乳癌目前為世界上每年被診斷出最多案例的癌症，而醫療影像對於乳癌篩檢是不可或缺的工具。目前最常用之醫療影像為乳房攝影，但最大缺點為對緻密性乳房的診斷敏感度不佳，可能造成錯誤診斷；反觀超音波影像可以很好地觀察緻密性乳房中的病變。傳統超音波檢測十分仰賴操作人員的經驗，而自動化掃描為一克服此問題的方法，另外掃描時換能器的移動策略中，徑向掃描比起光柵掃描花費較少的時間。因此本研究欲利用徑向環形掃描之乳房超音波影像進行三維重建、自動病灶分割及分析，並提出一個三維乳房病變偵測系統，以減少時間人力成本的消耗。實驗首先在內有模擬病變之乳房假體上擷取一系列的二維乳房超音波影像，重建後獲得三維超音波影像，再從其中利用深度學習分割方法找出病灶實體並分析之，最後提供了病灶位置等資訊。結果顯示以徑向環形掃描的影像亦可以有效地重建出三維影像，另外為減少檢查的時間，將徑向掃描之張數從360張減少至72張(20%)，病變分割效果亦可達到71.10%之準確率。在病灶分割部分對於不同骨幹的U-net網路架構(VGG16、RestNet50、EfficientNet b6)及不同損失函數(交叉熵、Dice loss、Focal loss)進行比較，最高可達骰子係數81.08%的分割準確率。目前大多數之臨床乳癌診斷僅參考二維超音波影像中之二維感興趣之區域；本研究進一步能夠在三維超音波影像中區分三維的病灶實體。最後，本研究提出一個初步的三維乳房病變偵測系統，期許在未來能協助醫生進行診斷，並降低對於超音波操作人員的經驗依賴及提高判斷的一致性。

Breast cancer is the most diagnosed cancer worldwide nowadays. Medical imaging modalities are indispensable for breast cancer screening. The most commonly used modality is mammograms; however, its low sensitivity in dense breasts might cause a misdiagnosis. In contrast, ultrasound imaging can better observe the lesions in dense breast, but it would be affected by the user operations easily. A solution to overcome this problem is automatic scanning. As for the probe moving strategy during scanning, radial scanning cost less time than raster scanning. Thus, this study would use radial scanning breast ultrasound images for three-dimensional reconstruction, automatic lesion segmentation and lesion analysis, and further a system is proposed to reduce the time and labor costs. A set of two-dimensional images is acquired from a breast phantom with simulated lesions and reconstructed into a three-dimensional image firstly. Then it is segmented to find lesion entities by deep learning methods. Lastly, the entities are analyzed to provide lesion information such as location. The results show that radial scanning images can be also reconstructed well. The examination time can be reduced by reducing 80% of radial scanning frames, and the accuracy of lesion segmentation can still reach 71.10%. In addition, this study also compares the segmentation performance of U-net with different network backbones (VGG16、RestNet50、EfficientNet b6) and loss functions (Cross entropy、Dice loss、Focal loss). The best accuracy reaches 81.08% using the dice coefficient. Most conventional breast cancer diagnosis only use two-dimensional images to find region of interest; This study can further distinguish three-dimensional lesion entities in a three-dimensional image. Furthermore, a three-dimensional breast lesion detection system is proposed for assisting doctors to make a diagnostic decision, reducing the operator-dependency, and improving the consistency of diagnosis in the future.

[1] "Breast Anatomy." National Institutes of Health, USA. https://training.seer.cancer.gov/breast/anatomy/ (accessed Feb 28, 2021).
[2] H. Sung, J. Ferlay, R. L. Siegel, M. Laversanne, I. Soerjomataram, A. Jemal, and F. Bray, "Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries," CA: a cancer journal for clinicians, vol. 71, no. 3, pp. 209-249, 2021.
[3] "乳癌防治." 衛生福利部 國民健康署. https://www.hpa.gov.tw/Pages/Detail.aspx?nodeid=614&pid=1124 (accessed Feb 27, 2021).
[4] About Breast Cancer: The American Cancer Society medical and editorial content team, American Cancer Society. [Online]. Available: https://www.cancer.org/content/dam/CRC/PDF/Public/8577.00.pdf. Accessed on: Feb 27, 2021.
[5] "Breast cancer." World Health Organization. https://www.who.int/news-room/fact-sheets/detail/breast-cancer (accessed Feb 26, 2021).
[6] Breast Cancer Risk and Prevention: The American Cancer Society medical and editorial content team, American Cancer Society. [Online]. Available: https://www.cancer.org/content/dam/CRC/PDF/Public/8578.00.pdf. Accessed on: Feb 27, 2021.
[7] Breast Cancer Screening (IARC Handbooks of Cancer Prevention., no. 15). Lyon, France: International Agency for Research on Cancer (IARC) Working Group on the Evaluation of Cancer-Preventive Strategies., 2016.
[8] H. J. Burstein. "Tumor, Node, Metastasis (TNM) staging classification for breast cancer." UpToDate. Waltham, MA UpToDate Inc. https://www.uptodate.com/contents/tumor-node-metastasis-tnm-staging-classification-for-breast-cancer (accessed Feb 10, 2020).
[9] Breast Cancer Early Detection and Diagnosis: The American Cancer Society medical and editorial content team, American Cancer Society. [Online]. Available: https://www.cancer.org/content/dam/CRC/PDF/Public/8579.00.pdf. Accessed on: Feb 27, 2021.
[10] K. K. Shung, Diagnostic ultrasound: Imaging and blood flow measurements. CRC press, 2015.
[11] F. Mohamed and C. V. Siang, "A survey on 3D ultrasound reconstruction techniques," Artificial Intelligence—Applications in Medicine and Biology, 2019.
[12] D. E. Octorina Dewi, M. Mohd. Fadzil, A. A. Mohd. Faudzi, E. Supriyanto, and K. W. Lai, "Position Tracking Systems for Ultrasound Imaging: A Survey," in Medical Imaging Technology: Reviews and Computational Applications, K. W. Lai and D. E. Octorina Dewi Eds. Singapore: Springer Singapore, 2015, pp. 57-89.
[13] H. Gao, Q. Huang, X. Xu, and X. Li, "Wireless and sensorless 3D ultrasound imaging," Neurocomputing, vol. 195, pp. 159-171, 2016.
[14] J. F. Chen, J. B. Fowlkes, P. L. Carson, and J. M. Rubin, "Determination of scan‐plane motion using speckle decorrelation: Theoretical considerations and initial test," International Journal of Imaging Systems and Technology, vol. 8, no. 1, pp. 38-44, 1997.
[15] S. S. Kaplan, "Automated Whole Breast Ultrasound," (in English), Radiol. Clin. N. Am., Article vol. 52, no. 3, p. 539, May 2014.
[16] J. C. M. van Zelst and R. M. Mann, "Automated Three-dimensional Breast US for Screening: Technique, Artifacts, and Lesion Characterization," (in English), Radiographics, Article vol. 38, no. 3, pp. 663-683, May-Jun 2018.
[17] R. Rosensweig, P. M. Foy, C. Colebeuglet, A. B. Kurtz, and B. B. Goldberg, "RADIAL SCANNING OF THE BREAST - AN ALTERNATIVE TO THE STANDARD ULTRASOUND TECHNIQUE," (in English), J. Clin. Ultrasound, Note vol. 10, no. 4, pp. 199-201, 1982.
[18] R. Selvi, "Equipment Settings and Examination Technique," in Breast Diseases: Imaging and Clinical Management, R. Selvi Ed. New Delhi: Springer India, 2015, pp. 99-110.
[19] C. Jäggi-Wickes, P. Brasier-Lutz, S. Schaedelin, R. Burian, C.-A. Schoenenberger, and R. Zanetti-Dällenbach, "Comparison of radial and meander-like breast ultrasound with respect to diagnostic accuracy and examination time," (in eng), Arch Gynecol Obstet, vol. 301, no. 6, pp. 1533-1541, 2020.
[20] M. H. Yap, G. Pons, J. Marti, S. Ganau, M. Sentis, R. Zwiggelaar, A. K. Davison, and R. Marti, "Automated Breast Ultrasound Lesions Detection Using Convolutional Neural Networks," (in English), IEEE J. Biomed. Health Inform., Article vol. 22, no. 4, pp. 1218-1226, Jul 2018.
[21] K. Drukker, M. L. Giger, K. Horsch, M. A. Kupinski, C. J. Vyborny, and E. B. Mendelson, "Computerized lesion detection on breast ultrasound," (in English), Med. Phys., Article vol. 29, no. 7, pp. 1438-1446, Jul 2002.
[22] M. H. Yap, E. A. Edirisinghe, and H. E. Bez, "A novel algorithm for initial lesion detection in ultrasound breast images," (in English), J. Appl. Clin. Med. Phys, Article vol. 9, no. 4, pp. 181-199, 2008.
[23] K. Doi, "Computer-aided diagnosis in medical imaging: Historical review, current status and future potential," Computerized Medical Imaging and Graphics, vol. 31, no. 4, pp. 198-211, 2007.
[24] J. Shiraishi, Q. Li, D. Appelbaum, and K. Doi, "Computer-aided diagnosis and artificial intelligence in clinical imaging," (in eng), Semin Nucl Med, vol. 41, no. 6, pp. 449-62, Nov 2011.
[25] J. Z. Cheng, D. Ni, Y. H. Chou, J. Qin, C. M. Tiu, Y. C. Chang, C. S. Huang, D. Shen, and C. M. Chen, "Computer-Aided Diagnosis with Deep Learning Architecture: Applications to Breast Lesions in US Images and Pulmonary Nodules in CT Scans," (in English), Sci Rep, Article vol. 6, p. 13, Apr 2016, Art no. 24454.
[26] W. Al-Dhabyani, M. Gomaa, H. Khaled, and A. Fahmy, "Dataset of breast ultrasound images," (in English), Data Brief, Article; Data Paper vol. 28, p. 104863, Feb 2020, Art no. 104863.
[27] P. Yakubovskiy, "Segmentation Models," GitHub repository, 2019. [Online]. Available: https://github.com/qubvel/segmentation_models.
[28] A. Buslaev, V. I. Iglovikov, E. Khvedchenya, A. Parinov, M. Druzhinin, and A. A. Kalinin, "Albumentations: fast and flexible image augmentations," Information, vol. 11, no. 2, p. 125, 2020.
[29] D. E. Robinson and P. C. Knight, "INTERPOLATION SCAN CONVERSION IN PULSE-ECHO ULTRASOUND," (in English), Ultrason. Imaging, Article vol. 4, no. 4, pp. 297-310, 1982.
[30] A. Fenster, D. B. Downey, and H. N. Cardinal, "Three-dimensional ultrasound imaging," (in English), Phys. Med. Biol., Review vol. 46, no. 5, pp. R67-R99, May 2001.
[31] Z. Wang and A. C. Bovik, "Mean squared error: Love it or leave it? A new look at signal fidelity measures," IEEE signal processing magazine, vol. 26, no. 1, pp. 98-117, 2009.
[32] Z. Wang, A. C. Bovik, H. R. Sheikh, and E. P. Simoncelli, "Image quality assessment: From error visibility to structural similarity," (in English), IEEE Trans. Image Process., Article vol. 13, no. 4, pp. 600-612, Apr 2004.
[33] R. Szeliski, Computer vision: algorithms and applications. Springer Science & Business Media, 2010.
[34] N. Otsu, "A threshold selection method from gray-level histograms," IEEE transactions on systems, man, and cybernetics, vol. 9, no. 1, pp. 62-66, 1979.
[35] M. Kass, A. Witkin, and D. Terzopoulos, "Snakes: Active contour models," International journal of computer vision, vol. 1, no. 4, pp. 321-331, 1988.
[36] N. Plath, M. Toussaint, and S. Nakajima, "Multi-class image segmentation using conditional random fields and global classification," in Proceedings of the 26th Annual International Conference on Machine Learning, 2009, pp. 817-824.
[37] S. Minaee, Y. Y. Boykov, F. Porikli, A. J. Plaza, N. Kehtarnavaz, and D. Terzopoulos, "Image segmentation using deep learning: A survey," IEEE Trans. Pattern Anal. Mach. Intell., 2021.
[38] J. Long, E. Shelhamer, and T. Darrell, "Fully convolutional networks for semantic segmentation," in Proceedings of the IEEE conference on computer vision and pattern recognition, 2015, pp. 3431-3440.
[39] V. Badrinarayanan, A. Kendall, and R. Cipolla, "Segnet: A deep convolutional encoder-decoder architecture for image segmentation," IEEE Trans. Pattern Anal. Mach. Intell., vol. 39, no. 12, pp. 2481-2495, 2017.
[40] O. Ronneberger, P. Fischer, and T. Brox, "U-net: Convolutional networks for biomedical image segmentation," in International Conference on Medical image computing and computer-assisted intervention, 2015: Springer, pp. 234-241.
[41] K. He, G. Gkioxari, P. Dollár, and R. Girshick, "Mask r-cnn," in Proceedings of the IEEE international conference on computer vision, 2017, pp. 2961-2969.
[42] A. Benali Amjoud and M. Amrouch, "Convolutional Neural Networks Backbones for Object Detection," Cham, 2020: Springer International Publishing, in Image and Signal Processing, pp. 282-289.
[43] K. Simonyan and A. Zisserman, "Very deep convolutional networks for large-scale image recognition," arXiv preprint, 2014.
[44] K. He, X. Zhang, S. Ren, and J. Sun, "Deep residual learning for image recognition," in Proceedings of the IEEE conference on computer vision and pattern recognition, 2016, pp. 770-778.
[45] M. Tan and Q. Le, "Efficientnet: Rethinking model scaling for convolutional neural networks," in International Conference on Machine Learning, 2019: PMLR, pp. 6105-6114.
[46] W. Al-Dhabyani, A. Fahmy, M. Gomaa, and H. Khaled, "Deep Learning Approaches for Data Augmentation and Classification of Breast Masses using Ultrasound Images," (in English), Int. J. Adv. Comput. Sci. Appl., Article vol. 10, no. 5, pp. 618-627, May 2019.
[47] S. Jadon, "A survey of loss functions for semantic segmentation," in 2020 IEEE Conference on Computational Intelligence in Bioinformatics and Computational Biology (CIBCB), 2020: IEEE, pp. 1-7.
[48] C. H. Sudre, W. Li, T. Vercauteren, S. Ourselin, and M. J. Cardoso, "Generalised dice overlap as a deep learning loss function for highly unbalanced segmentations," in Deep learning in medical image analysis and multimodal learning for clinical decision support: Springer, 2017, pp. 240-248.
[49] T. Y. Lin, P. Goyal, R. Girshick, K. M. He, and P. Dollar, "Focal Loss for Dense Object Detection," (in English), IEEE Trans. Pattern Anal. Mach. Intell., Article vol. 42, no. 2, pp. 318-327, Feb 2020.
[50] A. Kaehler and G. Bradski, Learning OpenCV 3: computer vision in C++ with the OpenCV library. " O'Reilly Media, Inc.", 2016.
[51] M. B. Dillencourt, H. Samet, and M. Tamminen, "A GENERAL-APPROACH TO CONNECTED-COMPONENT LABELING FOR ARBITRARY IMAGE REPRESENTATIONS," (in English), J. Acm, Article vol. 39, no. 2, pp. 253-280, Apr 1992.
[52] D. B. Downey, A. Fenster, and J. C. Williams, "Clinical utility of three-dimensional US," Radiographics, vol. 20, no. 2, pp. 559-571, 2000.
[53] M. G. del Carmen, E. F. Halpern, D. B. Kopans, B. Moy, R. H. Moore, P. E. Goss, and K. S. Hughes, "Mammographic Breast Density and Race," American Journal of Roentgenology, vol. 188, no. 4, pp. 1147-1150, 2007.
[54] Y. He, H. Yu, X. Liu, Z. Yang, W. Sun, Y. Wang, Q. Fu, Y. Zou, and A. Mian, "Deep Learning based 3D Segmentation: A Survey," arXiv preprint, 2021.