在癌症的分期上，縱膈腔淋巴結之分布及型態的評估非常重要，而胸腔電腦斷層掃描則是臨床上檢測的主要工具。在現今的檢測流程中，為了從患者的電腦斷層掃描中找出可能異常的腫大淋巴結，放射科醫生需要檢查每張斷層掃描影像。為了輔助醫生檢測並減少所耗的時間和錯誤標記，在本篇研究中，我們提出一個在胸腔電腦斷層攝影上全自動偵測腫大的縱膈腔淋巴結之系統。首先，我們利用三維分水嶺切割將電腦斷層掃描分成許多的三維區域。接著，利用兩個斑點偵測器在二維切片上偵測可疑的斑點狀區域，再經由淋巴結在電腦斷層掃描影像上的大小、密度以及藉由切割出淋巴結於解剖學上周遭的器官，例如氣管、主動脈弓來過濾不可能為淋巴結的可疑區域，並且將可疑的斑點狀區域以二維可疑點表示。最後，將三維區域與二維的可疑點對應並過濾後，獲得了一組可能為淋巴結的三維區域，再對這些區域取出幾何形狀、亮度的統計資訊以及亮度之直方圖等特徵資訊來訓練分類器，移除偽陽性之偵測。我們使用了25個對比增強的電腦斷層攝影影像來評估提出的系統，其中有77個被標記的腫大淋巴結。根據實驗的結果，我們提出的系統達到100%、90.90%、80.51%、70.12%以及61.00%的敏感性，其中分別對應每組影像平均有15.12、11.2、8.32、6.16以及3.76個偽陽性之偵測。在臨床上，我們提出的系統可以達到100%的敏感性，這帶給醫生更可靠的建議。

The assessment of involvement of mediastinal lymph nodes is an important evaluation criterion in cancer staging with computed tomography (CT) as the imaging tool. In the clinical workflow, radiologists need to examine all CT slices to detect abnormal enlarge lymph nodes. In this study, a fully automatic computer-aided detection system was proposed to detect mediastinal lymph nodes. With the assistance of the proposed system, the examination time and missed lymph nodes can be reduced. First, a mediastinal volume of interest (VOI) was extracted to restrict the possible locations of lymph nodes. Second, the 3-D watershed transform was performed to obtained morphology information. Third, two blob detectors were combined to identify suspicious regions. Finally, quantitative features were extracted from the overlapped regions in the 2-D suspicious point and 3-D suspicious region and combined in a classifier to discriminate the real lymph nodes from other tissues. The proposed system achieved the sensitivities of 100%, 90.90%, 80.51%, 70.12%, and 61.00% with 15.12, 11.2, 8.32, 6.16 and 3.76 false positives per volume, respectively. In the clinical use, the sensitivities of 100% achieved by our proposed system can provide more reliable recommendations to radiologists.

口試委員審定書 i
致謝 ii
摘要 iii
Abstract iv
Table of Contents v
List of Figures vi
List of Tables x
Chapter 1 Introduction 1
Chapter 2 Materials 4
Chapter 3 Computer-aided lymph node detection 5
3.1 Mediastina VOI locating 7
3.2 3-D Suspicious Region Extraction 11
3.3 Blob-like Map Construction 14
3.4 2-D Suspicious Point Extraction 17
3.4.1 Intensity 17
3.4.2 Location 18
3.4.2.1 Airway Tree Segmentation 18
3.4.2.2 Aortic Arch Segmentation 20
3.4.3 Bounding box 27
3.5 Lymph Node Candidate Determination 29
3.6 Classification on Lymph Node Candidates 31
3.6.1 Morphology 31
3.6.2 Statistic 33
3.6.3 Histogram 33
Chapter 4 Experimental Results and Discussion 34
4.1 Evaluation Methodology 34
4.2 Results 35
4.3 Discussion 45
Chapter 5 Conclusion and Future Work 46
References 47

[1]R. Siegel, J. Ma, Z. Zou, and A. Jemal, "Cancer statistics, 2014," CA: a cancer journal for clinicians, vol. 64, pp. 9-29, 2014.
[2]S. Kligerman and G. Abbott, "A radiologic review of the new TNM classification for lung cancer," American Journal of Roentgenology, vol. 194, pp. 562-573, 2010.
[3]T. Suwatanapongched and D. Gierada, "CT of thoracic lymph nodes. Part I: anatomy and drainage," British journal of radiology, vol. 79, pp. 922-928, 2006.
[4]J. Feulner, S. K. Zhou, M. Hammon, J. Hornegger, and D. Comaniciu, "Lymph node detection and segmentation in chest CT data using discriminative learning and a spatial prior," Med Image Anal, vol. 17, pp. 254-70, Feb 2013.
[5]E. Eisenhauer, P. Therasse, J. Bogaerts, L. Schwartz, D. Sargent, R. Ford, J. Dancey, S. Arbuck, S. Gwyther, and M. Mooney, "New response evaluation criteria in solid tumours: revised RECIST guideline (version 1.1)," European journal of cancer, vol. 45, pp. 228-247, 2009.
[6]L. Schwartz, J. Bogaerts, R. Ford, L. Shankar, P. Therasse, S. Gwyther, and E. Eisenhauer, "Evaluation of lymph nodes with RECIST 1.1," European journal of cancer, vol. 45, pp. 261-267, 2009.
[7]L. Dornheim and J. Dornheim, "Automatische detektion von lymphknoten in ct-datensatzen des halses," in Bildverarbeitung fur die Medizin 2008, ed: Springer, 2008, pp. 308-312.
[8]M. Feuerstein, D. Deguchi, T. Kitasaka, S. Iwano, K. Imaizumi, Y. Hasegawa, Y. Suenaga, and K. Mori, "Automatic mediastinal lymph node detection in chest CT," in SPIE medical imaging, 2009, pp. 72600V-72600V-11.
[9]M. Feuerstein, B. Glocker, T. Kitasaka, Y. Nakamura, S. Iwano, and K. Mori, "Mediastinal atlas creation from 3-D chest computed tomography images: application to automated detection and station mapping of lymph nodes," Med Image Anal, vol. 16, pp. 63-74, Jan 2012.
[10]A. Barbu, M. Suehling, X. Xu, D. Liu, S. K. Zhou, and D. Comaniciu, "Automatic detection and segmentation of lymph nodes from CT data," IEEE Trans Med Imaging, vol. 31, pp. 240-50, Feb 2012.
[11]T. Kitasaka, Y. Tsujimura, Y. Nakamura, K. Mori, Y. Suenaga, M. Ito, and S. Nawano, "Automated extraction of lymph nodes from 3-D abdominal CT images using 3-D minimum directional difference filter," in Medical Image Computing and Computer-Assisted Intervention–MICCAI 2007, ed: Springer, 2007, pp. 336-343.
[12]S. Beucher, "Watershed, hierarchical segmentation and waterfall algorithm," in Mathematical morphology and its applications to image processing, ed: Springer, 1994, pp. 69-76.
[13]D. G. Lowe, "Distinctive image features from scale-invariant keypoints," International journal of computer vision, vol. 60, pp. 91-110, 2004.
[14]J. Matas, O. Chum, M. Urban, and T. Pajdla, "Robust wide-baseline stereo from maximally stable extremal regions," Image and vision computing, vol. 22, pp. 761-767, 2004.
[15]A. Rosenfeld and J. L. Pfaltz, "Sequential operations in digital picture processing," Journal of the ACM (JACM), vol. 13, pp. 471-494, 1966.
[16]S. Tan, E. Teo, and H. Chua, "Quantitative three-dimensional anatomy of cervical, thoracic and lumbar vertebrae of Chinese Singaporeans," European Spine Journal, vol. 13, pp. 137-146, 2004.
[17]K. Mori, J.-i. Hasegawa, J. Toriwaki, H. Anno, and K. Katada, "Recognition of bronchus in three-dimensional X-ray CT images with applications to virtualized bronchoscopy system," in Pattern Recognition, 1996., Proceedings of the 13th International Conference on, 1996, pp. 528-532.
[18]L. Vincent and P. Soille, "Watersheds in digital spaces: an efficient algorithm based on immersion simulations," IEEE transactions on pattern analysis and machine intelligence, vol. 13, pp. 583-598, 1991.
[19]N. Salman, "Image Segmentation Based on Watershed and Edge Detection Techniques," Int. Arab J. Inf. Technol., vol. 3, pp. 104-110, 2006.
[20]P. Perona and J. Malik, "Scale-space and edge detection using anisotropic diffusion," Pattern Analysis and Machine Intelligence, IEEE Transactions on, vol. 12, pp. 629-639, 1990.
[21]H. Kong, H. C. Akakin, and S. E. Sarma, "A Generalized Laplacian of Gaussian Filter for Blob Detection and Its Applications," IEEE T. Cybernetics, vol. 43, pp. 1719-1733, 2013.
[22]K. Mikolajczyk, T. Tuytelaars, C. Schmid, A. Zisserman, J. Matas, F. Schaffalitzky, T. Kadir, and L. Van Gool, "A comparison of affine region detectors," International journal of computer vision, vol. 65, pp. 43-72, 2005.
[23]R. C. Gonzalez, R. E. Woods, and S. L. Eddins, Digital image processing using MATLAB vol. 2: Gatesmark Publishing Knoxville, 2009.
[24]D. Comaniciu and P. Meer, "Mean shift: A robust approach toward feature space analysis," Pattern Analysis and Machine Intelligence, IEEE Transactions on, vol. 24, pp. 603-619, 2002.
[25]N. Krishnamurthy, Introduction to computer graphics: Tata McGraw-Hill, 2002.
[26]K. F. Mulchrone and K. R. Choudhury, "Fitting an ellipse to an arbitrary shape: implications for strain analysis," Journal of structural geology, vol. 26, pp. 143-153, 2004.
[27]M. Hu, "Visual-Pattern Recognition by Moment Invariants," Ire Transactions on Information Theory, vol. 8, pp. 179-&;, 1962.
[28]D. W. Hosmer Jr, S. Lemeshow, and R. X. Sturdivant, Applied logistic regression: Wiley. com, 2013.
[29]F. A. Sadjadi and E. L. Hall, "Three-dimensional moment invariants," Pattern Analysis and Machine Intelligence, IEEE Transactions on, pp. 127-136, 1980.
[30]W. Cheung and G. Hamarneh, "N-sift: N-dimensional scale invariant feature transform for matching medical images," in Biomedical Imaging: From Nano to Macro, 2007. ISBI 2007. 4th IEEE International Symposium on, 2007, pp. 720-723.
[31]T. Kadir and M. Brady, "Saliency, scale and image description," International Journal of Computer Vision, vol. 45, pp. 83-105, 2001.
[32]K. Mikolajczyk and C. Schmid, "An affine invariant interest point detector," in Computer Vision—ECCV 2002, ed: Springer, 2002, pp. 128-142.
[33]K. Mikolajczyk and C. Schmid, "Scale &; affine invariant interest point detectors," International journal of computer vision, vol. 60, pp. 63-86, 2004.
[34]H. Bay, A. Ess, T. Tuytelaars, and L. Van Gool, "Speeded-up robust features (SURF)," Computer vision and image understanding, vol. 110, pp. 346-359, 2008.
[35]J.-M. Morel and G. Yu, "ASIFT: A new framework for fully affine invariant image comparison," SIAM Journal on Imaging Sciences, vol. 2, pp. 438-469, 2009.
[36]R. Vazquez Martin, "Affine image region detection and description," Journal of Physical Agents, vol. 4, pp. 45-54, 2010.
[37]C. Duanggate, B. Uyyanonvara, S. S. Makhanov, S. Barman, and T. Williamson, "Object detection with feature stability over scale space," Journal of Visual Communication and Image Representation, vol. 22, pp. 345-352, 2011.