臺灣博碩士論文加值系統

乳癌是女性癌症中最常出現的一種，超過40 歲的女性得到乳癌將有極高的致死率。如能提早發現便可提早治療以減低乳癌的致死率。
現今有一些技術被用於乳癌診斷， 例如： 乳腺X 光攝影檢查術
（mammography），核磁共振造影（magnetic resonance imaging, i.e., MRI）與超音波技術（sonograms, i.e., ultrasound images）。由於超音波影像技術有3 大優點：方便性，非侵入性與較低成本花費。所以超音波影像技術已成為現今罹患乳癌高危險女性群的主要乳癌掃描方法之一。超音波乳癌掃描技術在乳癌診斷占了很重要的地位。
超音波乳癌掃描技術目前已經從傳統二維影像技術發展為三維體積資料影像技術，而一份三維體積資料影像是由數百張二維影像所建構出來的。發展電腦輔助偵測系統(computer-aided detection, CAD)來自動偵測乳癌於一連串二維乳房超音波影像的技術，將會幫助醫生與放射師的病理診斷，並且對於乳癌掃描技術有莫大的幫助。
雖然目前已有不少電腦輔助偵測乳癌的研究發表，但是因為超音波影像本身既有的超音波假影問題，增加了偵測結果的錯誤率。一個可減少錯誤率的方法是從乳房解剖學上的資訊來做分析。就解剖學而言，大部分腫瘤發生的位置是位在脂肪層與肌肉層中間的乳房組織。
在這篇文獻中，我們將提出一個新的偵測演算法，首先會先偵測超音波影像中的脂肪層與肌肉層，接著再對兩層之間的區域即乳房組織做乳癌偵測的動作。
肌肉層偵測的概念主要是去確認影像中有豐富水平資訊的區域，這些區域可藉由賈柏濾波器（Gabor filters）和相位對稱性（phase symmetry）的計算得到。接著利用自動選定臨界值技術（automatic thresholding selection）和拓樸學關係性（topological relation）去偵測影像中脂肪層的區域。一旦影像中的脂肪層與肌肉層皆被確認出，接著偵測被此兩層所夾的中間區域裡的具高機率為腫瘤的可疑組織。
利用乳房解剖學資訊的分析將可以提高乳癌偵測的準確率。最後本論文藉由60 張超音波影像來進行準確率計算。由結果可以知道，大多數的囊腫（cysts）與腫瘤（lesions）皆可以很成功地被偵測出，證明本論文所發展的演算法具有高的準確率。

Breast cancer is one of the most frequent types of cancer found in females, and it has the highest incident rate of all cancers among females over the age of 40. Therefore, it is important to detect and treat it at an early stage.
There are several breast cancer diagnosis techniques, like mammography, magnetic resonance imaging (MRI), sonograms (ultrasound images), etc. With the benefits of ultrasound, convenience, non-invasiveness and relatively low cost, ultrasound images are considered as useful information on the screening of females at high risk for breast cancer.
Ultrasound breast cancer screening technique plays an important role in the field of the breast diagnosis. Ultrasound breast cancer screening technique has progressed from a 2D single image to a 3D volume data image which is composed of hundreds of 2D images.
Automatic detection of breast lesions in a series of 2D breast sonograms is great help for breast cancer screening; and thus the development of computer-aided detection (CAD) is needed. It provides a convenient way for doctors and radiologists to detect breast cancer while using ultrasound images. A lot of methods for computer-aided detection systems using ultrasound images have been developed by many researchers around the world.
While several approaches have been proposed previously, the false positive rate still tends to be too high for practical use because of the sonographic artifacts. One possible way to reduce the false positive rate is to incorporate the anatomic information into the
decision-making strategy. For anatomic information, most lesions appearing in breast tissue are located in-between the fat layer and the muscle layer in sonograms In this thesis, we present a new detection algorithm for identification of the fat and muscle layers first, and use the in-between region of those two layers to detect lesions.
For muscle layer detection, the basic idea is to identify the area with rich horizontal strip texture patterns by a newly developed texture descriptor computed by the Gabor filters and phase symmetry techniques. For fat layer detection, an automatic thresholding approach in collaboration with a topological relation is proposed. Once the fat and muscle layers are determined, the hypoechoic regions in-between these two layers are more likely to be a breast lesion.
Analysis of anatomic information of breast may increase the accuracy of lesion detection in a single breast ultrasound image. Having examining 60 ultrasound test cases, almost all the cysts and lesions could be detected successfully. The experimental results prove the accuracy of the proposed algorithm.

中文摘要 1
Abstract 3
Contents 5
List of Figures 6
List of Tables 8
1 Introduction 9
1.1 Introduction 10
1.2 Motivation 13
1.3 Organization 16
1.4 Contribution 17
1.5 Data source 18
2 Reference literature 19
3 Method and Material 25
3.1 Gabor Filter 28
3.2 Phase Symmetry 37
3.3 Entropy 43
3.4 Thresholding 48
4 Implementation and Results 53
5 Conclusion and Discussion 85
Ref 88

Sickles EA. Breast imaging: from 1985 to the present. Radiology 2000; 151: 1–16.
[2] Warner E, Plewes DB, Shumak RS, Catzavelos GC, Di Prospero LS, Yaffe MJ,
Ramsay GE, Chart PL, Cole DEC, Taylor GA, Cutrara M, Samuels TH, Murphy JP, Murphy JM, Narod SA. Comparison of breast magnetic resonance imaging, mammography, and ultrasound for surveillance of women at high risk of hereditary breast cancer. Journal of Clinical Oncology 2001; 19:3524–3531.
[3] Buchberger W, DeKoekkoek-Doll P, Springer P, Obrist P, Dunser M. Incidental
findings on sonography of the breast: clinical significance and diagnostic workup. American Journal of Roentgenology 1999; 173: 921–927.
[4] Berg WA, Gilbreath PL. Multicentric and multifocal cancer: whole breast US in
preoperative evaluation. Radiology 2000; 214: 59–66.
[5] Moon WK, Im JG, Koh YH, Noh DY, Park IA. US of mammographically detected
clustered microcalcifications. Radiology 2000; 217: 849–854.
[6] Zonderland HM, Coerkamp EG, Hermans J, van de Vijver MJ, van Voorthuisen
AE. Diagnosis of breast cancer: contributions of US as an adjunct to mammography. Radiology 1999; 213: 413–422.
[7] Bassett LW, Israel M, Gold RH, Ysrael C. Usefulness of mammography and
sonography in women<35 years old. Radiography 1991; 180: 831.
[8] Kolb TM, Lichy J, Newhouse JH. Occult cancer in women with dense breasts:
detection with screening US—diagnostic yield and tumor characteristics. Radiology 1998; 207: 191–199.
[9] Giger ML, Al-Hallaq H, Huo Z, Moran C, Wolverton DE, Chan CW, Zhong W.
Computerized analysis of lesions in US images of the breast. Academic Radiology 1999; 6 (11):665–674.
[10]Chiang HK, Tiu CM, Hung GS, Wu SC, Chang TY, Chou YH. Stepwise logistic
regression analysis of tumor contour features for breast ultrasound diagnosis. Ultrasound in Medicine and Biology 2001; 27 (11):1493–1498.
[11]Anant Madabhushi, Metaxas DN. Combining low-, high-level and empirical
domain knowledge for automated segmentation of ultrasonic breast lesions. IEEE Transactions on Medical Imaging 2004; 22(2):155-169.
[12]Drukker K, Giger ML, Horsch K, Kupinski MA, Vyborny CJ, Mendelson EB.
Computerized lesion detection on breast ultrasound. Medical Physics 2002; 29: 1438–1446.
[13]Drukker K, Giger ML, Mendelson EB. Computerized analysis of shadowing on
breast ultrasound for improved lesion detection. Medical Physics 2003; 30(7): 1833-1842.
[14]Ikedo Y, Fukuoka D, Hara T, Fujita H, Takada E, Endo T, Morita T.
Computer-aided detection system of breast masses on ultrasound images. Proc. of SPIE 2006; 6144:1801-1808.
[15]Chang RF, Chang-Chien KC, Chen HJ, Chen DR, Takada E, Moon WK. Whole
breast computer-aided screening using free-hand ultrasound. IPPR Conference on Computer Vision, Graphics and Image Processing 2005; 1281:1075-1080.
[16]Oppenheim AV, Lim JS. The important of phase in signals. Proc IEEE 1981;
69(5):529-541.
[17]Morrone MC, Owens RA. Feature detection from local energy. Pattern Recognition
Letters 1987; 6:303-313.
[18]Grau V, Noble AJ. Adaptive multi-scale ultrasound compounding using phase
information. Proc MICCAI 2005. Lecture Notes in Computer Science 2005; 3749:589-596.
[19]Kovesi PD. Invariant measures of image features from phase information. PhD
thesis, The University of Western Australia 1996.
[20]Kovesi PD. Symmetry and asymmetry from local phase. Proc Tenth Australian
Joint Conference on Artificial Intelligence 1997: 185-190.
[21]Miguel MP, Noble JA. 2D+T boundary detection in echocardiography. Medical
Image Analysis 2000; 4(1):21-30.
[22]Gerardo SO, Declerck J, Miguel MP, Noble JA. Automatic 3D echocardiographic
image analysis. Proc MICCAI 2000. Lecture Notes in Computer Science 2000; 1935: 687-696.
[23]Daugman JG. Two-dimensional spectral analysis of cortical receptive field profiles.
Vision Research 1980; 20:847– 856.
[24]Young R. The Gaussian derivative theory of spatial vision: analysis of cortical cell
receptive field lineweighting profiles. Tech. Rep. GMR-4920 (General Motor Research), Warren, Mich. 1985.
[25]Malik J, Perona P. Preattentive texture discrimination with early vision Mechanisms
mechanisms . Journal of the Optical Society of America 1990; A7: 923–932.
[26]Jain AK, Farrokhnia F. Unsupervised texture segmentation using Gabor filters.
Pattern Recognition 1991; 24(12):1167–1186.
[27]Wagner RF, Smith SW, Sandrik JM, Lopez H. Statistics of speckle in ultrasound
B-scans. IEEE Transactions of Sonics and Ultrasonics 1983; 30(3): 156-163.
[28]Kovesi PD. Image features from phase congruency. Videre: A Journal of Computer
Vision Research 1999: 1(3), 1-26.
[29]Venkatesh S, Owens R. On the classification of image features. Pattern Recognition
Letters 1990; 11:339-349.
[30]Otsu N. A threshold selection method from gray-level histogram. IEEE
Transactions on Systems, Man, and Cybernetics 1979; 9(1): 62-66.
[31]Liao PS, Chen TS, Chung PC. A fast algorithm for multilevel thresholding.
Journal of Information Science and Engineering 2002; 17, 713-727.
[32]Pollen DA, Ronner SF. Visual cortical neurons as localized spatial frequency filters.
IEEE Transactions on Systems, Man, and Cybernetics 1983; 13:907–916.