臺灣博碩士論文加值系統

超音波影像中的甲狀腺結節具有不同的成份及模糊的邊界，導致放射線專科醫師難以描繪出完整的結節形狀及判別出結節的成份；因此，本文提出一個自動切割結節並且分類結節成份的方法來改善診斷上的困難及降低誤診率。
我們使用一個決策樹演算法切割出可疑的結節區域，之後修復結節的樣貌並除去不必要的區域，使切割出的結節更為精緻，最後利用階層式的支援向量機來分類結節成分，此支援向量機由三個分兩類的支援向量機構成，分別含有經特徵挑選工具挑選後的最佳特徵來辨識不同的成分。
實驗結果比較不同的切割方法，同時評估結節中成份分類的結果，兩者皆顯示本論文提出的方法具有良好的效率。

Heterogeneous thyroid nodules which have distinct histopathlogical components and vague boundaries in ultrasound images result in a difficult task when radiologists and physicians manually draw a complete shape of nodule, or distinguish what kind of component a nodule has. Hence, an automatic segmentation and classification method is required to improve this tough work and decrease the misdiagnosis rate in this thesis.
We use decision tree algorithm to segment the possible nodular area. And the segmentation is refined to recover the nodular shape and remove extra region. Finally, a classification method based on a hierarchical support vector machine (SVM) combined with five binary-SVM is applied to find the distribution of components in the nodular lesion. Each SVM classifier recognizes the different components according to the corresponding optimal features which are extracted by a feature selection tool. Experimental results compared other segmentation algorithms and evaluated the classification of the components. Both showed good performances of the proposed approach.

摘要 ---------------------------------------------------------------------i
ABSTRACT --------------------------------------------------------------------ii
誌謝 -------------------------------------------------------------------iii
Contents --------------------------------------------------------------------iv
List of Tables -----------------------------------------------------------vi
List of Figures ----------------------------------------------------------vii
Chapter 1 Introduction -------------------------------------------------------1
1.1 Motivation -------------------------------------------------1

1.2 Objective of The Thesis ------------------------------------2

1.3 System Architecture-----------------------------------------3

1.4 Organization of The Thesis ---------------------------------3

Chapter 2 Related Image Processing and Preprocessing--------------------------5

2.1 Image Enhancement in Spatial Domain-------------------------5

2.1.1 Median Filter--------------------------------------5
2.1.2 Histogram Equalization-----------------------------6
2.1.3 Locating Suspicious Nodular Area ------------------7
2.2 Feature Extraction---------------------------------------------------8
2.2.1 Gray Level Co-occurrence Matrix--------------------------------------9
2.2.2 Statistic Feature Matrix--------------------------------------------13
2.2.3 Gray Length Run Length Matrix---------------------------------------13
2.2.4 Law’s Texture Energy Measures--------------------------------------15
2.2.5 Neighboring Gray Level Dependence Matrix----------------------------16
2.2.6 Homogeneity---------------------------------------------------------18
2.2.7 Histogram-----------------------------------------------------------19
2.2.8 Block Difference of Inverse Probability-----------------------------19
2.2.9 Discrete Cosine Transform-------------------------------------------20
2.2.10 Normalized Multi-scale Intensify Different--------------------------21
2.2.11 Haar Wavelet--------------------------------------------------------21
Chapter 3 The Proposed Method------------------------------------------------23
3.1 The Segmentation Algorithm -----------------------------------------23
3.1.1 Decision Tree-------------------------------------------------------24
3.1.2 Data Discretization ------------------------------------------------26
3.1.3 Entropy-------------------------------------------------------------27
3.1.4 Gain Ratio----------------------------------------------------------28
3.1.5 Feature Selection for Splitting Branches----------------------------29
3.1.6 Pruning-------------------------------------------------------------32
3.2 Nodular Shape Refinement -------------------------------------------35
3.2.1 Erosion and Dilation------------------------------------------------38
3.2.2 Labeling------------------------------------------------------------38
3.2.3 Region filling -----------------------------------------------------41
3.2.4 Region Growing Based on Convex Hull --------------------------------43
3.2.5 Boundary Extraction-------------------------------------------------45
3.3 Classifying the Histopathological Components in Nodules-------------45
3.3.1 Introduction about Support Vector Machine---------------------------46
3.3.2 Kernel Functions----------------------------------------------------49
3.3.3 Feature Selection---------------------------------------------------50
3.3.4 Hierarchical SVM----------------------------------------------------51
Chapter 4 Experimental Results and Discussion--------------------------------53
4.1 Experimental Environment and Data Source-------------------54
4.2 Selection of the Optimal Features for Internal nodes-------56
4.3 Comparison of Other Methods for Segmentation---------------59
4.4 Performance of Segmentation--------------------------------74
4.5 Performance of Classification------------------------------78
Chapter 5 Conclusions -------------------------------------------------------87
References-------------------------------------------------------------------88

[1]S. Tsantis, N. Dimitropoulos, D. Cavouras and G. Nikiforidis, 2006, “A hybrid Muti-scale model For Thyroid Nodule Boundary Detection Ultrasound Images”, Computer Method And Program In Biomedicine, vol. 84, pp. 86-98.
[2]S. J. Chen, S. N. Yu, J. E. Tzeng, Y. T. Chen, K. Y. Chang, K. S. Cheng, F. T. Hsiao and C. K. Wei, 2008, “Characterization Of The Major Histopathological Components Of Thyroid Nodules Using Sonographic Texture Features For Clinical Diagnosis And Management”, Ultrasound in Med. & Biol., vol. 35, pp. 201-208.
[3]S. Tsantis, D. Cavouras, I. Kalatzis, N. Piliouras, N. Dimitropoulos, and G. Nikiforidis, 2005, “Development of A Support Vector Machine-based Image Analysis System For Assessing The Thyroid Nodule Malignancy Risk on Ultrasound”, Ultrasound in Med. & Biol., vol. 31, pp. 1451-1459.
[4]C. K. Yeh, Y. S. Chen, W. C. Fan and Y. Y. Liao, 2009, “A disk expansion segmentation method for ultrasonic breast lesions”, Pattern recognition, vol. 42, pp. 596-606.
[5]M. Kass, A. Witkin and D. Terzopoulos, 1987, “Snake: active contour model”, International Journal of Computer Vision, pp.321 – 331.
[6]S. Tsantis, N. Dimitropoulos, M. Ioannidou, D. Cavouras, G. Nikiforidis, “Inter-scale wavelet analysis for speckle reduction in thyroid ultrasound images”, Computerized Medical Imaging and Graphics, vol. 31, pp. 117-127
[7]R. M. Haralick, K. Shanugam and I. Dinstein, 1973, “Textural Features for Image Classification”, IEEE Trans. Sys., Man and Cyb., vol. 3, pp. 610-621.
[8]C. M. Wu and Y. C. Chen, 1992, “Statistical feature matrix for texture analysis”, CVGIP: Graphical Models and Image Processing, vol. 54, pp. 407-419.
[9]M. M. Galloway, 1975, “Texture Analysis Using Gray Run Lengths”, Computer Graphics and Image Processing, vol. 4, pp. 172-179.
[10]C. J. Sun and W. G. Wee, 1983, “Neighboring gray level dependence matrix for texture classification”, Computer Vision, Graphics, and Image Processing, vol. 23, pp. 341-352.
[11]L. H. Siew, R. M. Hodgson and E. J. Wood, 1988, “Texture measures for carpet wear assessment”, IEEE Trans. Patt. Anal. Machine Intell., vol. 10, No. 1, pp.92 – 105.
[12]Y. D. Chum, S. Y. Seo, 2003, “Image Retrieval using BDIP and BVLC Moments”, IEEE Trans. Cir. & Sys. for Video Tech., vol. 13, pp. 951-957.
[13]E. L. Chen, P.C. Chung, C. L. Chen, H. M. Tsai, C. I. Chang, 1998, “An automatic diagnostic system for CT liver image classification”, IEEE Trans. Biol. Eng., vol. 45, pp. 783-794.
[14]M. Antonini, M. Barlaud, P. Mathieu and I. Daubechies, 1992, “Image coding using wavelet transform”, IEEE Trans. Imag. Proc., vol. 1, pp.205 – 220.
[15]W. H. Chao, Y. Y. Chen, C. W. Cho, S. H. Lin, Y. Y. I. Shih and S. Tsang, 2008, “Improving segmentation accuracy for magnetic resonance imaging using a boosted decision tree”, Journal of Neuroscience Methods, vol. 175, pp. 206-217,.
[16]S. Dua, H. Singh and H.W. Thompson, 2009, “Associative classification of mammograms using weighted rules”, Expert system with application, vol.36, pp.9250 – 9259.
[17]B. Samanta, G. Bird, M. Kuijpers, R. Zimmerman, G. Jarvik, G. Wernovsky, R. Clancy, D. Licht, J. Gaynor, C. Nataraj, 2009, “Prediction of periventricular leukomalacia. Part I: Selection of hemodynamic features using logistic regression and decision tree algorithms”, Intell. Intel. in Med., vol. 46, pp. 201-215.
[18]L. G. Shapiro and G. C. Stockman, 2001, Computer Vision, Prentice-Hall, NJ.
[19]R. C. Gonzalez and R. E. Woods, 2002, Digital Image Processing, 2nd, Prentice-Hall International Edit.
[20]S. Haykin, 1998, Neural Networks, Prentice-Hall, NJ.
[21]S. Z. Li, J. Yan and H. J. Zhang, 2001, “Learning Illumination-Invariant View Subspaces of Object Appearances “, Technical Report.
[22]H. Wechsler, P. Phillips, V. Bruce, F. Soulie and T. Huang, 1998, “Face Recognition: From Theory to Applications”, Springer-Verlag Telos.
[23]J.C. Burges, 1998, “A Tutorial on Support Vector Machines for Pattern Recognition”, Data Mining and Knowledge Discovery, vol. 2, pp. 121-167.
[24]A. J. Smola and B. Scholkopf, 2004, “A tutorial on support vector regression”, Statistics and Computing, vol. 14, pp. 199-222.
[25]C. Y. Chang, M. F. Tsai and S. J. Chen, 2008, “Classification of the Thyroid Nodules Using Support Vector Machines,” IEEE Inter. Join. Conf. on Neural Network, pp.3093-3098.
[26]C.C. Chang and C.J. Lin, LIBSVM: a library for support vector machines, 2001. Software available at http://www.csie.ntu.edu.tw/~cjlin/libsvm
[27]J. Dehmeshki, H. Amin, M. Valdivieso, and X. Ye, 2008, “Segmentation of Pulmonary Nodules in Thoracic CT Scans: A Region Grwoing Approach,” IEEE Trans. Med. Imag., vol. 27, pp.467-480.
[28]S. Shen, W. Sandham, M. Granat, and A. Sterr, 2005, “MRI Fuzzy Segmentation of Brain Tissue Using Neighborhood Attraction With Neural-Network Optimization”, IEEE Trans. BIOM., vol. 9, pp.459-467.
[29]A. P. DHAWAN, 2003, Medical Image Analysis, John Wiley & Sons, Inc.
[30]R. D. Boer, H. A. Vrooman, F. V. Lijn, M. W. Vernooij, M. A. Ikram, A.V. Lugt, M.B. Breteler and W. J. Niessen, 2009, “White matter lesion extension to automatic brain tissue segmentation on MRI”, NeuroImage, vol. 45, pp.1151-1161.