臺灣博碩士論文加值系統

胸部 X 光檢查 (Chest X-ray) 是臨床上最常進行的項目之一，也是一般身體健康檢查必要之檢查項目，由於目前臨床上胸部 X 光檢查之判讀主要由放射科醫師或相關醫護人員進行，在繁忙的臨床工作中不免有遺漏之處，因此本研究目的在於利用深度學習輔助計算心胸比， 以協助臨床醫護人員胸部 X 光檢查之判讀。
心胸比測量方式多元，為評估與檢查心臟疾病最基本之檢查工具以及評估心臟是否有擴大之重要常用指標，目前計算方式主要由醫護人員於心臟劃一條垂直參考線並量出心臟左緣及右緣間之距離再除以胸部寬度， 並依據其值將心臟分為輕度、中度及重度增大。一般人心胸比小於 0.5，表示「較好的」、「較健康的」，而對於慢性心臟衰竭者、長期高血壓或尿毒症長期透析之患者而言，常會出現心胸比大於 0.5 之狀況，表示患者心臟有擴大跡象，其不僅容易出現呼吸喘、透析中低血壓，甚至可能造成心律不整，而為了減輕醫護人員計算心胸比所導致之工作負荷並使患者獲得完善與精準之醫療照護，本研究整體模型精確性達到 78%之表現，期許透過此表現輔以提升臨床工作效率、降低失誤率等狀況發生。

Chest X-ray is one of the most common and the necessary item of health examination in clinical examinations. However, the examination analysis is the professional skills, health care workers must spend lots of time during their busy clinical works and cause trouble. Therefore, the purpose of this study use Deep Learning to assist health care workers calculate the cardiothoracic ratio (CTR) in Chest X-ray.
The cardiothoracic ratio is measured in a variety of ways. It is the most basic tool for assessing and examining heart disease and an important common index for assessing whether the heart is enlarged. At present, the method is mainly by medical workers to draw a vertical reference line on the heart and measure the left and right edges of the heart. The distance between the limbs is divided by the chest width, and the heart is classified into mild, moderate and severe enlargement according to its value. Generally, the CTR is less than 0.5, which means "better" and "healthier", however, patients with chronic heart failure、long-term hypertension or long-term dialysis patients with uremia the value of CTR often over 0.5 , which indicate that patient has signs of enlargement, which is not only prone to asthma, hypotension during dialysis, and may even cause arrhythmia.
In order to decrease health care worker’s working loading and providing the accurate medical care to patient. In this study, the accuracy of the model reaches 78%. According to this perspective, we expect that this performance will help improve clinical work efficiency and reduce error rates.

圖目錄
表目錄
第一章、緒論
1.1研究背景
1.2研究動機
1.3研究目的與問題
1.4預期成果與貢獻
第二章、文獻探討
2.1 胸部X光影像
2.2 深度學習於胸部X光及醫學影像之應用
2.3 物件偵測文獻探討
2.4 物件偵測演算法
2.4.1 YOLOv1
2.4.2 YOLOv2
2.4.3 YOLOv3
2.4.4 YOLOv4
第三章、材料與研究方法
3.1 實驗材料
3.2 研究架構與流程
3.3研究方法步驟說明
3.3.1 影像前處理
3.3.2 深度學習模型訓練-YOLOv4模型
第四章、實驗評估
4.1 實驗環境
4.2 實驗評估指標
4.2.1 專家判斷
4.2.2 檢測準確率
4.2.3 混淆矩陣
4.2.4召回率(Recall)
4.2.5檢測精度(AP)
4.3 實驗結果
第五章、目前進度與未來工作計畫
5.1 結果說明
5.2 未來展望
參考文獻

大林慈濟醫院影像醫學科(2010年4月21日)。X光在醫學上的應用(X-ray in Medical Application)。https://dl.tzuchi.com.tw/medi/node/74
于承宇 (2022年3月22日)。心臟肥大｜定義、原因、症狀、診斷、治療。Hello醫師。https://helloyishi.com.tw/heart-health/heart-disease/cardiomegaly/
中華民國醫事放射學會(2014年6月17日)。認識胸部X光簡易判讀。http://www.twsrt.org.tw/twsrt/upfile/files/2014617144517.pdf
台灣腎臟醫學會(2022年4月30日) 。血液透析及腹膜透析訪視作業評量標準說明_111年確認版。https://www.tsn.org.tw/hemodialysis.html?archiveSubCategoryId=AR0001-2
何明霖 (2020年8月)。胸部X光。肺科醫師 Pulmonologist。https://www.chestdoctor.org/%E5%89%AF%E6%9C%AC-%E8%83%B8%E9%83%A8x%E5%85%89
沐風流雲 (2021年12月8日) 。 什麼是anchor-based 和anchor free？。 知乎 。 https://www.zhihu.com/question/356551927
杜杏慧(2013年3月31日)。胸部X光加照側面，提升診斷準確度。防癆雜誌，19-20。https://www.tb.org.tw/uploads/102_sp/05.pdf
陳芳毓 (2021年3月16日)。智慧醫療懶人包》八張圖，一次搞懂智慧醫療的定義、應用與台灣的挑戰。未來城市。https://futurecity.cw.com.tw/article/1916?rec=i2i&from_id&from_index=3
鄭宏益 (2007年10月15日)。心臟大小。隨意窩。 https://blog.xuite.net/hyc0629/twblog/119764144#
練冠霆（2019年11月7日）。發現X射線。科技大觀園。https://reurl.cc/anmKnD
蔣西然 (2021) 。 深度學習人工智慧技術在醫學影像輔助分析中的應用。 中國醫療集團設備， 36 (6)， 164。 https://www.hea.com.tw/magazineDetail.asp?id=121
Bochkovskiy, A., Wang, C. Y., & Liao, H. Y. (2020). YOLOv4: Optimal Speed and Accuracy of Object Detection. Computer Science, 2004. https://doi.org/10.48550/arXiv.2004.10934
Chamveha, I., Promwiset, T., Tongdee, T., Saiviroonporn, P., & Chaisangmongkon, W. (2020). Automated Cardiothoracic Ratio Calculation And Cardiomegaly Detection Using Deep Learning Approach. Image and Video Processing. https://doi.org/10.48550/arXiv.2002.07468
Chen, C. M., Chou, Y. H., Tagawa, N., & Do, Y. (2013). Computer-Aided Detection and Diagnosis in Medical Imaging. Computational and Mathematical Methods in Medicine. https://doi.org/10.1155/2013/790608
Chan, H. P., Doi, K., Galhotra, S., Vyborny, C. J., Mahon, H. M., & Jokich, P. M. (1987). Image Feature Analysis and Computer-Aided Diagnosis in Digital Radiography. Medical Physics, 14(4), 538-48. http://dx.doi.org/10.1118/1.596065
Cha, K. H., Hadjiiski, L. M., Samala, R. K., Chan, H. P., Cohan. R. H., Caoili, E. M., Paramagul, C., Alva, A., & Weizer, A. Z. (2016). Bladder Cancer Segmentation in CT for Treatment Response Assessment：Application of Deep-Learning Convolution Neural Network—A Pilot Study. Tomography, 2(4), 421-429. http://dx.doi.org/ 10.18383/j.tom.2016.00184
Danzer, C. S. (1919). The Cardiothoracic Ratio：An Index of Cardiac Enlargement Bibliography. The American Journal of the Medical Sciences, 157(4), :513
Gupte, T., Niljikar, M., Gawali, M., Kulkarni, V., Kharat, A., & Pant, A. (2021, Aug 5-6). Deep Learning Models for Calculation of Cardiothoracic Ratio from Chest Radiographs for Assisted Diagnosis of Cardiomegaly. Conference on Computer Vision and Pattern Recognition [Conference presentation]. International Conference on Artificial Intelligence, Big Data, Computing and Data Communication Systems. South Africa. http://icabcd.org/2021/
Ghafoorian, M., Karssemeijer, N., Heskes, T., Inge, W. M. van Uden., Sanchez, C. I., Litjens, G., Frank, E. de Leeuw, Bram, van Ginneken, Marchiori, E., & Platel, B. (2017). Location Sensitive Deep Convolutional Neural Networks for Segmentation of White Matter Hyperintensities. Scientific Reports, 7(1),5110. http://dx.doi.org/10.1038/s41598-017-05300-5
Gaddam, P. C. S. K., & Sunkara, P., (2016). Advanced Image Processing Using Histogram Equalization and Android Application Implementation [Unpublished doctoral dissertation]. Blekinge Tekniska Högskola. https://www.diva-portal.org/smash/get/diva2:1063170/FULLTEXT01.pdf
Hegde, M. (2013). A comprehensive Approach to Congenital Heart Diseases. Jaypee Brothers Medical Publishers. doi:10.5005/jp/books/12075_13
Khush Patel (2019年5月28日)。 Custom Object Detection using TensorFlow from Scratch 。 Medium。 https://towardsdatascience.com/custom-object-detection-using-tensorflow-from-scratch-e61da2e10087
Kooi, T., Ginneken, B., Karssemeijer, N., & Heeten, A. (2017). Discriminating Solitary Cysts from Soft Tissue Lesions in Mammography Using a Pretrained Deep Convolutional Neural Network. Medical Physics, 44(3), 1017-1027. http://dx.doi.org/ 10.1002/mp.12110
Li, Z. N., Hou, Z. H., Chen. C., Hao, Z., An, Y. Q., Liang, S., & Lu, A. B. (2019). Automatic Cardiothoracic Ratio Calculation With Deep Learning. Institute of Electrical and Electronics Engineers Access, 7, 2169-3536. https://doi.org/10.1109/ACCESS.2019.2900053
Nowak, S., Faron, A., Luetkens, J. A., Geißler, H. L., Praktiknjo, M., Block, W., Thomas, D., & Sprinkart, A. M. (2020). Fully Automated Segmentation of Connective Tissue Compartments for CT-Based Body Composition Analysis: A Deep Learning Approach. Investigative Radiology, 55(6), 357-366. http://dx.doi.org/ 10.1097/RLI.0000000000000647
Perrin, D. P., Bueno, A., Rodriguez, A., Marx, G., & Nido, P. (2017, Feb 11-16). Application of Convolutional Artificial Neural Networks to Echocardiograms for Differentiating Congenital Heart Diseases in a Pediatric Population [Conference presentation]. 2017 SPIE Medical Imaging Conference, Orlando, FL, United States. https://spie.org/news/spie-professional-magazine-archive/2016-october/science-of-deep-learning-precision-medicine-at-spie-medical-imaging-2017?SSO=1
Redmon, J., Divvala, S., Girshick, R., & Farhadi, A. (2016). You Only Look Once: Unified, Real-Time Object Detection. Conference on Computer Vision and Pattern Recognition, 12, 1063-6919. http://dx.doi.org/10.1109/CVPR.2016.91
Redmon, J., & Farhadi, A. (2016). YOLO9000：Better , Faster , Stronger. Conference on Computer Vision and Pattern Recognition, 09, 1063-6919. http://dx.doi.org/ 10.1109/CVPR.2017.690
Redmon, J., & Farhadi, A. (2018). YOLOv3: An Incremental Improvement. Computer Science, 1804. https://doi.org/10.48550/arXiv.1804.02767
Ravi, D., Wong, C., Deligianni, F., Berthelot, M., Perez, J. A., Lo, B., & Yang, G. Z. (2017). Deep learning for health informatics. IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS, 21(1), 4-21. http://dx.doi.org/ 10.1109/JBHI.2016.2636665
Tommy Huang (2018年7月10日)。機器學習\統計方法: 模型評估-驗證指標(validation index)。Medium。https://chih-sheng-huang821.medium.com/%E6%A9%9F%E5%99%A8%E5%AD%B8%E7%BF%92-%E7%B5%B1%E8%A8%88%E6%96%B9%E6%B3%95-%E6%A8%A1%E5%9E%8B%E8%A9%95%E4%BC%B0-%E9%A9%97%E8%AD%89%E6%8C%87%E6%A8%99-b03825ff0814
Tommy Huang (2018年9月4日)。深度學習-物件偵測:You Only Look Once (YOLO)。Medium。https://chih-sheng-huang821.medium.com/%E6%B7%B1%E5%BA%A6%E5%AD%B8%E7%BF%92-%E7%89%A9%E4%BB%B6%E5%81%B5%E6%B8%AC-you-only-look-once-yolo-4fb9cf49453c
Tubiana, M. (1996). Wilhelm Conrad Röntgen and the discovery of X-rays. Bulletin de L'Académie Nationale de Médecine, 180(1), 97-108
Truszkiewicz, K., Poreba, R., & Gac, P. (2016). Radiological Cardiothoracic Ratio in Evidence-Based Medicine. Signal Processing and Communications Applications Conference , 10(9). https://doi.org/10.3390/jcm10092016
Taçyildiz, Z.C., Kiliç, E., Budak, A., & Karataş, H. (2021). (2021, Jun 9-11). Cardiothoracic Ratio Calculation and Cardiomegaly Detection Based on Object Detection [Conference presentation]. 2021 Signal Processing and Communications Applications Conference, Istanbul, Turkey. https://ieeexplore.ieee.org/xpl/conhome/9477764/proceeding
Yotsueda, R., Taniguchi, M., Tanada, S., Eriguchi, M., Fujisaki, K., Torisu, K., Masutani, K., Hirakata, H., Kitazono, T., & Tsuruya, K. (2017). Cardiothoracic Ratio and All-Cause Mortality and Cardiovascular Disease Events in Hemodialysis Patients: The Q-Cohort Study. American Journal of Kidney Diseases, 70(1), 84-92. http://dx.doi.org/10.1053/j.ajkd.2016.11.026