乳癌在近年來一直是僅次於子宮頸癌，排名第二位的癌症。每年台灣地區的婦女，約有2200名新發生的乳癌個案，對國內婦女生命已造成相當大的威脅。乳房X光攝影是早期偵測乳癌最有效的方法之一，它在腫塊的偵側已經證實可比婦女自己摸到腫塊提早平均1.7年。因此40至49歲的婦女每兩年至少做一次乳房X光攝影，可以早期發現乳癌、早期治療乳癌和防止乳癌的擴散，及降低乳癌的死亡率。

　　本論文以歐洲MIAS協會所提供的MiniMammographic Database為研究素材，進行乳房腫塊的偵測辨識。整個研究方法步驟包括了腫瘤區塊的分割、特徵值的粹取、特徵值的最佳組合選擇，以及腫瘤的偵測方法。

　　腫瘤區塊的分割可以有效地反應可疑腫瘤區塊的亮度變化，並且去除不必要的背景區域。其次，本研究所共粹取212個特徵值，分別為碎形維度1個、緊密度1個、灰階值強度統計圖4個、空間灰階相關特徵值176個、紋路頻譜16個和紋路特徵編碼14個。為了找出最具有分析價值的特徵值組合，我們以PCA演算法決定應有的特徵值的個數，之後藉由基因演算法決定特徵值的最佳組合選擇。最後腫瘤的偵測方法採用了線性辨識分析和三個不同類型的類神經網路(倒傳遞類神經網路、機率類神經網路與輻射基底函數)為分類器。

　　整體而言，實驗數據顯示基因演算法和逐步性演算法所選取的特徵值組，與四種不同分類器的可能組合的實驗數據{DR, FAR, CR；Detection Rate, False Alarm Rate, Correct Classification Rate}非常接近，這說明二個特徵值選擇法均能有效找到最佳特徵值的組合。再者，將二個特徵值選擇法所得的特徵值組饋入倒傳遞類神經網路與機率類神經網路，脂肪病例的辨識效能都超過95%；在緻密線體病例及脂肪線體病例上，正確分類率則落在80% ~ 90%之間。

　　Breast cancer recently ranks the second leading cause in cancers among women in Taiwan, and only falls behind the Cervical cancer. Every year, near 2200 new cases of breast cancer occur in Taiwan so that it becomes a vast threat to women’s life in Taiwan. Mammography is one of useful screening methods to early detect breast cancer. It has been approved that its effectiveness of finding masses in women’s breast is earlier than that of self detection by 1.7 years. Thus, American Cancer Society recommends that the women with ages from 40 to 49 should take a routine mammography examination every two years and one year afterwards. The goal of this suggestion is “detecting breast cancer early, curing it in time, to prevent the spread of breast cancer and to reduce the mortality of breast cancer.”

　　In this study, the MiniMammographic Database provided by the Mammographic Image Analysis Society is used to evaluate our system. The methods of our system include (1) the segmentation of masses, (2) the extraction of features, (3) the optimal grouping of features, and (4) the detection of masses by LDA (Linear Discriminant Analysis) and three different neural network classifiers (BPN, PNN and RBFN；Back Propagation Neural Network, Probabilistic Neural Network, Radial Basis Function Neural Network).

　　There are 212 features extracted from the segmented part of a suspicious region of a mammogram. They are composed of one from fractal dimension, one from compactness, four generated from gray level histogram, 176 in accordance with spatial gray level dependence, 16 from texture spectrum, and 14 arising from texture feature coding method. For sake of finding the valuable combination of features used in the classification procedure, the genetic algorithm is applied to determine the best candidates. In regard to our experimental results, it clearly indicates that the feature vectors chosen by the genetic algorithm are almost equivalent to those chosen by the stepwise algorithm. Moreover, the effectiveness measures {DR, FAR, CR；Detection Rate, False Alarm Rate, Correct Classification Rate }derived from BPN and those derived from PNN are close to each other under the same learning samples and test samples. For the fatty cases, our system achieves the correct classification rate above 95%; for the dense-glandular and fatty-glandular cases, our system also attains correct classification rates falling into a degree of 80% ~ 90%, respectly. These results supportably validate that the proposed two feature selection methods can effectively seek the optimal combination of features.

中 文 摘 要 I
Abstract II
誌謝 IV
目錄 V
表目錄 VII
圖目錄 VIII

第一章 緒論 1
1.1 研究動機 1
1.2 研究目標 3
1.3 論文系統架構 5

第二章 資料庫說明 8
2.1 乳房X光片資料庫來源 8
2.2 實驗影像選取 13

第三章 腫瘤的自動分割 14
3.1 臨界值分割簡介 14
3.2 Otsu臨界值分割法 15
3.3 熵臨界值分割法 17
3.4 型態學的修正 20

第四章 特徵值擷取與分析 25
4.1 碎形維度 25
4.2 緊密度 26
4.3 灰階值強度統計圖 26
4.4 空間灰階相關特徵值 27
4.5 紋路頻譜 32
4.6 紋路特徵編碼 37

第五章 特徵值選取 44
5.1 主成份分析 44
5.2 基因演算法選取 47
5.3 逐步演算法選取 50

第六章 腫瘤偵測方法 52
6.1 線性辨識分析 52
6.2 類神經網路 54

第七章 實驗結果 58

第八章 結論與未來發展 78

參考文獻 80

【1】洛夫, 林德賽, 郭雅君, 楊雅婷, ‘‘乳房聖經’’

【2】http://www.doh.gov.tw/NewVersion/index.asp

【3】http://www.doh.gov.tw/statistic/index.htm

【4】季瑋珠, 張金堅, “本土醫學資料庫之建立及衛生政策上
之應用”, 1993。

【5】X. Zhou, R. Gordon, “Detection of early breast cancer: an overview and future prospects,” Critical Review in Biomedical Engineering, 17, pp. 203-255, 1989.

【6】K. F. Schmidt, “A better breast test,”Science News, 143, pp. 392-393, 1993.

【7】Newsweek, May 10, p. 64, 1993.

【8】D. J. Vyborny, “Can computers help radiologists read mammograms,” Radiology, 191, pp. 315-317, 1994.

【9】“The national breast a and cervical cancer early detection program,” CDC U.S. Department of health and human services, 1997.

【10】R.H. Gold, L.W. Bassett, “X-ray mammography : History, controversy, and state of art in mammography, thermography and ultrasound in breast cancer detection,” Grune&Straton, 1982.

【11】S.M. Lai, X. Li, W.F. Bischof, “On techniques for detecting circumscribed masses in mammograms,” IEEE Trans. Med. Image, Vol. 8, pp. 377-386, 1990.

【12】D. Brzakovic, X.M. Luo, P. Brzakovic, “An approach to automated detection of tumors in mammography,” IEEE Trans. Med. Image, Vol. 9, pp.233-241, 1990.

【13】F.F. Yin, M.L. Giger, K. Doi, C.D. Metz, C.J. Vyborny, R.A. Schmidt, “Computerized detection of masses in digital mammograms : Analysis of bilateral subtraction images,” Med. Phys. Vol. 18, pp. 955-963, 1991.

【14】I. Christoyianni, E. Dermatas, G. Kokkinakis, “Neural classification of abnormals tissue in digital mammography using statistical features of the texture,” Electronics, Circuits and Systems, 1999. Proceedings of ICECS '99. The 6th IEEE International Conference on , Volume: 1 , 1999.

【15】I. Christoyianni, E. Dermatas, G. Kokkinakis,
“Fast detection of masses in computer-aided mammography,” IEEE Signal Processing Magazine, pp. 54-64. January 2000.

【16】B. Aldrich, M. Desai, “Application of spatial grey level dependence methods to digitized mammograms,” IEEE, 1994.

【17】H.P. Chan, “Image feature analysis and computer-
aided diagnosis in digital radiology : Automated detection of microcalcifications in mammography,” Med. Phy., Vol. 14, No. 4, pp. 538-548, 1987.

【18】A.P. Dhawan, “Enhancement of mammographic features by optimal adaptive neighborhood image processing,” IEEE Trans. Medical Image, Vol. M2-5, No.5, March 1986.

【19】C.M. Kocur, S.K. Rogers, L.R. Myers, T. Burns, M Kabrisky., J.W. Hoffmeister, K.W. Bauer, J.M. Steppe, “Using neural networks to select wavelet features for breast cancer diagnosis,’’ IEEE Engineering in Med. and
Bio., Vol. 15, Issue 3, pp. 95-102, May/June 1996.

【20】H. Kobatake, Y. Yoshinaga, M. Murakai, “Automatic detection of malignant tumors mammogram,” IEEE Trans. Med. Image, Vol. 3, pp. 407-410, 1994.

【21】H. Kobatake, Y. Yoshinaga, ‘‘Detection of spicules on mammogram based on skeleton analysis,’’ IEEE Trans. Med. Image, Vol. 15, Issue 3, pp. 235-245, June 1996.

【22】M. Sameti, R.K. Ward, J. Morgan-Parkes, B. Palcic, ‘‘A method for detection of malignant masses in digitized mammograms using a fuzzy segmentation algorithm,’’ IEEE Engineering in Med. and Bio., Vol. 2, pp. 513-516, Nov. 1997.

【23】L. Hadjiiski, B. Sahiner, Heang-Ping Chan, N. Petrick, M. Helvie, ‘‘Classification of malignant and benign masses based on hybrid ART2LDA approach,’’ IEEE Trans. Med. Image, Vol.18, Issue 12, pp.1178-1187, Dec. 1999.

【24】N.R. Mudigonda, R.M. Rangayyan, J.E. Leo Desautels, ‘‘Gradient and texture analysis for the classification of mammographic masses,’’ IEEE Trans. Med. Image, Vol. 19, No. 10, pp. 1032-1043, Oct. 2000.

【25】B. Verma and J. Zakos, ‘‘A comuter-aided diagnosis system for digital mammograms based on fuzzy-neural and feature extraction techniques,’’ IEEE Trans. Biomedi, Vol. 5, No. 1, pp. 46-54, Mar. 2001.

【26】N. Otsu, ‘‘A threshold selection method from gray-level histogram,’’ IEEE Trans. Syst., Man, Cybern. , Vol. SMC-9, No. 1, pp.62-66, 1979.

【27】J. Kittler and J. Illingwroth, ‘‘On threshold selection using clustering criteria,’’ IEEE Trans. Syst., Man, Cybern. , Vol. SMC-15, No. 5, pp.652-655, 1985.

【28】P.K. Sahoo, S.Soltani and A. K. Wong, ‘‘A survey of thresholding techniques,’’ Comp. Vision, Graphics, Image Proc., Vol. 41, No. 2, pp.233-260, 1988.

【29】N.K. Pal and S.K. Pal, ‘‘Entropic thresholding,’’ Signal Processing. pp.97-108, 1989.

【30】C.Y. Enderwick and Ecangelia Micheli-Tzanakou, ‘‘Classification of mammographic tissue using shape and texture features,’’ Proce.-19th International Conference-IEEE/EMBS Oct. 30-Nov. 2, pp. 810-813, 1997.

【31】H. Kobatake, M. Murakami, H. Takeo and S. Nawano, ‘‘Computerized detection of malignant tumors on digital mammograms,’’ IEEE Trans. Med. Image, Vol.18, No. 5, pp. 369-378, May 1999.

【32】廖有千, ‘‘紋路特徵值分析應用於乳房X光片攝影之腫瘤偵測.’’ 國立成功大學資訊工程系碩士論文, 90年度.

【33】H. P. Chan, D. Wei, M. A. Helvie, B. Sahiner, D. D. Adler, M. M. Goodsitt, and N. Petrick, ‘‘Computer-aided classification of mammographic masses and normal tissue: Linear discriminant analysis in texture feature space,’’ Phys. Med. Biol. Vol. 40, No. 5, pp. 857-876, May 1995.

【34】D. Wei, H. P. Chan, M. A. Helvie, B. Sahiner, D. D. Adler, M. M. Goodsitt, and N. Petrick, ‘‘Classification of mass and normal breast tissue on digital mammograms: Multiresolution texture analysis,’’ Med. Phys. Vol. 22, pp.1501-1513, 1995.

【35】D. F. Specht, ‘‘Probabilistic neural network and the polynomial Adaline as complementary techniques for classification,’’ IEEE Trans. On Neural Networks, pp. 111-121.

【36】D. F. Specht, ‘‘Probabilistic neural networks,’’ Neural Networks, Vol. 3, pp.109-118, 1990.

【37】J. Moody and C. Darken, ‘‘Fast learning in networks of locally-tuned processing units,’’ Neural Comput, Vol. 1, pp.281-294, 1989.

【38】M. J. D. Powell, ‘‘Radial basis function for multivariable interpolation: A review,’’ in Algorithms for Approximation, eds., J. C. Mason and M. G. Cox, Oxford University Press, pp.143-167, 1987.

【39】A. K. Jain, “Fundamentals of Digital Image Processing,” Englewood Cliffs, NJ: Prentice-Hall, 1989.