在科技充斥且發達的現代社會，人類每天的生活也仰賴著科技進步所帶來的幫助，像是餐廳裡的自助點餐機、工廠裡的機械手臂，還有線上或者是醫院裡面的自動診斷系統，人們藉由科技為生活帶來了許多便利，也幫助我們更能有效率完成許多事務。
而在醫院中的自動診斷系統，要能發揮到真正幫助到人類的效果，就必須要有足 夠準確的判斷結果。本論文針對手骨 X 光影像腕骨分割問題進行探討，透過結合 ResNeSt 和 U-net 神經網路架構的優點，並且也在損失函數上進行利用 Dice Loss 和 Boundary Loss 的相加，使得分割結果能在邊界上有更好的表現，提出可快速、準確 的自動化分割出腕骨部分之系統，並期望若是將來可與手部 X 光影像的判讀做結合， 那麼將可為放射科醫護人員帶來便利，能更有效率、更準確地做出判斷，並且降低因 人為因素、或者影像不夠清晰所造成的誤判。

In a modern society with advanced technology, humans’ daily lives rely on the help brought by technology, such as self-service ordering machines in restaurants, robotic arms in factories, and automatic diagnosis systems online or in hospitals. People use technology to bring a lot of convenience to life, and also help us to complete many tasks more efficiently.
However, if the automatic diagnosis systems in the hospital can really help human beings, it must have sufficiently accurate judgment results. This thesis focuses on hand X-ray images. By combining the advantages of ResNeSt and U-net neural network architecture, and also using the addition of Dice Loss and Boundary Loss in the loss function, the segmentation results can be better on the boundary performance. This thesis proposes a system that can automatically segment the carpal bones quickly and accurately. If the hand X-ray images can be combined with the interpretation system in the future, it will enable radiologists to make more efficient and accurate judgments. Using machines to help humans make more objective judgments can reduce misjudgments caused by human factors or images that are not clear enough.

摘要 II
ABSTRACT III
誌謝 IV
圖目錄 VII
表目錄 IX
第一章 緒論 1
1.1 前言 1
1.2 研究動機/目的 4
1.3 名詞解釋 5
1.4 論文架構 8
第二章 研究方法 9
2.1 神經網路模型 9
2.1.1 U-Net [28] 9
2.1.2 FPN(Feature Pyramid Networks)[29] 10
2.1.3 PSPNet(Pyramid Scene Parsing Network)[31] 11
2.1.4 DeepLabV3+ [32] 12
2.1.5 PAN(Pyramid Attention Network for Semantic Segmentation)[34] 14
2.1.6 ResNeSt(Split-Attention Networks)[35] 16
2.2 損失函數 19
2.2.1 Dice Loss[38] 19
2.2.2 Boundary Loss [40] 20
第三章 資料與預處理 24
3.1 資料庫 24
3.2 資料預處理 25
第四章 實驗設計與結果分析 30
4.1 實驗設計 30
4.2 錯誤評估指標 31
4.3 實驗一:各神經網路架構實驗結果 34
4.4 實驗二:使用 RESNEST 結合 U-NET 架構實驗結果 36
4.5 實驗三:損失函數結合 BOUNDARY LOSS 實驗結果 39
4.6 結果分析 41
第五章 結論與未來展望 45
5.1 結論 46
5.2 未來展望 47
參考文獻 47

[1] Sonka, Milan, Vaclav Hlavac, and Roger Boyle. Image processing, analysis, and machine vision. Cengage Learning, 2014.
[2] Armi, Laleh, and Shervan Fekri-Ershad. "Texture image analysis and texture classification methods-A review." arXiv preprint arXiv:1904.06554 (2019).
[3] Solli, Martin, and Reiner Lenz. "Color semantics for image indexing." Conference on Colour in Graphics, Imaging, and Vision. Vol. 2010. No. 1. Society for Imaging Science and Technology, 2010.
[4] Yu, Honghai, and Stefan Winkler. "Image complexity and spatial information." 2013 Fifth International Workshop on Quality of Multimedia Experience (QoMEX). IEEE, 2013.
[5] Muthukumar, K., S. Poorani, and S. Sindhu. "Color Image segmentation using Similarity based Region merging and Flood Fill Algorithm." vol 5: 40-46 (2016).
[6] Omhover, Jean-François, Marcin Detyniecki, and Bernadette Bouchon-Meunier. "A
region-similarity-based image retrieval system." Proceedings of IPMU. Vol. 4. 2004.
[7] Sheela, S., and M. Sumathi. "Study and Theoretical Analysis of Various Segmentation
Techniques for Ultrasound Images." Procedia Computer Science 87 (2016): 67-73.
[8] Chauhan, Amit Singh, Sanjay Silakari, and Manish Dixit. "Image segmentation methods: A survey approach."2014 Fourth International Conference on Communication Systems and Network Technologies. IEEE, 2014.
[9] Ramella, Giuliana, and Gabriella Sanniti di Baja. "Color histogram-based image segmentation."International Conference on Computer Analysis of Images and
Patterns. Springer, Berlin, Heidelberg, 2011.
[10] Al-Amri, Salem Saleh, and Namdeo V. Kalyankar. "Image segmentation by using
threshold techniques." arXiv preprint arXiv:1005.4020 (2010).
[11] Marr, David, and Ellen Hildreth. "Theory of edge detection." Proceedings of the Royal
Society of London. Series B. Biological Sciences 207.1167 (1980): 187-217.
[12] Huertas, Andres, and Gerard Medioni. "Detection of intensity changes with subpixel accuracy using Laplacian-Gaussian masks." IEEE Transactions on Pattern Analysis
and Machine Intelligence 5 (1986): 651-664.
[13] Van Vliet, Lucas J., Ian T. Young, and Guus L. Beckers. "A nonlinear Laplace operator as edge detector in noisy images." Computer vision, graphics, and image processing 45.2 (1989): 167-195.
[14] Canny, John. "A computational approach to edge detection." IEEE Transactions on pattern analysis and machine intelligence6 (1986): 679-698.
[15] Gupta, Samta, and Susmita Ghosh Mazumdar. "Sobel edge detection algorithm." International journal of computer science and management Research 2.2 (2013): 1578-1583.
[16] Hosang, Jan, Rodrigo Benenson, and Bernt Schiele. "Learning non-maximum suppression." Proceedings of the IEEE conference on computer vision and pattern recognition. 2017.
[17] Adams, Rolf, and Leanne Bischof. "Seeded region growing." IEEE Transactions on pattern analysis and machine intelligence16.6 (1994): 641-647.
[18] Chen, Shiuh-Yung, Wei-Chung Lin, and Chin-Tu Chen. "Split-and-merge image segmentation based on localized feature analysis and statistical tests."CVGIP: Graphical Models and Image Processing 53.5 (1991): 457-475.
[19] Swain, Michael J., and Dana H. Ballard. "Color indexing." International journal of computer vision 7.1 (1991): 11-32.
[20] Deng, Jia, et al. "Imagenet: A large-scale hierarchical image database." 2009 IEEE conference on computer vision and pattern recognition. Ieee, 2009.
[21] Long, Jonathan, Evan Shelhamer, and Trevor Darrell. "Fully convolutional networks for semantic segmentation." Proceedings of the IEEE conference on computer vision and pattern recognition. 2015.
[22] 蔡惠予, “台灣醫療輻射造成國民劑量調查方法研究”, 行政院原子能委員會輻射 偵測中心 (2018): 39-45
[23] Iannaccone, G. "WW Greulich and SI Pyle: Radiographic atlas of skeletal development of the hand and wrist. I volume-atlante di 256 pagine. Stanford University Press, Stanford, California, 1959."Acta geneticae medicae et gemellologiae: twin research8.4 (1959): 513-513.
[24] Li, Chen-Ming, “The study of fully automatic bone age development using the area ratio of carpals”, Master Thesis, EE NTHU. 2011
[25] Lin, Chih-Yang, et al. "Global-and-local context network for semantic segmentation of street view images." Sensors 20.10 (2020): 2907.
[26] Dumoulin, Vincent, and Francesco Visin. "A guide to convolution arithmetic for deep learning." arXiv preprint arXiv:1603.07285(2016).
[27] Wang, Mingzhu, and Jack CP Cheng. "A unified convolutional neural network integrated with conditional random field for pipe defect segmentation." Computer‐ Aided Civil and Infrastructure Engineering 35.2 (2020): 162-177.
[28] Ronneberger, Olaf, Philipp Fischer, and Thomas Brox. "U-net: Convolutional networks for biomedical image segmentation." International Conference on Medical image computing and computer-assisted intervention. Springer, Cham, 2015.
[29] Lin, Tsung-Yi, et al. "Feature pyramid networks for object detection." Proceedings of the IEEE conference on computer vision and pattern recognition. 2017.
[30] Bebis, George, and Michael Georgiopoulos. "Feed-forward neural networks." IEEE Potentials 13.4 (1994): 27-31.
[31] Zhao, Hengshuang, et al. "Pyramid scene parsing network." Proceedings of the IEEE conference on computer vision and pattern recognition. 2017.
[32] Chen, Liang-Chieh, et al. "Encoder-decoder with atrous separable convolution for semantic image segmentation." Proceedings of the European conference on computer vision (ECCV). 2018.
[33] Chen, Liang-Chieh, et al. "Rethinking atrous convolution for semantic image segmentation." arXiv preprint arXiv:1706.05587(2017).
[34] Li, Hanchao, et al. "Pyramid attention network for semantic segmentation." arXiv preprint arXiv:1805.10180 (2018).
[35] Zhang, Hang, et al. "Resnest: Split-attention networks."arXiv preprint arXiv:2004.08955 (2020).
[36] He, Kaiming, et al. "Deep residual learning for image recognition." Proceedings of the IEEE conference on computer vision and pattern recognition. 2016.
[37] Xie, Saining, et al. "Aggregated residual transformations for deep neural networks."Proceedings of the IEEE conference on computer vision and pattern recognition. 2017.
[38] Milletari, Fausto, Nassir Navab, and Seyed-Ahmad Ahmadi. "V-net: Fully convolutional neural networks for volumetric medical image segmentation." 2016 fourth international conference on 3D vision (3DV). IEEE, 2016.
[39] Sorensen, Th A. "A method of establishing groups of equal amplitude in plant sociology based on similarity of species content and its application to analyses of the vegetation on Danish commons." Biol. Skar. 5 (1948): 1-34.
[40] Kervadec, Hoel, et al. "Boundary loss for highly unbalanced segmentation."International conference on medical imaging with deep learning. PMLR, 2019.
[41] Boykov, Yuri, et al. "An integral solution to surface evolution PDEs via geo-cuts." European Conference on Computer Vision. Springer, Berlin, Heidelberg, 2006.
[42] Perez, Luis, and Jason Wang. "The effectiveness of data augmentation in image classification using deep learning." arXiv preprint arXiv:1712.04621 (2017).
[43] Buslaev, Alexander, et al. "Albumentations: fast and flexible image augmentations." Information 11.2 (2020): 125.
[44] Rezatofighi, Hamid, et al. "Generalized intersection over union: A metric and a loss for bounding box regression." Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2019.
[45] Joseph, Mickaël A., and Jansirani Natarajan. "The Carpal and Tarsal Bones of the Human Body: Arabic mnemonics." Sultan Qaboos University Medical Journal 20.2 (2020): e223.