
1.2 研究背景及相關研究………………………………………3
1.2.1 國外相關研究現況……………………………………3
1.2.2 國內相關研究現況……………………………………5
1.2.3 參考文獻之評述………………………………………6
1.3 研究目的……………………………………………………6
1.4 論文架構……………………………………………………7
第二章 乳房組織細胞病理及相關影像處理技術……………………8
2.1 乳房組織細胞病理…………………………………………9
2.1.1 乳癌的臨床表徵………………………………………9
2.1.2 常見之乳房良性病變………………………………10
2.1.3 正常之乳房切片細胞特徵…………………………12
2.1.4 常見之乳房惡性腫瘤………………………………13
2.1.5 麥瑪通切片取樣技術之步驟………………………14
2.1.6 乳房切片之製作步驟………………………………15
2.2 影像處理技術………………………………………………17
2.2.1 色彩空間……………………………………………17
2.2.1.1 RGB彩色模型…………………………………18
2.2.1.2 HIS彩色模型…………………………………18
2.2.1.3 灰階轉換………………………………………19
2.2.2 影像分割-臨界值法…………………………………20
2.2.3 形態學的修正………………………………………21
2.2.4 相鄰元素標記法……………………………………22
2.2.5 主動式輪廓模型（Active Contour Model）…………24
2.3 支援向量機(Support Vector Machine)…………………27
2.4 評估方式……………………………………………………31
2.5 ROC曲線……………………………………………………33
第三章 研究架構與方法……………………………………………35
3.1 研究影像與設備……………………………………………37
3.1.1 研究影像……………………………………………37
3.1.2 研究設備……………………………………………37
3.2 研究步驟……………………………………………………38
3.2.1 影像分析與前處理…………………………………38
3.2.2 以Otsu分割演算法分割出前景部分………………41
3.2.3 填補細胞核內部……………………………………44
3.2.4 去除多餘細胞質、紅血球與人工雜點……………45
3.2.5 分離重疊細胞核……………………………………47
3.2.5.1 計算重疊細胞核個別之重心…………………48
3.2.5.2 取出交界處之分割點，分割重疊細胞核……52
3.2.6 取得細胞核初始輪廓………………………………55
3.2.7多重主動式輪廓模型………………………………56
3.2.7.1使用主動式輪廓模型之動機…………………56
3.2.7.2主動式輪廓模型………………………………57
3.2.8 特徵參數量測………………………………………62
3.2.8.1輪廓點之曲率…………………………………62
3.2.8.2細胞核面積……………………………………63
3.2.8.3細胞核之緊密度（Compactness）……………64
3.2.8.4細胞核之亮度對比度（Contrast）……………65
3.3 細胞辨識……………………………………………………66
3.4 系統評估……………………………………………………67
第四章 實驗結果與討論……………………………………………69
4.1 細胞核之分割結果…………………………………………70
4.1.1 灰階轉換……………………………………………70
4.1.2 對比強化……………………………………………71
4.1.3 以Otsu分割演算法分離出細胞與背景……………72
4.1.4 去除多餘間質細胞、紅血球與人工雜點部分………72
4.1.5 填滿細胞核內部……………………………………74
4.1.6 分離重疊細胞核……………………………………75
4.1.7 初始輪廓偵測………………………………………77
4.1.8 以主動式輪廓模型逼近出最佳輪廓………………78
4.2 特徵參數量測與評估………………………………………80
4.2.1 輪廓點之曲率………………………………………81
4.2.2 細胞核面積…………………………………………81
4.2.3 細胞核之緊密度(Compactness)……………………82
4.2.4 細胞核之對比度（Contrast）………………………83
4.2.5 辨識結果……………………………………………83
4.3 辨識結果評估………………………………………………87
4.4 系統介面……………………………………………………93
第五章 結論與未來展望……………………………………………94
5.1 結論…………………………………………………………94
5.2 未來展望……………………………………………………95
參考文獻………………………………………………………………96

[1] A. Amini, Weymouth, and T.,Jain, R., “Using Dynamic Programming for Solving Variational Problems in Vision”, IEEE Trans. Pattern Analysis and Machine Intelligence, Vol.12, No.9, pp.855-866.,1992.

[2] R.C. Gonzalez and R.E. Woods, “Digital Image Processing / 2nd edition,” Prentice Hall, 2001

[3] M. Kass, Witkin, and D. Terzopoulos, “Snakes: Active Contour Models.”, International Journal of Computer Vision, Vol.1, pp.321-331, 1987.

[4] K.M. Lee, W. N. Street, “Automatic Image Segmentation and Classification Using On-line Shape Learning,” IEEE, 2000.

[5] M. Mignotte, and J. Meunier, “A Multiscale Optimization Approach for the Dynamic Contour-based Boundary Detection Issue”, Elsevier Science, Computerized Medical Imaging and Graphics, pp.265-275, 2001.

[6] F. Mokhtarian , R. Suomela, “Robust Image Corner Detection Through Curvature Scale Space,”IEEE Trans.on Pattern Analysis and Machine Intelligence, Volume 20, Issue 12, Dec. 1998 Page(s):1376 – 1381

[7] N. Otsu, “A Threshold Selection Method from Gray-Level Histogram”. IEEE Trans. System Man Cybernetics, SMC-9(1): 62-66,1979.

[8] H. Refai, T. K. Teague, “Automatic Count of Hepatocytes In Microscopic Images,” Image Processing, 2003. ICIP 2003. Proceedings. 2003 International Conference on Volume 2, 14-17 Sept. 2003 Page(s):II - 1101-4 vol.3

[9] A. Rosenfeld and E. Johnston, “Angle Detection on Digital Curves,” IEEE Transaction on Computers, Vol. C-22, pp. 875-878, 1973.

[10] A. Rosenfeld and A.C. Kak, “Digital Picture Processing”, Academic Press, NewYork, NY, 2nd edition, 1982.

[11] M. Sammouda, R. Sammouda, N. Niki, N. Yamaguchi, N. Moriyama, “Cancerous nuclei detection on digitized pathological lung color images,”Journal of Biomedical Informatics 35, 92-98, 2002.

[12] H. Shi, R. Ward, “Canny Edge Based Image Expansion”, Circuits and Systems, 2002. ISCAS 2002. IEEE International Symposium on Volume 1, 26-29 May 2002 Page(s):I-785 - I-788 vol.1

[13] L. T. Tan, C. P. Lian, C. J.S. deSilva, “Comparison Between Syntactic Pattern Recognition and The Randomized Hough Transform,” Neural Information Processing, 1999. Proceedings. ICONIP '99. 6th International Conference on Volume 3, 16-20 Nov. 1999 Page(s):873 - 877 vol.3.

[14] S. Taylor, N.C.a.J., An Introduction ro Support Vector Machines and Other Kernel-based Learning Methods. 2000: Cambridge University Press.

[15] J.M. Thijssen, B.J. Oosterveld, and R.F. Wagner, “Gray level transforms and lesion detectability in echographic images,” Ultrason. Imaging, vol.10, no.3, pp. 171-195, July, 1998

[16] J.P. Thiran and B. Macq, “Morphological Feature Extraction for the Classification of Digital Images of Cancerous Tissues,” IEEE Trans. On Biomedical Engineering, Vol. 43, No.10, October 1996.

[17] N. VAN VLIET, “Image Segmentation Applied to Cytology,”Technical Report n°0303, June 2003.

[18] W. H. Worberg, W. N. Street, “Computer-Generated Nuclear Features Compared With Axillary Lymph Node Status and Tumor Size as Indicators of Breast Cancer Survival,” Elsevier Science(USA), 2002.

[19] H.S. Wu, J. Barba, and J. Gil, “A Parametric Fitting Algorithm for Segmentation of Cell Images,” IEEE Trans. On Biomedical Engineering, Vol. 45, No.3, March 1998.

[20] H. Zhou, K.Z. Moa, “Adaptive Successive Erosion-based Cell Image Segmentation for P53 Immunohistochemistry in Bladder Inverted Papilloma,” Engineering in Medicine and Biology Society, 2005. IEEE-EMBS 2005. 27th Annual International Conference of the 2005 Page(s):6484 – 6487

[21] C.C. Chang and C.J. Lin，LIBSVM—A Library for Support Vector Machines，http://www.csie.ntu.edu.tw/~cjlin/libsvm/

[22] 朱娟秀，”乳癌病理診斷”，台灣醫學第五卷第五期，P.552-556，2001。

[23] 巫宗昇，”利用彩色資訊之肝切片電腦影像分割”，成功大學資訊工程研究所碩士論文，1992。

[24] 何炅哲，”心臟核醫影像的左心室輪廓偵測與血流測量”，嘉義大學資訊工程研究所碩士論文，2003。

[25] 林群雄,”使用多類支持向量機混合三角偵測做人臉辨識”，2006電子商務與數位生活研討會論文。

[26] 林儀信，”以組織學為基礎之乳癌細胞分割及辨識”，嘉義大學資訊工程研究所碩士論文，2006。

[27] 張祐育,”子宮內膜異位症之組織影像特徵參數分析”，南台科技大學電機工程研究所碩士論文，2004。

[28] 馮齡儀，”電腦輔助子宮頸抹片異常細胞辨識之初期研究”，中原大學醫學工程研究所碩士論文，2004。

[29] 鐘兆春，方勝雄，”臨床生理學與病理切片技術”，1999。

[30] 鐘國亮，”影像處理與電腦視覺”，2002。

[31] 石蠟切片，http://www.biox.cn/content/20050828/35748.htm

[32] 吳世重，”乳癌簡介”， http://www.cgmh.org.tw/intr/intr4/c8210/QA/BREAST/Breast.htm