摘要
　　乳房腫瘤通常是乳癌的重要徵兆，為了幫助放射診斷科醫師進行早期乳房腫瘤的篩檢，電腦輔助診斷工具的開發已經引發研究人員相當高的興趣。本論文主要探討乳房X光攝影與乳房核磁共振影像之乳房腫瘤電腦輔助診斷系統包含篩檢、分類與定位功能的建立與技術，以協助醫師做早期診斷與治療。論文中所提出的完整系統主要包括四項子系統，分別為：乳房X光影像電腦輔助腫瘤篩檢系統，提供乳房腫瘤偵測功能；乳房核磁共振影像組織分類與對比增強系統，輔助與提高醫師腫瘤篩檢的準確性；乳房X光影像電腦輔助腫瘤良惡性分類系統，輔助醫師進行腫瘤良惡性判斷；乳房病灶三維定位系統，提供切片手術細針定位之位置導引。相信藉由這四項子系統的整合，對於提昇臨床乳癌早期診斷的準確度能有很大的幫助。
　　在第一個子系統中，我們採用三組與腫瘤紋路相關的特徵群，分別為空間灰階相依(Spatial Gray Level Dependence, SGLD)、紋路頻譜(Texture Spectrum, TS)與紋路特徵編碼(Texture Feature Coding Method, TFCM)。總計十九種特徵值被使用來描述數位乳房X光影像上腫瘤與正常區域組織的特性。緊接著，在分類器的監測下，三種特徵選取法: 連續向後選取(Sequential Backward Selection, SBS)、連續向前選取(Sequential Forward Selection, SFS) 與連續浮動式搜尋法(Sequential Floating Search Method, SFSM) 被用來從十九種特徵值中挑選出次最佳子集合，以達到提升腫瘤偵測性能的目的。最後，系統的性能將在比較機率類神經網路(Probabilistic Neural Network, PNN)與支援向量機器(Support Vector Machine, SVM)兩種分類器後達到最佳狀態。
　　在最近的乳房檢查技術中，乳房核磁共振影像已經開始受到相當程度的注重。由於傳統顯影劑注射而得到的對比增強乳房核磁共振影像已經被證實在乳癌的偵測上具有高敏感度，第二個子系統將採用一個稱為卡門線性混合模型濾波法的頻譜偵測技術，來有效的將乳房核磁共振影像分類出四種主要組織，並以組織分離的高對比影像呈現出來，以替代傳統的顯影劑注射方式。除了期望減少病人因注射所造成的痛苦，亦可降低醫療資源的耗費。為了驗證系統的效能表現，我們使用了電腦虛擬假體與真實的乳房核磁共振兩種影像在與c-means (CM) 演算法的比較下進行一連串的實驗。系統的輸出結果在經過比較c-means (CM) 演算法與劑注射而得的對比增強乳房核磁共振影像之後，顯示本系統在頻譜偵測技術下所產生的組織分離高對比影像優於前面兩種影像。
　　因為形狀特徵是分辨良性與惡性腫瘤的重要指標。第三個子系統首先運用熵值門檻演算法在第一個子系統或第二個子系統所偵測到的腫瘤可疑區域中將腫瘤切割出來。然後，四種形狀特徵將由這些被分割出來的腫瘤中抽取出來，並輸入到機率類神經網路中以完成良惡性分類。為了瞭解系統的可用性，我們使用在台中榮民總醫院收集的資料庫影像來從事系統效能的評估，其結果顯示本系統具有高度的準確性。
　　在第四個子系統中，被偵測出的乳房病灶在經過一連串的計算與擠壓的校正之後計算出以乳頭為參考點的立體定位座標。然後，透過虛擬實境模組語言(virtual reality modeling language viewer, VRMLV)，病灶將以立體模型方式呈現在瀏覽器上以供醫師執行針吸生檢定位切片手術的初步指引。為了驗證系統所計算的病灶立體座標的正確性，我們收集了一組具有病灶同時出現在乳房X光攝影與乳房核磁共振影像的影像樣本。因此，實際的病灶立體座標將可由乳房核磁共振影像獲得確認，並做為進ㄧ步系統正確率評估之用。
　　對於電腦輔助醫療技術而言，本論文所提出的理論不僅驗證了運用醫學影像電腦輔助系統診斷早期乳癌的可能性，也提供了未來幾年內電腦輔助醫療系統發展之雛型。

Abstract
　Breast masses generally present as the major symptom of breast cancer. In order to assist radiologists in detecting masses in the early stage of breast cancer, it is highly desirable to develop a reliable computer aided diagnostic system as an assistant. This dissertation presents a computer-aided diagnostic system for mass screening, classification and localization, which performs mass screening in both of mammograms and the breast magnetic resonance imaging (MRI) followed by the classification and localization on those detected masses. Entire proposed system is consisted of four subsystems, including the mass screening system in mammograms, the contrast enhancement system for mass screening in breast MRIs, the classification system in detected masses, and the 3-D locating system of detected masses using cranio-caudal and medio-lateral oblique views.
　In the first subsystem, three groups of characteristics related to mass texture are adopted, namely, SGLD (Spatial Gray Level Dependence), TS (Texture Spectrum) and TFCM (Texture Feature Coding Method). Totally 19 texture features are offered to describe the characteristics of masses and normal textures on digitized mammograms. Next, under the testing by classifiers, three Feature Selection Methods--SBS (Sequential Backward Selection), SFS (Sequential Forward Selection) and SFSM (Sequential Floating Search Method) are used to find out suboptimal subset from 19 features in order to improve the performance of mass detection. Finally, the performances are compared when two classifiers—PNN (Probabilistic Neural Network) and SVM (Support Vector Machine) are applied--to reach the optimum performance.
　Among the most recent techniques of breast examination, a great attention is being paid to breast MRIs. Since the contrast-enhanced breast MRIs acquired by traditional contrast-injection has shown to be very sensitive in the detection of breast cancer, the second subsystem adopts a spectral signature detection technology, Kalman Filter-based Linear Mixing Method (KFLM), which could successfully classify breast MRIs into four major tissues and present the classified results in high contrast tissue-separated images. A series of experiments using real MRIs and phantoms are conducted and compared to the commonly used c-means (CM) method for performance evaluation. After compare with CM algorithm and contrast-injected breast MRIs, the results showed that the high contrast images generated by spectral signature detection technologies had a superior quality.
　In the third subsystem, the entropic thresholding technique is developed for mass segmentation from ROI (region of interest) of masses detected by first or second subsystem. Since the shape of masses is crucial in classification between benignancy and malignancy, then four shape features extracted from segmented masses are implemented in PNN for mass classification. To evaluate our designed system a data set collected from Taichung Veteran General Hospital, Taiwan, R.O.C. was used for performance evaluation. The results are encouraging and had shown to be promising in our system.
　For the fourth subsystem, the 3D localization is determined by a sequence of coordinate corrections of detected lesions using the breast nipple as a controlling point. Finally, the 3D visualization implements a virtual reality modeling language viewer (VRMLV) to view the exact location of the lesion as a guide for needle biopsy. In order to validate our proposed 3D localization system, a set of breast lesions, which appear both in mammograms and in MR Images is used for experiments where the depth of breast lesions can be verified by the MR images.
　In summary, this dissertation provides the odds for diagnosis of masses by using computer-aided diagnostic system in breast medical images. It also provides a prototype and applications of computer-aided diagnostic system for the future several years.

References
[1]K.F. Schmidt, “A better breast test,” Science News 143, pp. 392-393, 1993.
[2]The statistics of Department of Health, Executive Yuan, Taiwan ROC, http://www.doh.gov.tw/.
[3]Newsweek, pp. 64-64, May 10, 1993.
[4]D.J. Vyborny, “Can computers help radiologists read mammograms,” Radiology, vol. 191 (1994) 315-317.
[5] The National Breast and Cervical Cancer Early Detection Program, NBCCEDP, At-A-Glance, 1997 CDC U.S. Department of Health and Human Services.
[6]C.J. Vyborny, “can computers help radiologists read mammograms,” Radiology 191, 315-317 (1994).
[7]S.M. Lai, X.B. Li and W. F. Bischof, “On techniques for detecting circumscribed masses in mammograms,” IEEE Trans. Med. Imag., vol. 8, no. 4, pp. 377-386, 1989.
[8]D. Brzakovic, X.M. Luo and P. Brzakovic, “An approach to automated detection of tumors in mammograms,” IEEE Trans. Med. Imag., vol. 9, no. 3, pp. 233-241, 1990.
[9]W.P. Kegelmeyer Jr., J.M. Prunedu, P.D. Bourland, A. Hillis, M.W. Riggs And M.L. Nipper, “Computer-aided mammographic screening for spiculated lesions,” Radiology, vol. 191, no. 2, pp. 331-337, 1994.
[10]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,” Physical Medical Biology, vol. 40, no. 5, pp. 857-876, 1995.
[11]D. Wei, H.P. Chan, M.A. Helvie, B. Sahiner, N. Petrick, D.D. Adler and M.M. Goodsitt, “Classification of mass and normal breast tissue on digital mammograms: Multiresolution texture analysis,” Med. Phys., vol. 22, no. 9, pp. 1501-1513, 1995.
[12]I. Christoyianni, E. Dermatas and G. Kokkinakis, “Fast Detection of Masses in Computer-Aided Mammography,” Signal Processing Magazine, IEEE, vol. 17, no. 1, pp. 54-64, 2000.
[13]K. Bovis and S. Singh, “Detection of masses in mammograms using texture features,” Proceedings of 15th International Conference on Pattern Recognition 2, pp. 267-270, 2000.
[14]M. Sameti, R.K. Ward, B. Palcic and J. Morgan-Parkes, “Texture feature extraction for tumor detection in mammographic images,” IEEE Pacific Rim Conference on Communications, Computers and Signal Processing 2, pp. 831-834, 1997.
[15]B. Aldrich and M. Desai, “Application of Spatial Grey Level Dependence Methods to Digitized Mammograms,” Proceedings of the IEEE Southwest Symposium on Image Analysis and Interpretation, pp. 100-105, 1994.
[16]D.C. He and L. Wang, “Texture features based on texture spectrum,” Pattern Recognition, vol. 24, no. 5, pp. 391-399, 1991.
[17]D.C. He, L. Wang and J. Guibert, “Texture discrimination based on an optimal utilization of texture features,” Pattern Recognition, vol. 24, no. 2, pp. 141-146, 1988.
[18]M.H. Horng, Y.N. Sun and X.Z. Lin, “Texture feature coding method for classification of liver sonography,” Computerized Medical Imaging and Graphics, vol. 26, pp. 33-42, 2002.
[19]C.J.C. Burges, ”A tutorial on Support Vector Machines for Pattern Recognition,” Data Mining and Knowledge Discovery 2, pp. 121-167, 1998.
[20]C. Campbell, “Algorithmic Approaches to Training Support Vector Machines: A Survey,” ESANN’2000 proceedings, pp. 27-36, 2000.
[21]P. Pudil, F.J. Ferri, J. Novovicova and J. Kittler, “Floating search methods for feature selection with nonmonotonic criterion functions,” Proceedings of the 12th IAPR International Conference on Pattern Recognition, vol. 2, pp. 279-283, 1994.
[22]P. Pudil, K. Fuka, K. Beranek and P. Dvorak, “Potential of artificial intelligence based feature selection methods in regression models,” Proceedings of Computational Intelligence and Multimedia Applications, ICCIMA '99, pp. 159-163, 1999.
[23]S.J. Reeves and Z. Zhe, “Sequential Algorithms for Observation Selection,” IEEE Transaction on Signal Processing, vol. 47, no. 1, pp. 123-132, 1999.
[24]BI-RADS Breast Imaging Reporting and Data System – Magnetic Resonance Imaging, First Edition, American College of Radiology, 2003.
[25]C.K. Kuhl, R.K. Schmutzler, C.C. Leutner, A. Kempe, E. Wardelmann, A. Hocke, M. Maringa, U. Pfeifer, D. Krebs and H. H. Schild, “Breast MR Imaging Screening in 192 Women Proved or Suspected to Be Carriers of a Breast Cancer Susceptibility Gene: Preliminary Results1,” Radiology, vol. 215, pp. 267-279, 2000.
[26]S. Heywang-Kobrunner and R. Beck, “Contrast-enhanced MRI of the breast,” 2nd edition, Springer-Verlag, 1996.
[27]C.I. Chang, and C. Brumbley, "A Kalman filtering approach to multispectral image classification and detection of changes in signature abundance," IEEE Trans.on Geoscience and Remote Sensing, vol. 37, pp. 257-268, Jan. 1999.
[28]S. Haykin, “Adaptive Filtering Theory,” 3rd ed. Englewood Cliffs, NJ: Prentice-Hall, 1996.
[29]C.M. Wang, C.C. Chen, Y.N. Chung, S.C. Yang, “Detection of Spectral Signatures in Multispectral MR Images for Classification,” IEEE Trans. on Medical Imaging, vol. 12, no. 1, pp. 50-61, Jan. 2003
[30]C.M. Wang, C.C. Chen, S.C. Yang, “An Unsupervised Orthogonal Subspace Projection Approach to MR Image Classification,” Optical Engineering, vol. 41, pp. 1546-1557, July. 2002
[31]C.M. Wang, S.C. Yang, “Orthogonal Subspace Projection-Base Approaches to Classification of MR Images Sequences,” Computerized Medical Imaging and Graphics, vol. 25, no. 6, pp. 465-476, Dec. 2001.
[32]BI-RADS Breast Imaging Reporting and Data System – Mammography, Fourth Edition, American College of Radiology, 2003.
[33]L. Shen, R.M. Rangayyan, J.E.L. Desautels, “Application of shape analysis to mammographic calcifications,” IEEE Tran. on Medical Imaging, Vol. 13, Issue 2, pp. 263-274, June 1994.
[34]R.M. Rangayyan, N.R. Mudigonda, J.E.L. Desautels, “Boundary modelling and shape analysis methods for classification of mammographic masses,” Med. Biol. Eng. Comput., vol. 38, pp. 487-496, 2000.
[35]R.M. Rangayyan, et al., “Measures of acutance and shape for classification of breast tumors,” IEEE Trans. Med. Imag., vol. 16, no. 6, pp. 799-810, 1997.
[36]L. Shen, R.M. Rangayyan, “Detection and classification of mammographic calcifications,” Int. J. Pattern Recogn., vol. 7, no. 6, pp. 1413-1416, 1993.
[37]T.M. Nguyen, R.M. Rangayyan, “Fractal analysis of contours of mammographic masses,” Proceedings of the IASTED International Conference on Biomedical Engineering, pp. 186-191 (Innsbruck, Austria), Feb. 16-18, 2005.
[38]B.S. Sahiner, et al., “Improvement of mammographic mass characterization using spiculation measures and morphological features,” Med. Phys., vol. 28, no. 7, pp. 1455-1465, 2001.
[39]L.T. Niklason, B.T. Christian, L.E. Niklason, D.B. Kopans, P.J. Slanetz, D.E. Castleberry, B.H. Opsahl-Ong, C.E. Landberg, and B. Giambattist, “Digital breast tomosynthesis: potentially a new method for breast cancer screening,” in Proc. 4th Int. Workshop Digital Mammography, Nijmegen, The Metherlands, pp. 51-56, 1998.
[40]A.D. Maidment, E.F. Conant, S.A. Feig, C.W. Piccoli, and M. Albert, “3-dimension analysis of breast calcifications,” Proc. 3rd Int. Workshop Digital Mammography, Chicago, IL, pp. 245-250, 1996.
[41]Y.H. Chang, W.F. GOOD, J.H. Sumkin, B. Zheng and D. Gur, “Computerized localization of breast lesion from two views an experimental comparison of two methods,” Investigative Radiology, vol. 34, no. 9, pp. 585-588, 1999.
[42]B.H. Cho, J.H. Woo, W.K. Mun, I.Y. Kim and S.I. Kim, “The localization and visualization of breast lesion in digitized Mammogram,” Proceedings of the 22th Annual EMBS International Conference, Chicago IL, pp. 23-28, 2000.
[43]M. Yam, M. Brady, R. Highnam, C. Behrenbruch, R. English and Y. Kita, “Three-dimensional reconstruction of microcalcification clusters from two mammographic views,” IEEE Trans. on Med. Imag., vol. 20, pp. 479-489, 2001.
[44]Y. Kita, E. Tohno, R.P. Highnam and M. Brady, “A CAD system for the 3D location of lesions in mammograms,” Medical Image Analysis, vol. 6, pp. 267-273, 2002.
[45]http://www.wiau.man.ac.uk/services/MIAS/MIASweb.html
[46]C.C. Chang, and C.J. Lin, LIBSVM: a library for support vector machines, software available at http://www.csie.ntu.edu.tw/~cjlin/libsvm, (2005).
[47]R. Schalkoff, Pattern Recognition: Statistical, Structural and Neural Approaches. New York: Wiley, 1992.
[48]R.O. Duda and P.E. Hart, “Pattern Classification and Scene Analysis,” New York: Wiley, 1973.
[49]C.I. Chang and H. Ren, “An experiment-based quantitative and comparative analysis of hyperspectral target detection and image classification algorithms,” IEEE Trans. Geosci. Remote Sensing, vol. 38, pp. 1044-1063, Mar. 2000.
[50]C.I. Chang, H. Ren, Q. Du, S.S. Chiang, and A. Ifarraguerri, “An ROC analysis for subpixel detection,” presented at the IEEE 2001 International Geoscience and Remote Sensing Symp., Sydney, Australia, July 24-28, 2001.
[51]C.E. Metz, “ROC methodology in radiological imaging,” Investigat Radiol., vol. 21, pp. 720-723, 1986.
[52]N.R. Pal and S.K. Pal, “Entropic thresholding,” Signal processing, vol. 16, pp. 97-108, 1989.
[53]C.S. LO, P.C. Chung, S.K. Lee, C.I. Chang, T. Lee, G..C. Hsu and C.W. Yang, “Off-line mammography screening system embedded with hierarchically-coarse-to-fine techniques for the detection and segmentation of clustered microcalcifications,” IEICE Trans. on Information and Systems, E83-D, no. 12, pp. 2161-2173, 2000.
[54]T. Cover and J. Thomas, “Elements of Information Theory,” New York: John Wiley and Sons, 1991.
[55]P.S. Liao, B.C. Hsu, C.S. Lo, P.C. Chung, T.S. Chen, S.K. Lee, L. Cheng and C.I. Chang, “Automatic Detection of Microcalcifications in Digital Mammograms by Entropy Thresholding,” Proceeding of 18th Annual International Conference IEEE Engineering in Medicine and Biology Society, pp. 88-89, October 31-November 3, 1996.
[56]G. Cardenosa, “Breast Imaging Companion 2/ed,” Williams & Wilkins, 2001.
[57]D.B. Kopans, “Breast Imaging 2/ed,” Lippincott-Raven, 1998.
[58]D.B. Kopans, “Standardized mammography reporting,” Radiologic Clinics of North America, vol. 30, no. 1, pp. 257-264, 1992.
[59]C.J. D’Orsi and D.B. Kopans, “Mammographic feature analysis,” Semin Roentgenol, vol. 28, pp. 204-230, 1993.
[60]N. Petrick, H.P. Chan, B. Sahiner and D. Wei, “An Adaptive Density-Weighted Contrast Enhancement Filter for Mammographic Breast Mass Detection,” IEEE Trans. on Med. Imag., vol. 15, no. 1, pp. 59-67, 1996.
[61]J. Kraus and P.P. Deimel, “Full-Width-Half-Maximum and Confinement of Optical Modes in Multiple-Quantum-Well Laser Structure,” Journal of LightWave Technology , IEEE, vol. 11, no. 11, pp. 1802-1805, 1993.
[62]R. Chandrasekhar, and Y. Attikiouzel, “A simple method for automatically locating the nipple on mammograms,” IEEE Trans. on Med. Imag., vol. 16, no. 5, pp. 483-494, 1997.