研究結果指出，緻密的乳房通常含有較多的結締組織和各種非脂肪組織，而乳房緻密度這項遺傳性特徵，與乳癌發生率有著高相關性，以及能否透過乳房X光攝影或生理檢查發現乳癌，有著比其他遺傳性因子更重大的影響力，基於上述原因，本論文提出基於經驗模態分解的乳房X光片腺體強化方法，以達到有效的自動切割乳房影像腺體，並進而利用相關特徵擷取對乳房影像腺體密度做分類，達到輔助醫師診斷的功用。
在腺體切割以及分類系統的流程上，大致可分為下列步驟。首先，本研究使用快速自適性二維經驗模態分解來強化乳房X光片的腺體部份，此時強化後的影像共存著腺體、皮膚線、以及脂肪三個部份。接著我們利用形態學的方法來去除皮膚線，進而使用k-means演算法來將腺體與脂肪分類、並將脂肪的部份移除，第一次切割結果顯示乳房X光片中腺體的位置。之後我們利用改進過後的區塊成長演算法，使它能夠根據每張乳房X光片的情況，自適性地調整閥值做以自原始影像第二次的腺體切割。
在本論文中，腺體切割結果被賦予一個專家所驗證的分數，每張乳房影像都透過放射科醫師的目視檢查以判斷切割的結果，這部分我們經由專家驗證，有75% 以上的影像符合醫師所判定的標準。而在分類系統的部分，我們針對BIRADS所定義的密度類別，希望能藉由相關特徵擷取做一個分類的動作，根據乳房影像的特性，我們萃取了碎形維度，形態特徵以及紋理特徵。經過主成份分析的轉換後，特徵的維度會減少而且分類的效率也會提高。運用k-nearest neighbor (kNN) 分類器結合以上萃取的特徵，對應到資料應屬的類別，準確率有86%；而藉由類神經網路的分類器則有97%。

Literatures indicate that dense breast in mammography generally contains more glandular tissue and less fatty tissue. The breast density, which is hereditary characteristics, is also related to the chance of getting breast cancer in many studies. Detecting the breast cancer from mammogram is more important than other genetic factors. An effective and automatic segmentation method to determine the glandular tissue from mammogram becomes a fundamental and important issue for further breast density related research. Therefore, this paper proposes an empirical mode decomposition-based mammogram gland enhancement method to perform efficiently automatic segmentation of gland. First, the proposed method uses fast and adaptive bidimensional empirical mode decomposition (FABEMD) to enhance the mammogram for dense glandular tissue, skin lines, and fatty tissue. Second, the skin lines are removed from the image by using morphology technologies. The segmented results are used to be the coordinate locations of the glandular tissue in the mammogram. Third, we adopt k-means algorithm to classify glandular and fatty tissue in the image to determine a threshold. Fourth, we improve a region growing method to adaptively tune the threshold from the original mammogram. The segmented results which get point 4 or 5 occupy 75% in our mammogram database. According to the characteristic of the dense tissue of each BIRADS density category, we extract the fractal dimension, morphology features, and texture features. The experimental results show that the accuracy rate of the PCA+BPN classifier is about 97% which is significantly higher than the 86% accuracy rate of the PCA+kNN classifier.

[1] American College of Radiology (1998), Illustrated Breast Imaging Reporting and Data System BIRADS, Philadelphia: American College of Radiology, 1st edition.
[2] Boyd, N. F., J. W. Byng, R. A. Jong, E. K. Fishell, L. E. Little, A. B. Miller, G. A. Lockwood, and D. L. Tritchler (1995), “Quantitative classification of mammographic densities and breast cancer risks: results from the Canadian National Breast Screening Study,” Journal of the National Cancer Institute, vol. 87, pp. 670-675.
[3] Boyd, N. F., G. A. Lockwood, J. W. Byng, D. L. Tritchler, and M. J. Yaffe (1998), “Mammographic densities and breast cancer risk,” Breast Disease, vol. 10, pp. 113-126.
[4] Byng, J. W., N. F. Boyd, E. Fishell, R. A. Jong, and M. J. Yaffe (1994), “The quantitative analysis of mammographic densities,” Physics in Medicine and Biology, vol. 39, pp. 1629-1638.
[5] Chen Shun-Yu (2004), Multivariate data analysis, Taipei: Hwa Tai Publishing, 3rd edition.
[6] Christoyianni, I., E. Dermatas, and G. Kokkinakis (5-8 Sept., 1999), “Neural classification of abnormals tissue in digital mammography using statistical features of the texture,” International Conference on Electronics, Circuits and Systems, vol. 1, pp. 117-120, Paphos, Cyprus.
[7] Del Carmen, M. G., E. F. Halpern, D. B. Kopans, B. Moy, R. H. Moore, P. E. Goss, and K. S. Hughes (2007), ”Mammographic breast density and race,” American Journal of Roentgenology, vol. 188, no. 4, pp. 1147-1150.
[8] Enderwick, C.Y and Ecangelia Micheli-Tzanakou (30 Oct.-2 Nov.,1997), “Classification of mammographic tissue using shape and texture features,” International Conference on IEEE Engineering in Medicine and Biology Society, vol. 2, pp. 810-813, Chicago, USA.
[9] Gonzalez, R. C. and R. E. Woods (2007), Digital Image Processing, US: Prentice Hall, 3rd edition.
[10] Huang, N. E., Z. Shen, S. R. Long, M. C. Wu, H. H. Shih, Q. Zheng, N. C. Yen, C. C. Tung, and H. H. Liu (1998), “The empirical mode ecomposition and the Hilbert spectrum for nonlinear and nonstationary time series analysis,” Proceedings of the Royal Society of London Series A: Mathematical, Physical and Engineering Sciences, vol. 454, no. 1971 pp. 903–1005.
[11] Harvey, J. A. and V. E. Bovbjerg (2004), “Quantitative assessment of mammographic breast density: relationship with breast cancer risk,” Radiology, vol. 230, pp. 29-41.
[12] Kallenberg, M. G., M. Lokate, C. H. Van Gils, and N. Karssemeijer (2011), “Automatic breast density segmentation: an intergration of different approaches,” Physics in Medicine and Biology, vol. 56, pp. 2715-2729.
[13] Keir, B. and S. Sameer (2002), “Classification of mammographic breast density using a combined classifier paradigm,” Proceedings of Medical Image Understanding and Analysis Conference, pp. 177-180.
[14] Kobatake, H., M. Murakami, H. Takeo, and S. Nawano (1999), “Computerized detection of malignant tumors on digital mammograms,” IEEE Transactions on Medical Imaging, vol. 18, no. 5, pp. 369-378.
[15] Lin, Ching-Hung, Jau-Yu Liau, and Yen-Shen Lu (2009), “Molecular subtypes of breast cancer emerging in young women in Taiwan: evidence for more than just westernization as a reason for the disease in Asia,” Cancer Epidemiology Biomarkers and Prevention, vol. 18, pp. 1807-1814.
[16] Liao Yu-China (2000), “The mass detection in mammography using texture analysis,” Master Thesis, National Cheng Kung University, Tainan, Taiwan.
[17] Liu, Qingqing, Li Liu, Yanli Tan, Jian Wang, Xueyun Ma, and Hairi Ni (15 Oct.- 17 Oct., 2011), “Mammogram density estimation using sub-region classification,” International Conference on Biomedical Engineering and Informatics, vol. 1, pp. 356-359, Shanghai, China..
[18] Muhimmah, I. and R. Zwiggelaar (8-11 Nov., 2007), “Mammographic density classification using multiresolution histogram information,” International Special Topic Conference on Information Technology in Biomedicine, vol. 3, pp. 59-65, Loannina, Greece.
[19] Oliver, A., J. Freixenet, R. Marti, J. Pont, E. Perez, E. R. E. Denton, and R. Zwiggelaar (2008), “A novel breast tissue density classification methodology,” IEEE Transactions on Information Technology in Biomedicine, vol. 12, no. 1, pp. 55-65.
[20] Oza, A. M. and N. F. Boyd (1993), “Mammographic parenchymal patterns: a marker of breast cancer risk,” Epidemiologic Reviews, vol. 15, pp. 196-208.
[21] Peitgen, H-O. and D. Squpe (1988), The Science of Fractal Images, New York: Springer-Veriag, 1st edition.
[22] Pitas, I. (2000), Digital Image Processing Algorithms and Applications, Canada: Wiley-Interscience, 1st edition.
[23] Rilling, G., P. Flandrin, and P. Goncalves (2004), “Empirical mode decomposition as a filter bank,” IEEE Signal Processing Letters, vol. 11, no. 2, pp. 112-114.
[24] Saftlas, A. F. and M. Szklo (1987), “Mammographic parenchymal patterns and breast cancer risk,” Epidemiologic Reviews, vol. 9, pp. 146-174.
[25] Saha, P. K., J. K. Udupa, E. F. Conant, P. Charkraborty, and D. Sullivanmm (2001), “Breast tissue density quantification via digitized mammograms,” IEEE Transactions on Medical Imaging, vol. 20, no. 8, pp. 792-803.
[26] Tabar, L., T. Tot and P. B. Dean (2005), Breast Cancer-The Art and Science of Early Detection with Mammography, Stuttgart: GermanyL Georg Thieme Verlag, 1st edition.
[27] Sarkar, N. and B. B. Chaudhuri (1991), “An efficient approach to estimate fractal dimension of textual images,” Pattern Recognition, vol. 25, no. 9, pp. 1035-1041.
[28] Sharif, B. M. A., R. A. Reza, and F. K. Jesmin (2008), “Fast and adaptive bidimensional empirical mode decomposition using order-statistics filter based envelope estimation,” The European Association for Signal Processing Journal on Advances in Signal Processing, vol. 2008, no. 3, Article ID 728356, pp. 1-18.
[29] Warner , E., G. Lockwood, M. Math, D. Tritchler, and N. F. Boyd (1992), “The risk of breast cancer associated with mammographic parenchymal patterns: a meta-analysis of the published literature to examine the effect of method of classification,” Cancer Detection and Prevention, vol. 16, pp. 63-72.
[30] Yu Nan-Su (2004), “Optimal feature selection: the mass detection in mammograms,” Master Thesis, National Cheng Kung University, Tainan, Taiwan.