臺灣博碩士論文加值系統

　　癌症為國人死因之首，核子醫學骨骼造影是偵測癌症骨轉移的利器。醫師要從大量影像中找出異常病變區域並且判斷其確切位置，是一項十分耗費精力與時間的工作。因此，本論文希望能夠以深度學習為基礎，開發出一套核子醫學骨骼造影自動定位系統，以便減輕醫師的工作量。

　　本篇採用YOLOv3和SSD300架構為基底，修改其內部的卷積神經網路，透過不同參數進行骨骼造影之腰椎定位以提高自動定位的準確度。接著我們用影像前處理標記樣本腰椎相對位置，來訓練卷積神經網路處理影像資料，使其快速有效的處理大量的影像。

　　The malignancies are the number one cause of death among Taiwanese population, and bone scintigraphy is a fine tool to detect the bone metastases of the malignancies. Finding out the abnormal lesions and locating the precise location are energy-consuming and time-consuming task for the physicians. Therefore, this thesis is looking forward to use deep learning as a basis to develop an auto positioning system of bone scintigraphy in order to mitigate the workload of the physicians.

　　Based on the YOLOv3 and SSD300 architecture, this thesis modifies its internal Convolution Neural Network for the positioning of lumbar spine in bone scintigraphy by different parameters to increase the accuracy of auto positioning. Then we mark the relative position of sample lumbar spine by image preprocessing to train the Convolution Neural Network to process the image data, and make it process a large amount of images rapidly and effectively.

摘 要 i
Abstract ii
誌 謝 iv
目 錄 v
圖 目 錄 vii
表 目 錄 viii
第1章 緒論 1
1.1. 研究動機與背景 1
1.2. 研究目的 2
1.3. 研究範圍與限制 2
1.4. 研究架構 3
第2章 文獻回顧 4
2.1. 醫學影像 4
2.1.1. 核子醫學影像 5
2.1.2. 骨骼造影 7
2.2. 機器學習 9
2.2.1. 學習分類模式 9
2.3. 神經網路模型CNN 10
2.4. 深度學習 11
2.4.1. 深度學習框架 12
2.4.2. Keras 12
2.4.3. TensorFlow 13
2.5. 物件偵測 14
2.5.1. YOLO 14
2.5.2. SSD 17
第3章 研究方法 18
3.1. 骨骼造影影像擷取 18
3.1.1. 腰椎影像擷取 19
3.1.2. 影像擷取限制 19
3.2. 腰椎影像訓練樣本 21
3.2.1. 腰椎位置定義 22
3.2.2. 標記腰椎樣本 24
3.3. 偵測模型訓練流程 26
3.3.1. YOLOv3模型訓練 27
3.3.2. SSD300模型訓練 29
第4章 實驗結果與探討 31
4.1. 實驗環境詳細規格 31
4.2. YOLOv3模型訓練結果 32
4.2.1. YOLOv3模型分類正確率之驗證 33
4.3. SSD300模型訓練結果 38
4.3.1. SSD300模型分類正確率之驗證 39
4.4. 腰椎自動定位結果比較 44
第5章 結論與未來展望 45
5.1. 結論 45
5.2. 未來展望 46
參考文獻 47

[1]衛生福利處統計部.[Online]. Available: https://dep.mohw.gov.tw/DOS/np-1776-113.html.
[2]Robert E.Coleman, JanetBrown, IngunnHolen, 2020, Abeloff's Clinical Oncology, Sixth Edition, Pages 809-830.e3
[3]González-Sistal, À., Sánchez, A. B., Carnero, M. H., & Morell, A. R. (2011). Advances in medical imaging applied to bone metastases (Vol. 16). chapter.
[4]Greco, M. (2020). Diagnostic Radioentomology. In Modern Beekeeping. IntechOpen.
[5]O'Malley, J. P., & Ziessman, H. A. (2020). Nuclear Medicine and Molecular Imaging: The Requisites E-Book. Elsevier.
[6]Azhari, H., Kennedy, J. A., Weiss, N., & Volokh, L. (2020). Nuclear Medicine: Planar and SPECT Imaging. In From Signals to Image (pp. 159-215). Springer, Cham.
[7]Wadah Ali, 2017, Development of Quality Control Phantom for Gamma Camera and SPECT System, Page 8.
[8]Brenner, A. I., Koshy, J., Morey, J., Lin, C., & DiPoce, J. (2012, January). The bone scan. In Seminars in nuclear medicine (Vol. 42, No. 1, pp. 11-26). WB Saunders.
[9]Bombardieri E, Aktolun C, Baum RP, et al. Bone scintigraphy: procedure guidelines for tumour imaging. Eur J Nucl Med Imaging 2003; 30:99-106.
[10]Love C, Din AS, Tomas MB, Kalapparambath TP, Palestero CJ. Radionuclide bone imaging: an illustrative review. RadioGraphics 2003; 23:341–358.
[11]Alpaydin, E. (2020). Introduction to machine learning. MIT press.
[12]Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2012). Imagenet classification with deep convolutional neural networks. In Advances in neural information processing systems (pp. 1097-1105).
[13]Kumar, A., & Verma, D. (2020, January). Number Plate Reorganization using Image Processing and machine learning Approaches: A Review. In 2020 International Conference on Computer Communication and Informatics (ICCCI) (pp. 1-4). IEEE.
[14]Ge, Z., Song, Z., Ding, S. X., & Huang, B. (2017). Data mining and analytics in the process industry: The role of machine learning. Ieee Access, 5, 20590-20616.
[15]Kucukyilmaz, T., Cambazoglu, B. B., Aykanat, C., & Baeza-Yates, R. (2017). A machine learning approach for result caching in web search engines. Information Processing & Management, 53(4), 834-850.
[16]Socher, R., Bauer, J., Manning, C. D., & Ng, A. Y. (2013, August). Parsing with compositional vector grammars. In Proceedings of the 51st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) (pp. 455-465).
[17]Stilgoe, J. (2018). Machine learning, social learning and the governance of self-driving cars. Social studies of science, 48(1), 25-56.
[18]Bhowmick, A., & Hazarika, S. M. (2016). Machine learning for e-mail spam filtering: review, techniques and trends. arXiv preprint arXiv:1606.01042.
[19]Erickson, B. J., Korfiatis, P., Akkus, Z., & Kline, T. L. (2017). Machine learning for medical imaging. Radiographics, 37(2), 505-515.
[20]Tzanakou, E. M. (Ed.). (2017). Supervised and unsupervised pattern recognition: feature extraction and computational intelligence. CRC press.
[21]Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2012). Imagenet classification with deep convolutional neural networks. In Advances in neural information processing systems (pp. 1097-1105).
[22]A Beginner's Guide To Understanding Convolutional Neural Networks. [Online]. Available: https://adeshpande3.github.io/A-Beginner%27sGuide-To-Understanding-Convolutional-Neural-Networks/.
[23]Y. LeCun et al., "Deep learning," nature, vol. 521, no. 7553, p. 436, 2015.
[24]Artificial Neural Network with Keras — An Example. [Online]. Available: https://medium.com/@cdabakoglu/artificial-neural-network-with-keras-d858f82f90c5.
[25]" Keras," [Online]. Available: https://keras.io/.
[26]"TensorFlow," [Online]. Available: https://www.tensorflow.org/.
[27]Redmon, J., Divvala, S., Girshick, R., & Farhadi, A. (2016). You only look once: Unified, real-time object detection. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 779-788).
[28]Redmon, J., & Farhadi, A. (2017). YOLO9000: better, faster, stronger. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 7263-7271).
[29]Yi, Z., Yongliang, S., & Jun, Z. (2019). An improved tiny-yolov3 pedestrian detection algorithm. Optik, 183, 17-23.
[30]Liu, W., Anguelov, D., Erhan, D., Szegedy, C., Reed, S., Fu, C. Y., & Berg, A. C. (2016, October). Ssd: Single shot multibox detector. In European conference on computer vision (pp. 21-37). Springer, Cham.
[31]Perez, H., Tah, J. H., & Mosavi, A. (2019). Deep learning for detecting building defects using convolutional neural networks. Sensors, 19(16), 3556.
[32]Adedokun, O. A., & Burgess, W. D. (2012). Analysis of paired dichotomous data: A gentle introduction to the McNemar test in SPSS. Journal of MultiDisciplinary Evaluation, 8(17), 125-131.
[33]Jung, Y. (2018). Multiple predicting K-fold cross-validation for model selection. Journal of Nonparametric Statistics, 30(1), 197-215.