臺灣博碩士論文加值系統

摘要

乳癌是目前台灣女性所有癌症發病率排名的第四位，對國內婦女造成相當大的威脅。而早期診斷、早期發現與早期治療是減少死亡率以及延長患者存活年限的最佳方法，目前以乳房Ｘ光攝影為早期乳癌診斷最有效方法。美國癌症學會（American Cancer Society）建議40至49歲的婦女每兩年至少需做一次乳房X光攝影檢查，而50歲以上的婦女每年至少做一次乳房X光攝影檢查。
目前已知有百分之五十的乳癌是由微鈣化所形成，而另一半則肇因於腫瘤。在過去，微鈣化偵測已經被廣泛的探討與研究，許多電腦輔助診斷工具也已經被發展出來。相反地，由於腫瘤屬於乳房組織的一部份，所以其偵測的困難度更甚於微鈣化。另一個困難處是腫瘤在乳房X光攝影的特徵表現上，通常比微鈣化來得複雜與多變，不同於微鈣化可以灰階值亮度與對比辨識。乳房腫瘤的偵測必須找腫瘤組織與正常組織之間的差異性，以作為判斷的依據。
本論文中，將研究腫瘤在紋路特徵上的各種不同表現，並進一步分析以達到腫瘤偵測的目的。擺脫傳統上利用灰階值作為診斷依據的方法，採用了不與灰階值強度直接相關的紋路頻譜(Texture Spectrum)和紋路特徵編碼(Texture Feature Coding Method)。
利用各種不同的演算法來描述腫瘤組織和正常組織的特徵值後，再將特徵值用來訓練類神經網路，做為腫瘤偵測的依據。本論文也嘗試以三種類神經網路：倒傳遞類神經網路（Back Propagation Neural Network），輻射基底函數類神經網路（Radial Basis Function Neural Network），和機率類神經網路（Probabilistic Neural Network）來找出最佳的判斷結果。
　　本論文使用的實驗影像資料，是由歐洲 The Mammographic Image Analysis Society (MIAS)所提供的MIAS MiniMammographic Database。這組資料庫中，有161位受試者的左右乳房影像，共322張數位影像。其中有115張影像是經過醫師判讀與病理證實具有腫瘤或鈣化等異常組織，其餘207張影像是健康正常的影像。
根據實驗數據顯示，利用紋路特徵去分析腫瘤影像所做的判斷，可以達到90％的偵測率(Detection Rate)，只有10％的判斷錯誤警示(False Alarm Rate)。由此可知，本論文所使用的紋路特徵非常適合去描述腫瘤影像與正常影像之間的差異性。而利用本論文所提出的判斷方式，可以使系統做出正確的偵測判斷。應用在臨床醫學上，可以輔助醫師對於乳房腫瘤的診斷，提高診斷的正確率，以降低乳癌對於生命的危害。

Abstract

Breast cancer is now the fourth leading cause in cancer among women in Taiwan. According to statistics, one out of eight women may develop breast cancer in their lifetimes. The mortality and incidence rates continue to rise. At the present time, mammography is the only effective means to be used for screening of early breast cancer detection. American Cancer Society recommends that women with ages between 40 and 49 should take mammograms every two years and one year afterwards.
It is known that 50% of breast cancer is generally caused by microcalcifications while the other half resulting from masses. Detection of microcalcifications has been studied and investigated extensively in the past. Many computer-aided diagnostic tools have also been developed. On the other hand, the detection of breast masses has not made as much progress as the detection of microcalcifications. This is due to the fact that detecting masses is much more difficult than detection of microcalcifications. The difficulty arises in that mammographic characteristics of masses are generally more complicated than that of microcalcifications. Unlike microcalcifications which can be specified by intensity and contrast, breast masses are usually described by the tissues’s textures.
This thesis investigates different texture-based approaches to mass detection. Two methods are proposed for mass feature extraction, which are “Texture Spectrum” and “Texture Feature Coding Method”.
After feature extraction and selection, neural networks will be used to detect masses using extracted texture features. Three types of neural networks will be analyzed, back propagation neural network, radial basis function neural network and probabilistic neural network.
The database used to evaluate our system is provided by the Mammographic Image Analysis Society (MIAS). The MIAS MiniMammographic Database is available in the public domain. It contains 322 mammograms with 115 cases of tumors or calcifications that are biopsy-proven and verified by radiologist. The experiment results show that our texture feature-based neural network approaches can achieve a 95% detection rate with a 10% false alarm rate. These methods can be applied clinics for assisting the doctors to diagnose the breast tumor, and raising the diagnostic accuracy.

目錄


中文摘要……………………………………………………………………………Ⅰ
英文摘要……………………………………………………………………………Ⅲ
目錄…………………………………………………………………………………Ⅵ
表目錄………………………………………………………………………………Ⅷ
圖目錄………………………………………………………………………………Ⅸ

第一章　緒論………………………………………………………………………１
1.1 研究動機…………………………………………………………………１
1.2 研究目標…………………………………………………………………５
1.3 系統架構…………………………………………………………………７

第二章　ＭＩＡＳ資料庫…………………………………………………………９

第三章　紋路特徵值擷取與分析………………………………………………１４
3.1 影像前處理……………………………………………………………１４
3.2 灰階值強度統計圖……………………………………………………１６
3.3 空間灰階相關特徵值…………………………………………………１８
3.4 紋路頻譜………………………………………………………………２１
3.4.1黑白對稱性……………………………………………………２２
3.4.2幾何對稱性……………………………………………………２３
3.4.3方向維度………………………………………………………２４
3.4.4方位特徵………………………………………………………２５
3.4.5中央對稱性……………………………………………………２９
3.5 紋路特徵編碼…………………………………………………………３０
　　　 3.5.1粗糙度…………………………………………………………３３
3.5.2同質性…………………………………………………………３４
3.5.3平均收斂值……………………………………………………３４
3.5.4變異量…………………………………………………………３５
3.5.5熵………………………………………………………………３６
3.5.6相似度…………………………………………………………３７
3.5.7規律性…………………………………………………………３７

第四章　腫瘤偵測實驗結果……………………………………………………４０
4.1 類神經網路分類………………………………………………………４０
4.2 實驗資料取樣…………………………………………………………４４
4.3 特徵值選取……………………………………………………………４５
4.4 實驗結果………………………………………………………………４７

第五章　結論……………………………………………………………………５８

參考文獻…………………………………………………………………………６０

附錄………………………………………………………………………………６４

參考資料

[1] http://www.doh.gov.tw/90-info/info-5.htm
[2] http://www.doh.gov.tw/newdoh/90-org/org-10/org10-2/89/org10-89.htm
[3] 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
[4] http://www.fmmu.edu.cn/web/hepa/text/brestca.htm
[5] 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.
[6] “The national breast a and cervical cancer early detection program”, CDC U.S. Department of health and human services, 1997.
[7] 陳啟明, 雷永耀, 彭芳谷, “乳房疾病”, 九州圖書文物有限公司, 1982
[8] L.W. Bassett, “Mammographic analysis of calcifications”, Radiological Clinics of North America, Vol. 30, No.1, pp. 93-105, 1992
[9] 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
[10] 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
[11] H. Kobatake, Y. Yoshinaga, M. Murakai, “Automatic detection of malignant tumors mammogram”, IEEE Trans. Med. Image, Vol. 3, pp. 407-410, 1994.
[12] 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.
[13] W.P. Kegelmeyer, J.M. Prunedu, P.D. Bourland, A. Hillis, M.W. Riggs. M.L. Nipper, “Computer-aided mammographic screening for speculated lesions”, Radiol. Vol. 191, pp. 331-337, 1994.
[14] H.P. Chan, D. Wei, M.A. Helvie, B. Sahiner, D.D. Adler, M.M. Goodsitt, N. Petrick, “Computer-aided classification of mammographic masses and normal tissue : Linear discriminant analysis in texture feature space”, Phys. Med. Biol. Vol. 40, pp. 857-876, 1995.
[15] 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.
[16] I. Christoyianni, E. Dermatas, G. Kokkinakis, “Fast detection of masses in computer-aided mammography” IEEE Signal Processing Magazine, pp. 54-64. January 2000.
[17] B. Aldrich, M. Desai, “Application of spatial grey level dependence methods to digitized mammograms”, IEEE, 1994.
[18] 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, Jul/Aug 1987.
[19] A.P. Dhawan, “Enhancement of mammographic features by optimal adaptive neighborhood image processing”, IEEE Trans. Medical Image, Vol. M2-5, No.5, March 1986.
[20] R.M. Haralick, K. Shanmugam, I. Dinstein, “Textural features for image classification”, IEEE Transaction on system, mam, and cybernetics, Vol. SMC-3, No. 6, November 1973.
[21] D.C. He, L. Wang, “Texture features based on texture spectrum”,Pattern Recognition, Vol.24, No.5, pp.391-399, 1991.
[22] D.C. He, L. Wang, J. Guibert, “Texture discrimination based on an optimal utilization of texture features”, Pattern Recognition, Vol. 24, No. 2, pp. 141-146, 1988.
[23] M.H. Horng, Y.N. Sun, X.Z. Lin, “Texture feature coding method for classification of liver sonography” Computerized Medical Imaging and Graphics 26, pp 33-42, 2002.
[24] http://www.wiau.man.ac.uk/services/MIAS/MIASweb.html
[25] D.C. He, L. Wan, “Texture feature extraction from texture spectrum”, IEEE, pp.1987-1990.
[26] P.C. Chung “An algorithm for detection and segmentation of clustered microcalcifications on mammograms,” 2nd Medical Engineering Week of the World. May 26-30, Taipei, Taiwan, ROC. pp. 102, 1996.
[27] C.M. Wu, Y.C. Chen. K.S. Hsieh, “Texture feature for classification of ultrasonic liver images”, IEEE Tans. Med. Imaging, Vol. 11, No. 2, pp. 141-152, 1992.
[28] L. Wang, D.C. He, “A new statistical approach for texture analysis”, Photogrammetric Engng Remote Sensing 56, pp. 61-66, 1990
[29] J.M. Coggins, A.K. Jain, “A spatial filtering approach to texture analysis”, Pattern Recognition Lett., No. 3, 195-203, 1985.
[30] J.W.V. Miller, J.B. Farson, Y. Shin, “Spatially invariant image sequences” IEEE Trans, Image Processing, Vol. 1, pp.148-161, Apr. 1992.
[31] N.K. Pal, “Entropic thresholding” Signal processing. 16, 97-108, 1989
[32] G. Cardenosa, “Breast imaging companion” Lippincott-Raven, New York, 1997
[33] D.F. Specht, ”Probabilistic neural networks and the polynomial Adaline as complementary techniques for classification” IEEE Trans. On Neural Networks , 111-121.
[34] D.F. Specht, “Probabilistic neural networks ,” Neural Networks ,3,109-118, 1990.