民國101年衛生福利部公佈的10大死因當中，發現惡性腫瘤已經長達30年居首。當中乳癌則排名女性死亡癌症的第4名，顯示乳癌的防範與治療還有很大的努力空間。因此，我們的目的是想透過紋路分析，建立乳癌的電腦輔助診斷方法。
　　可以透過動態磁振造影(DCE-MRI)，得到每一個voxel的動態曲線，接著將動態曲線分類到幾種不同的曲線型態，曲線型態的類別則是由自身的動脈輸入函數推導而得。比起一般的紋路分析，本研究將影像中每個voxels的灰階取代成曲線型態，再進行紋路特徵的計算。
　　本研究基於68名乳房腫瘤病患，共有81個腫瘤，當中包含31個良性腫瘤與50個惡性腫瘤。本文提出的方法將產生42個紋路特徵與14個基本統計特徵。紋路分析的結果顯示，未移除背景時，最佳分類精度為0.74；移除背景影響後，最佳分類精度提升為0.77。使用2個輸入來訓練類神經網路時，可以進一步提升最佳分類精度達0.83。

Cancer has been the number one cause of death in Taiwan for thirty years. Currently, breast cancer ranks number 4 on the list of all cancer deaths in women. Apparently, there is plenty of room for improvement in preventing and treating breast cancer. Hence, via texture analysis, the goal of this study is to develop a computer-aided-diagnostic method for breast cancer.
　　Via images acquired by dynamic contrast-enhanced MRI (DCE-MRI), the kinetic curve of each voxel is divided into different levels according to its correlation with the arterial input function. Compared to the conventional texture analysis method that studies the variation pattern of the gray level of the image voxels, this work replaces the gray level by the level of the kinetic curve associated with the image voxels.
　　The proposed texture analysis method is applied to 68 subjects suffered from breast tumors. Among the 81 tumors of these patients, 31 are benign and 50 are malignant. The proposed approach generates 42 texture features and 14 statistical features. With the background retained, the best classification accuracy is 0.74. The accuracy can be improved to 0.77 by removing the background. By using two features as the neural network inputs, the classification accuracy can be further improved to 0.83.

誌謝 ii
摘要 iii
Abstract iv
目錄 vi
圖目錄 viii
表目錄 ix
第一章 緒論 1
1.1 研究背景 1
1.2 研究目的 5
1.3 乳房腫瘤診斷技術[4-7] 5
1.4 文獻回顧 9
1.5 論文架構 10
第二章 分析方法與原理 11
2.1 紋路分析(Texture Analysis)[8] 11
2.1.1 灰階共生矩陣(GLCM) 11
2.1.2 紋路特徵 13
2.2 門檻值演算法(Threshold Algorithms) 16
第三章 分析方法與步驟 18
3.1 資料庫來源 18
3.2 資料收集 19
3.3 Kinetic Curve 21
3.4 Arterial Input Function 21
3.5 Curve Pattern 22
3.6 Color coding 23
3.7 Color codes的紋路分析 24
3.8 移除背景影響 25
3.9 3維紋路特徵計算(萃取) 26
3.10 基本統計 29
第四章 結果與討論 31
4.1 紋路分析結果 32
4.1.1 不移除背景影響 35
4.1.2 移除背景影響 40
4.2 基本統計結果 41
4.3 綜合分類效果 45
4.4 類神經網路分類結果 51
第五章 結論與未來展望 53
參考文獻 55
附錄 57

[1] 衛生福利部統計處. (2012). 民國101年主要死因分析. Available: http://www.mohw.gov.tw/cht/DOS/DisplayStatisticFile.aspx?d=13689&s=1
[2] 癌症防治組. (2011). 民國100年癌症登記報告. Available: http://www.hpa.gov.tw/BHPNet/Web/Stat/StatisticsShow.aspx?No=201404160001
[3] 台灣癌症登記中心. Available: http://tcr.cph.ntu.edu.tw/main.php?Page=A1
[4] 乳癌檢查與知識網. Available: http://www.mdesign.tw/display/breast/index.php
[5] 乳癌防治基金會. Available: http://www.breastcf.org.tw/
[6] 林惠文, 顏啟華, and 應宗和‧「乳癌的篩檢與診斷」‧基層醫學，21卷，頁352-358，200612。
[7] 黃獻樑 and 程劭儀‧「乳癌的篩檢」‧基層醫學，21卷，頁225-228，200608。
[8] R. M. Haralick, K. Shanmugam, and I. H. Dinstein, "Textural Features for Image Classification," Systems, Man and Cybernetics, IEEE Transactions on, vol. SMC-3, pp. 610-621, 1973.
[9] B. Aldrich and M. Desai, "Application of spatial grey level dependence methods to digitized mammograms," in Image Analysis and Interpretation, 1994., Proceedings of the IEEE Southwest Symposium on, 1994, pp. 100-105.
[10] V. Goldberg, A. Manduca, D. L. Ewert, J. J. Gisvold, and J. F. Greenleaf, "Improvement in specificity of ultrasonography for diagnosis of breast tumors by means of artificial intelligence," Med Phys, vol. 19, pp. 1475-81, Nov-Dec 1992.
[11] B. S. Garra, B. H. Krasner, S. C. Horii, S. Ascher, S. K. Mun, and R. K. Zeman, "Improving the distinction between benign and malignant breast lesions: the value of sonographic texture analysis," Ultrason Imaging, vol. 15, pp. 267-85, Oct 1993.
[12] M. Sameti, R. K. Ward, J. Morgan-Parkes, and B. Palcic, "A method for detection of malignant masses in digitized mammograms using a fuzzy segmentation algorithm," in Engineering in Medicine and Biology Society, 1997. Proceedings of the 19th Annual International Conference of the IEEE, 1997, pp. 513-516 vol.2.
[13] C. Ruggiero, F. Bagnoli, R. Sacile, M. Calabrese, G. Rescinito, and F. Sardanelli, "Automatic recognition of malignant lesions in ultrasound images by artificial neural networks," in Engineering in Medicine and Biology Society, 1998. Proceedings of the 20th Annual International Conference of the IEEE 1998, pp. 872-875.
[14] I. Christoyianni, E. Dermatas, and G. Kokkinakis, "Neural classification of abnormal tissue in digital mammography using statistical features of the texture," in Electronics, Circuits and Systems, 1999. Proceedings of ICECS '99. The 6th IEEE International Conference on, 1999, pp. 117-120 vol.1.
[15] I. Christoyianni, E. Dermatas, and G. Kokkinakis, "Fast detection of masses in computer-aided mammography," Signal Processing Magazine, IEEE, vol. 17, pp. 54-64, 2000.
[16] B. J. Woods, B. D. Clymer, T. Kurc, J. T. Heverhagen, R. Stevens, A. Orsdemir, et al., "Malignant-lesion segmentation using 4D co-occurrence texture analysis applied to dynamic contrast-enhanced magnetic resonance breast image data," J Magn Reson Imaging, vol. 25, pp. 495-501, Mar 2007.
[17] W. Chen, M. L. Giger, H. Li, U. Bick, and G. M. Newstead, "Volumetric texture analysis of breast lesions on contrast-enhanced magnetic resonance images," Magn Reson Med, vol. 58, pp. 562-71, Sep 2007.
[18] K. Nie, J. H. Chen, H. J. Yu, Y. Chu, O. Nalcioglu, and M. Y. Su, "Quantitative analysis of lesion morphology and texture features for diagnostic prediction in breast MRI," Acad Radiol, vol. 15, pp. 1513-25, Dec 2008.
[19] S. C. Agner, S. Soman, E. Libfeld, M. McDonald, K. Thomas, S. Englander, et al., "Textural kinetics: a novel dynamic contrast-enhanced (DCE)-MRI feature for breast lesion classification," J Digit Imaging, vol. 24, pp. 446-63, Jun 2011.