Publications-Theses
Article View/Open
Publication Export
-
題名 以知識圖譜解譯人工智慧肺癌影像診斷之脈絡
Using knowledge graph to interpret the context of artificial intelligence lung cancer image diagnosis作者 梁芸瑄
Liang, Yun-Shiuan貢獻者 羅崇銘
Lo, Chung-Ming
梁芸瑄
Liang, Yun-Shiuan關鍵詞 知識圖譜
詮釋資料
肺癌
醫學影像
支氣管鏡檢查
Knowledge graph
Metadata
Lung cancer
Medical image
Bronchoscopy日期 2021 上傳時間 2-Sep-2021 16:34:38 (UTC+8) 摘要 癌症是全球人口主要死亡原因之一，在2020年就有將近一千萬人因此死亡，而在這一千萬人之中佔比最大的便是肺癌。根據衛生福利部統計，在2019年癌症依舊為導致國人死亡的最大原因，是第二名心臟疾病的2.5倍。在這其中，肺癌長期位居癌症死亡率的第一名，並在近十年間更有增長的趨勢。 多數癌症引發的死亡率如此之高，最大的問題就在於前期症狀不明顯、難以察覺，無法即刻進行積極地治療。為了解決此問題，醫學影像應用在早期的診斷之助益不可小覷。其可為病情留下紀錄，做醫師研讀、複查之用，更重要的是醫學影像是以非侵入性方式取得組織內部情形，這樣的方式與解剖相比可大量減輕對病患的傷害。 在肺部的診斷中，醫師會針對病人的情況採取不同的醫學影像檢查方式，而一般會使用到的影像學檢查有：胸部X光、胸部電腦斷層掃描檢查(computed tomography, CT)、核磁共振檢查(magnetic resonance imaging, MRI)和正子放射斷層掃描(positron emission tomography, PET)。除此之外，還有更深入的影像檢查：支氣管鏡檢查、支氣管鏡超音波(endobronchial ultrasound, EBUS)、胸腔內視鏡檢查等。 然而醫學影像用途廣泛，相關數據的產生快速增長。要處理大量產生的醫學影像會是不小的負擔，因此，若能以人工智慧(Artificial Intelligence, AI)電腦輔助診斷系統作為第二閱片者輔佐醫生進行影像的判讀，不失為提高診斷穩定性的方法之一。電腦輔助診斷系統(computer-aided diagnosis, CAD)提供了具有一致性的量化作業，使診斷過程更有效率以及更加精確，在如此優異的表現下仍舊保證了快速的決策速度。 然而，不同類型的肺部影像檢查有著不同的形態與特徵，不是那麼容易讓使用者可以理解，而且電腦輔助診斷系統解析了大量的檢查影像資訊會形成許多零散的知識。觀察在不同儀器下的成像，可以發現他們的特徵皆不同，而醫生和人工智慧觀察影像時所關注的重點也不同，因此需要有一個系統連結影像彼此之間的關係，以此幫助使用者理解，並從中學習、找到需要的資訊。 傳統上以詮釋資料作為描述肺部影像的特徵，但詮釋資料雖然可以將影像轉為具統一性的文字描述，卻有著過於平面的問題，在擁有大量資料時會無法快速呈現整體重點，因此需要由知識圖譜來進行組織、呈現不同影像與特徵之間的關聯性，進而發掘出其所代表的影像診斷可能情況。 研究首先搜集電腦輔助診斷系統對白光支氣管鏡檢查、自體螢光支氣管鏡檢查與支氣管鏡彈性超音波檢查的成像診斷結果與醫生診斷肺癌影像所使用的判讀規則，了解兩者對不同檢查中腫瘤觀察的細節，並參考肺癌腫瘤分期方法權衡對肺癌影響的要點後將細節描述轉化為研究中所使用的詮釋資料。以及搜集電腦輔助診斷系統所使用的肺部影像照片，2015年9月至2017年4月對雙和醫院的70名患者進行了篩查產生的白光支氣管鏡檢查與自體螢光支氣管鏡檢查所產生之肺部影像，以及2019年2月至12月對雙和醫院患者進行支氣管鏡彈性超音波檢查影像共114張做研究材料。接著擷取詮釋資料重點建構知識圖譜，將詮釋資料的重點置於知識圖譜中，並建構資訊之間的關聯性。同時，不同層級與節點間有著關聯的描述，讓使用者得知資訊彼此間的關係，進而找出其中的關聯性以產生具邏輯的知識圖譜。 知識圖譜建立後，可以讓使用者了解醫生與人工智慧如何進行肺部影像診斷，以及分辨出不同影像間的共同與相異處為何。未來可將實驗領域移轉到人體其他不同部位的醫學影像上進行知識圖譜的製作與分析，提供一個科技新工具輔助醫學影像診斷。
Cancer is one of the main causes of death in the global population. Nearly 10 million people have died from this cause in 2020, and lung cancer accounts for the largest proportion of these million people. According to statistics from the Ministry of Health and Welfare, cancer is the leading cause of death for citizens in 2019, 2.5 times that of the second leading cause, heart disease. Among them, lung cancer has long ranked first in cancer mortality, which has been increasing in the past ten years. The mortality rate caused by most cancers is high. The biggest problem is that the early symptoms are not obvious, difficult to detect, and cannot be treated immediately. In order to solve this problem, the application of medical imaging in early diagnosis cannot be underestimated. It can keep a record of the condition for doctors to study and review, and more importantly, medical imaging is a non-invasive way to obtain the internal situation of the tissue. Compared with anatomy, this method can greatly reduce the harm to the patient. In the diagnosis of the lungs, doctors will adopt different medical imaging examination methods according to the patient`s condition, and the imaging examinations generally used are: chest X-ray, chest computed tomography, magnetic resonance imaging and positron emission tomography. In addition, there are more in-depth imaging examinations: bronchoscopy, bronchoscopy ultrasound, thoracic endoscopy, etc. As medical imaging is widely used, the production of related data is growing rapidly. It is not a small burden to process a large number of generated medical images. Therefore, if the Artificial Intelligence computer-aided diagnosis system can be used as the second reader to assist the doctor in the interpretation of the images, it will improve the stability of the diagnosis. One of the methods─computer-aided diagnosis─provides consistent quantitative operations, making the diagnosis process more efficient and accurate, and still ensuring rapid decision-making under such excellent performance. Different types of lung imaging examinations have different shapes and characteristics, which are not so easy for users to understand. The computer-aided diagnosis system analyzes a large amount of examination image information to form a lot of scattered knowledge. Observing the imaging under different instruments, we can find that their characteristics are different, and the focus of the doctor and AI when observing the image is also different. Therefore, a system is needed to link the relationship between the images to help users understand, learn and locate the needed information. Traditionally, metadata is used to describe the characteristics of lung images. Although interpretive data can transform the image into a unified text description, it has the problem of being too flat. When there is a large amount of data, it will not be able to quickly present the overall focus. To address this challenge, the knowledge graph is used to organize and present the correlation between different images and features, and then explore the possible image diagnosis situations. The research first collects the imaging diagnosis results of white light bronchoscopy, autofluorescent bronchoscopy, and bronchoscopy elastic ultrasonic examination by the computer-aided diagnosis system and the interpretation rules used by doctors to diagnose lung cancer images. This allows investigation of the detaile tumor characteristics in the two different examinations. Reference to the lung cancer staging method is made to weigh the main points of the impact on lung cancer, and convert the detailed description into the metadata used in the study. As well as the collection of lung imaging photos used by the computer-aided diagnosis system, from September 2015 to April 2017, 70 patients in Shuanghe Hospital were screened for white light bronchoscopy and autofluorescent bronchoscopy. Data source consists of a total of 114 images of lungs and bronchoscopy ultrasound examinations performed on patients in Shuanghe Hospital from February to December 2019. Research proceudres include: extract the metadata and construct the knowledge graph, place the emphasis of the metadata in the knowledge graph, and construct the correlation between the information. At the same time, there are related descriptions between different levels and nodes, allowing users to know the relationship between the information, and then find the connection among them to generate a logical knowledge graph. After the knowledge graph is established, users can understand how doctors and artificial intelligence perform lung imaging diagnosis, and distinguish the similarities and differences between different images. In the future, the experimental field can be applied to medical images of other different parts of the human body for the production and analysis of knowledge graph. This research provides a new technological tool to assist medical imaging diagnosis.參考文獻 [1] World Health Organization. (2021). Cancer. Available: https://www.who.int/news-room/fact-sheets/detail/cancer[2] K. Everington. (2019). Taiwan has 15th highest lung cancer rate in world. Available: https://www.taiwannews.com.tw/en/news/3825780[3] J. M. Croswell, D. F. Ransohoff , and B. S. Kramer. Principles of Cancer Screening: Lessons from History and Study Design Issues, Semin. Oncol., vol. 37, p.202-215. doi: 10.1053/j.seminoncol.2010.05.006.[4] 衛生福利部國民健康署（2020）。106年癌症登記報告。取自 https://www.hpa.gov.tw/Pages/Detail.aspx?nodeid=269&pid=12235[5] T. Yorifuji and S. Kashima, "Air pollution: another cause of lung cancer," Lancet Oncol., vol. 14, p. 788-789. doi: 10.1016/S1470-2045(13)70302-4.[6] J. L. Mumford, et al., "Lung cancer and indoor air pollution in Xuan Wei, China," Science, vol. 235, p. 217-220. doi: 10.1126/science.3798109.[7] Fredrik Nyberg, et al., "Urban Air Pollution and Lung Cancer in Stockholm," Epidemiology vol. 11, No. 5, pp. 487-495. Available: https://www.jstor.org/stable/3703988?seq=1[8] P. Broderick, et al., "Deciphering the Impact of Common Genetic Variation on Lung Cancer Risk: A Genome-Wide Association Study," Cancer Res., vol. 69, p. 6633-6641. doi: 10.1158/0008-5472.CAN-09-0680.[9] T. S. Panchabhai, M. Ghobrial and A. C. Mehta, "History of Bronchoscopy: The Evolution of Interventional Pulmonology," Interventions in Pulmonary Medicine, p. 609-621. doi: 10.1007/978-3-319-58036-4_39.[10] S. Leong, T. Shaipanich, S. Lam and K. Yasufuku, "Diagnostic bronchoscopy--current and future perspectives," J. Thorac. Dis., vol. 5, p. S498-S510. doi: 10.3978/j.issn.2072-1439.2013.09.08.[11] H. Müller, N. Michoux, D. Bandon and A. Geissbuhler, "A review of content-based image retrieval systems in medical applications—clinical benefits and future directions," Int. J. Med. Inf., vol. 73, p. 1-23. doi: 10.1016/j.ijmedinf.2003.11.024.[12] S. Jha and E. J. Topol, "Adapting to Artificial Intelligence: Radiologists and Pathologists as Information Specialists," JAMA, vol. 316, p. 2353-2354. doi: 10.1001/jama.2016.17438.[13] E. El-Dahshan, H. M. Mohsen, K. Revett and A. M. Salem, "Computer-aided diagnosis of human brain tumor through MRI: A survey and a new algorithm," Expert Syst. Appl., vol. 41, p. 5526-5545. doi: 10.1016/j.eswa.2014.01.021.[14] F. Pesapane, M. Codari and F. Sardanelli, "Artificial intelligence in medical imaging: threat or opportunity? Radiologists again at the forefront of innovation in medicine," Eur. Radiol. Exp., vol. 2(1), p. 35. doi: 10.1186/s41747-018-0061-6.[15] G. Chartr, et al., "Deep Learning: A Primer for Radiologists," RadioGraphics, vol. 37, p. 2113-2131. doi: 10.1148/rg.2017170077.[16] The American College of Radiology. ACR BI-RADS Atlas® 5th Edition. Available: https://www.acr.org/Clinical-Resources/Reporting-and-Data-Systems/Bi-Rads[17] Visual Resources Association Data Standards Committee. (2002). VRA Core Categories, Version 3.0. Available: https://marciazeng.slis.kent.edu/metadatabasics/Portuguese/schemas/VRA3.htm[18] Phillips, John T, "Metadata - Information about electronic records," Sch. J., vol. 29, p. 52.[19] J. Greenberg, "Understanding Metadata and Metadata Schemes," Cat. Classif. Q., vol. 40, p. 17-36. doi: 10.1300/J104v40n03_02.[20] The International Federation of Library Associations and Institutions,., Digital Libraries: Metadata Resources. Available: https://www.ifla.org/node/9337[21] 余顯強，「以資訊處理觀點論Metadata之本質與意涵」，教育資料與圖書館學，卷 45(2)，頁 249-266。[22] Brad, E., "Metadata and Its Application." Library and Information Science Journals. Available: https://www.questia.com/library/journal/1G1-93085596/metadata-and-its-application[23] 陳昭珍、蕭伯瑜（2006）。全集層次詮釋資料之應用研究:以EAD為例（學位論文）。取自國立臺灣師範大學圖書館機構典藏。（系統編號GN0692150049）[24] 王德華、袁萍、田榮華、徐亮（2014），醫學影像檔案資料的現代化管理，中國病案，15，14-15。[25] 郭莉珠（2004）。技術研究現狀與發展方向，檔案學通訊，1，74-76。[26] A. Kumar, J. Kim, L. Wen, M. Fulham and D. Feng, "A graph-based approach for the retrieval of multi-modality medical images," Med. Image Anal., vol. 18(2), p. 330-342. doi: 10.1016/j.media.2013.11.003.[27] T. Yu, et al., "Knowledge graph for TCM health preservation: Design, construction, and applications, " Artif. Intell. Med., vol. 77, p. 48-52. doi: 10.1016/j.artmed.2017.04.001.[28] S. M. S. Hasan, D. Rivera, X.-C. Wu, E. B. Durbin, J. B. Christian and G. Tourassi, "Knowledge Graph-Enabled Cancer Data Analytics, " IEEE J. Biomed. Health Inform., vol. 24, p. 1952-1967. doi: 10.1109/JBHI.2020.2990797.[29] 張幟（2018）。Neo4j圖形資料庫權威指南：傲視大數據時代的先端利器。臺北市：深石數位。[30] P.-H. Feng, Y.-T. Lin and C.-M. Lo, "A machine learning texture model for classifying lung cancer subtypes using preliminary bronchoscopic findings," Med. Phys., vol. 45, p. 5509-5514. doi: 10.1002/mp.13241.[31] P. H. Feng, T. T. Chen, Y. T. Lin, S. Y. Chiang and C. M. Lo, "Classification of lung cancer subtypes based on autofluorescence bronchoscopic pattern recognition: A preliminary study," Comput. Methods Programs Biomed., vol. 163, p. 33-38. doi: 10.1016/j.cmpb.2018.05.016.[32] Y. Li, et al., "Comparison of the autofluorescence bronchoscope and the white light bronchoscope in airway examination," Chin. J. Cancer, vol. 29, p. 1018-22. doi: 10.5732/cjc.010.10298.[33] A. K. Banerjee. Detection of early lung cancer: Autofluorescence bronchoscopy and investigational modalities - UpToDate. Available: https://www.uptodate.com/contents/detection-of-early-lung-cancer-autofluorescence-bronchoscopy-and-investigational-modalities[34] F. J. F. Herth, R. Eberhardt, D. Anantham, D. Gompelmann, M. W. Zakaria and A. Ernst, "Narrow-Band Imaging Bronchoscopy Increases the Specificity of Bronchoscopic Early Lung Cancer Detection," J. Thorac. Oncol., vol. 4, p. 1060-1065. doi: 10.1097/JTO.0b013e3181b24100.[35] A. Y. Altonbary, H. Hakim and A. M. El-Shamy, "Diagnostic Efficacy of Endoscopic Ultrasound Elastography in Differentiating Solid Pancreatic Lesions: A Single-Center Experience," Clin. Endosc., vol. 52, p. 360-364. doi: 10.5946/ce.2018.160.[36] S. Mittal, A. Mohan, V. Hadda and K. Madan, "Endobronchial ultrasound elastography in mediastinal lymphadenopathy: Report of two cases and systematic review of literature," Lung India, vol. 36, p. 149-153. doi: 10.4103/lungindia.lungindia_349_17.[37] H.-Y. He, M. Huang, J. Zhu, H. Ma and X.-D. Lyu, "Endobronchial Ultrasound Elastography for Diagnosing Mediastinal and Hilar Lymph Nodes," Chin. Med. J. (Engl.), vol. 128, p. 2720-2725. doi: 10.4103/0366-6999.167296.[38] Nakajima, T., et al., "Elastography for Predicting and Localizing Nodal Metastases during Endobronchial Ultrasound," Respiration, vol. 90, pp. 499-506. doi: 10.1159/000441798.[39] G. A. Silvestri, et al., "Methods for Staging Non-small Cell Lung Cancer: Diagnosis and Management of Lung Cancer, 3rd ed: American College of Chest Physicians Evidence-Based Clinical Practice Guidelines," Chest, vol. 143. doi: 10.1378/chest.12-2355.[40] 三軍總醫院（2021）。肺癌的診斷與處置。取自https://wwwv.tsgh.ndmctsgh.edu.tw/unit/10041/14484[41] 連啟惇（2012）。早期肺癌檢查新利器--窄頻染色及自體螢光支氣管鏡。取自 http://www.kmuh.org.tw/www/kmcj/data/10101/18.htm 描述 碩士
國立政治大學
圖書資訊與檔案學研究所
108155003資料來源 http://thesis.lib.nccu.edu.tw/record/#G0108155003 資料類型 thesis dc.contributor.advisor 羅崇銘 zh_TW dc.contributor.advisor Lo, Chung-Ming en_US dc.contributor.author (Authors) 梁芸瑄 zh_TW dc.contributor.author (Authors) Liang, Yun-Shiuan en_US dc.creator (作者) 梁芸瑄 zh_TW dc.creator (作者) Liang, Yun-Shiuan en_US dc.date (日期) 2021 en_US dc.date.accessioned 2-Sep-2021 16:34:38 (UTC+8) - dc.date.available 2-Sep-2021 16:34:38 (UTC+8) - dc.date.issued (上傳時間) 2-Sep-2021 16:34:38 (UTC+8) - dc.identifier (Other Identifiers) G0108155003 en_US dc.identifier.uri (URI) http://nccur.lib.nccu.edu.tw/handle/140.119/136920 - dc.description (描述) 碩士 zh_TW dc.description (描述) 國立政治大學 zh_TW dc.description (描述) 圖書資訊與檔案學研究所 zh_TW dc.description (描述) 108155003 zh_TW dc.description.abstract (摘要) 癌症是全球人口主要死亡原因之一，在2020年就有將近一千萬人因此死亡，而在這一千萬人之中佔比最大的便是肺癌。根據衛生福利部統計，在2019年癌症依舊為導致國人死亡的最大原因，是第二名心臟疾病的2.5倍。在這其中，肺癌長期位居癌症死亡率的第一名，並在近十年間更有增長的趨勢。 多數癌症引發的死亡率如此之高，最大的問題就在於前期症狀不明顯、難以察覺，無法即刻進行積極地治療。為了解決此問題，醫學影像應用在早期的診斷之助益不可小覷。其可為病情留下紀錄，做醫師研讀、複查之用，更重要的是醫學影像是以非侵入性方式取得組織內部情形，這樣的方式與解剖相比可大量減輕對病患的傷害。 在肺部的診斷中，醫師會針對病人的情況採取不同的醫學影像檢查方式，而一般會使用到的影像學檢查有：胸部X光、胸部電腦斷層掃描檢查(computed tomography, CT)、核磁共振檢查(magnetic resonance imaging, MRI)和正子放射斷層掃描(positron emission tomography, PET)。除此之外，還有更深入的影像檢查：支氣管鏡檢查、支氣管鏡超音波(endobronchial ultrasound, EBUS)、胸腔內視鏡檢查等。 然而醫學影像用途廣泛，相關數據的產生快速增長。要處理大量產生的醫學影像會是不小的負擔，因此，若能以人工智慧(Artificial Intelligence, AI)電腦輔助診斷系統作為第二閱片者輔佐醫生進行影像的判讀，不失為提高診斷穩定性的方法之一。電腦輔助診斷系統(computer-aided diagnosis, CAD)提供了具有一致性的量化作業，使診斷過程更有效率以及更加精確，在如此優異的表現下仍舊保證了快速的決策速度。 然而，不同類型的肺部影像檢查有著不同的形態與特徵，不是那麼容易讓使用者可以理解，而且電腦輔助診斷系統解析了大量的檢查影像資訊會形成許多零散的知識。觀察在不同儀器下的成像，可以發現他們的特徵皆不同，而醫生和人工智慧觀察影像時所關注的重點也不同，因此需要有一個系統連結影像彼此之間的關係，以此幫助使用者理解，並從中學習、找到需要的資訊。 傳統上以詮釋資料作為描述肺部影像的特徵，但詮釋資料雖然可以將影像轉為具統一性的文字描述，卻有著過於平面的問題，在擁有大量資料時會無法快速呈現整體重點，因此需要由知識圖譜來進行組織、呈現不同影像與特徵之間的關聯性，進而發掘出其所代表的影像診斷可能情況。 研究首先搜集電腦輔助診斷系統對白光支氣管鏡檢查、自體螢光支氣管鏡檢查與支氣管鏡彈性超音波檢查的成像診斷結果與醫生診斷肺癌影像所使用的判讀規則，了解兩者對不同檢查中腫瘤觀察的細節，並參考肺癌腫瘤分期方法權衡對肺癌影響的要點後將細節描述轉化為研究中所使用的詮釋資料。以及搜集電腦輔助診斷系統所使用的肺部影像照片，2015年9月至2017年4月對雙和醫院的70名患者進行了篩查產生的白光支氣管鏡檢查與自體螢光支氣管鏡檢查所產生之肺部影像，以及2019年2月至12月對雙和醫院患者進行支氣管鏡彈性超音波檢查影像共114張做研究材料。接著擷取詮釋資料重點建構知識圖譜，將詮釋資料的重點置於知識圖譜中，並建構資訊之間的關聯性。同時，不同層級與節點間有著關聯的描述，讓使用者得知資訊彼此間的關係，進而找出其中的關聯性以產生具邏輯的知識圖譜。 知識圖譜建立後，可以讓使用者了解醫生與人工智慧如何進行肺部影像診斷，以及分辨出不同影像間的共同與相異處為何。未來可將實驗領域移轉到人體其他不同部位的醫學影像上進行知識圖譜的製作與分析，提供一個科技新工具輔助醫學影像診斷。 zh_TW dc.description.abstract (摘要) Cancer is one of the main causes of death in the global population. Nearly 10 million people have died from this cause in 2020, and lung cancer accounts for the largest proportion of these million people. According to statistics from the Ministry of Health and Welfare, cancer is the leading cause of death for citizens in 2019, 2.5 times that of the second leading cause, heart disease. Among them, lung cancer has long ranked first in cancer mortality, which has been increasing in the past ten years. The mortality rate caused by most cancers is high. The biggest problem is that the early symptoms are not obvious, difficult to detect, and cannot be treated immediately. In order to solve this problem, the application of medical imaging in early diagnosis cannot be underestimated. It can keep a record of the condition for doctors to study and review, and more importantly, medical imaging is a non-invasive way to obtain the internal situation of the tissue. Compared with anatomy, this method can greatly reduce the harm to the patient. In the diagnosis of the lungs, doctors will adopt different medical imaging examination methods according to the patient`s condition, and the imaging examinations generally used are: chest X-ray, chest computed tomography, magnetic resonance imaging and positron emission tomography. In addition, there are more in-depth imaging examinations: bronchoscopy, bronchoscopy ultrasound, thoracic endoscopy, etc. As medical imaging is widely used, the production of related data is growing rapidly. It is not a small burden to process a large number of generated medical images. Therefore, if the Artificial Intelligence computer-aided diagnosis system can be used as the second reader to assist the doctor in the interpretation of the images, it will improve the stability of the diagnosis. One of the methods─computer-aided diagnosis─provides consistent quantitative operations, making the diagnosis process more efficient and accurate, and still ensuring rapid decision-making under such excellent performance. Different types of lung imaging examinations have different shapes and characteristics, which are not so easy for users to understand. The computer-aided diagnosis system analyzes a large amount of examination image information to form a lot of scattered knowledge. Observing the imaging under different instruments, we can find that their characteristics are different, and the focus of the doctor and AI when observing the image is also different. Therefore, a system is needed to link the relationship between the images to help users understand, learn and locate the needed information. Traditionally, metadata is used to describe the characteristics of lung images. Although interpretive data can transform the image into a unified text description, it has the problem of being too flat. When there is a large amount of data, it will not be able to quickly present the overall focus. To address this challenge, the knowledge graph is used to organize and present the correlation between different images and features, and then explore the possible image diagnosis situations. The research first collects the imaging diagnosis results of white light bronchoscopy, autofluorescent bronchoscopy, and bronchoscopy elastic ultrasonic examination by the computer-aided diagnosis system and the interpretation rules used by doctors to diagnose lung cancer images. This allows investigation of the detaile tumor characteristics in the two different examinations. Reference to the lung cancer staging method is made to weigh the main points of the impact on lung cancer, and convert the detailed description into the metadata used in the study. As well as the collection of lung imaging photos used by the computer-aided diagnosis system, from September 2015 to April 2017, 70 patients in Shuanghe Hospital were screened for white light bronchoscopy and autofluorescent bronchoscopy. Data source consists of a total of 114 images of lungs and bronchoscopy ultrasound examinations performed on patients in Shuanghe Hospital from February to December 2019. Research proceudres include: extract the metadata and construct the knowledge graph, place the emphasis of the metadata in the knowledge graph, and construct the correlation between the information. At the same time, there are related descriptions between different levels and nodes, allowing users to know the relationship between the information, and then find the connection among them to generate a logical knowledge graph. After the knowledge graph is established, users can understand how doctors and artificial intelligence perform lung imaging diagnosis, and distinguish the similarities and differences between different images. In the future, the experimental field can be applied to medical images of other different parts of the human body for the production and analysis of knowledge graph. This research provides a new technological tool to assist medical imaging diagnosis. en_US dc.description.tableofcontents 目錄 i圖目錄 iii表目錄 v誌謝 vi摘要 viiAbstract ix第一章 緒論 11.1背景介紹 11.2研究動機與目的 5第二章 文獻探討 72.1詮釋資料 72.2醫學知識圖譜文獻 10第三章 材料與研究方法 173.1影像資料與分析 203.1.1白光支氣管鏡檢查(white-light bronchoscopy, WLB) 213.1.2自體螢光支氣管鏡檢查 (autofluorescence bronchoscopy, AFB) 223.1.3支氣管鏡彈性超音波(endobronchial ultrasound elastography) 233.2量化特徵 253.2.1白光支氣管鏡檢查(white-light bronchoscopy, WLB) 253.2.2自體螢光支氣管鏡檢查(Autofluorescence bronchoscopy, AFB) 263.2.3支氣管鏡彈性超音波(endobronchial ultrasound elastography) 273.3詮釋資料建立與知識圖譜建構 283.3.1建立詮釋資料方便描述與建立關鍵詞彙 28第四章 實驗結果與討論 334.1 Neo4j 334.2結果 364.3分析 39第五章 結論及未來方向 40參考文獻 42 zh_TW dc.format.extent 2910961 bytes - dc.format.mimetype application/pdf - dc.source.uri (資料來源) http://thesis.lib.nccu.edu.tw/record/#G0108155003 en_US dc.subject (關鍵詞) 知識圖譜 zh_TW dc.subject (關鍵詞) 詮釋資料 zh_TW dc.subject (關鍵詞) 肺癌 zh_TW dc.subject (關鍵詞) 醫學影像 zh_TW dc.subject (關鍵詞) 支氣管鏡檢查 zh_TW dc.subject (關鍵詞) Knowledge graph en_US dc.subject (關鍵詞) Metadata en_US dc.subject (關鍵詞) Lung cancer en_US dc.subject (關鍵詞) Medical image en_US dc.subject (關鍵詞) Bronchoscopy en_US dc.title (題名) 以知識圖譜解譯人工智慧肺癌影像診斷之脈絡 zh_TW dc.title (題名) Using knowledge graph to interpret the context of artificial intelligence lung cancer image diagnosis en_US dc.type (資料類型) thesis en_US dc.relation.reference (參考文獻) [1] World Health Organization. (2021). Cancer. Available: https://www.who.int/news-room/fact-sheets/detail/cancer[2] K. Everington. (2019). Taiwan has 15th highest lung cancer rate in world. Available: https://www.taiwannews.com.tw/en/news/3825780[3] J. M. Croswell, D. F. Ransohoff , and B. S. Kramer. Principles of Cancer Screening: Lessons from History and Study Design Issues, Semin. Oncol., vol. 37, p.202-215. doi: 10.1053/j.seminoncol.2010.05.006.[4] 衛生福利部國民健康署（2020）。106年癌症登記報告。取自 https://www.hpa.gov.tw/Pages/Detail.aspx?nodeid=269&pid=12235[5] T. Yorifuji and S. Kashima, "Air pollution: another cause of lung cancer," Lancet Oncol., vol. 14, p. 788-789. doi: 10.1016/S1470-2045(13)70302-4.[6] J. L. Mumford, et al., "Lung cancer and indoor air pollution in Xuan Wei, China," Science, vol. 235, p. 217-220. doi: 10.1126/science.3798109.[7] Fredrik Nyberg, et al., "Urban Air Pollution and Lung Cancer in Stockholm," Epidemiology vol. 11, No. 5, pp. 487-495. Available: https://www.jstor.org/stable/3703988?seq=1[8] P. Broderick, et al., "Deciphering the Impact of Common Genetic Variation on Lung Cancer Risk: A Genome-Wide Association Study," Cancer Res., vol. 69, p. 6633-6641. doi: 10.1158/0008-5472.CAN-09-0680.[9] T. S. Panchabhai, M. Ghobrial and A. C. Mehta, "History of Bronchoscopy: The Evolution of Interventional Pulmonology," Interventions in Pulmonary Medicine, p. 609-621. doi: 10.1007/978-3-319-58036-4_39.[10] S. Leong, T. Shaipanich, S. Lam and K. Yasufuku, "Diagnostic bronchoscopy--current and future perspectives," J. Thorac. Dis., vol. 5, p. S498-S510. doi: 10.3978/j.issn.2072-1439.2013.09.08.[11] H. Müller, N. Michoux, D. Bandon and A. Geissbuhler, "A review of content-based image retrieval systems in medical applications—clinical benefits and future directions," Int. J. Med. Inf., vol. 73, p. 1-23. doi: 10.1016/j.ijmedinf.2003.11.024.[12] S. Jha and E. J. Topol, "Adapting to Artificial Intelligence: Radiologists and Pathologists as Information Specialists," JAMA, vol. 316, p. 2353-2354. doi: 10.1001/jama.2016.17438.[13] E. El-Dahshan, H. M. Mohsen, K. Revett and A. M. Salem, "Computer-aided diagnosis of human brain tumor through MRI: A survey and a new algorithm," Expert Syst. Appl., vol. 41, p. 5526-5545. doi: 10.1016/j.eswa.2014.01.021.[14] F. Pesapane, M. Codari and F. Sardanelli, "Artificial intelligence in medical imaging: threat or opportunity? Radiologists again at the forefront of innovation in medicine," Eur. Radiol. Exp., vol. 2(1), p. 35. doi: 10.1186/s41747-018-0061-6.[15] G. Chartr, et al., "Deep Learning: A Primer for Radiologists," RadioGraphics, vol. 37, p. 2113-2131. doi: 10.1148/rg.2017170077.[16] The American College of Radiology. ACR BI-RADS Atlas® 5th Edition. Available: https://www.acr.org/Clinical-Resources/Reporting-and-Data-Systems/Bi-Rads[17] Visual Resources Association Data Standards Committee. (2002). VRA Core Categories, Version 3.0. Available: https://marciazeng.slis.kent.edu/metadatabasics/Portuguese/schemas/VRA3.htm[18] Phillips, John T, "Metadata - Information about electronic records," Sch. J., vol. 29, p. 52.[19] J. Greenberg, "Understanding Metadata and Metadata Schemes," Cat. Classif. Q., vol. 40, p. 17-36. doi: 10.1300/J104v40n03_02.[20] The International Federation of Library Associations and Institutions,., Digital Libraries: Metadata Resources. Available: https://www.ifla.org/node/9337[21] 余顯強，「以資訊處理觀點論Metadata之本質與意涵」，教育資料與圖書館學，卷 45(2)，頁 249-266。[22] Brad, E., "Metadata and Its Application." Library and Information Science Journals. Available: https://www.questia.com/library/journal/1G1-93085596/metadata-and-its-application[23] 陳昭珍、蕭伯瑜（2006）。全集層次詮釋資料之應用研究:以EAD為例（學位論文）。取自國立臺灣師範大學圖書館機構典藏。（系統編號GN0692150049）[24] 王德華、袁萍、田榮華、徐亮（2014），醫學影像檔案資料的現代化管理，中國病案，15，14-15。[25] 郭莉珠（2004）。技術研究現狀與發展方向，檔案學通訊，1，74-76。[26] A. Kumar, J. Kim, L. Wen, M. Fulham and D. Feng, "A graph-based approach for the retrieval of multi-modality medical images," Med. Image Anal., vol. 18(2), p. 330-342. doi: 10.1016/j.media.2013.11.003.[27] T. Yu, et al., "Knowledge graph for TCM health preservation: Design, construction, and applications, " Artif. Intell. Med., vol. 77, p. 48-52. doi: 10.1016/j.artmed.2017.04.001.[28] S. M. S. Hasan, D. Rivera, X.-C. Wu, E. B. Durbin, J. B. Christian and G. Tourassi, "Knowledge Graph-Enabled Cancer Data Analytics, " IEEE J. Biomed. Health Inform., vol. 24, p. 1952-1967. doi: 10.1109/JBHI.2020.2990797.[29] 張幟（2018）。Neo4j圖形資料庫權威指南：傲視大數據時代的先端利器。臺北市：深石數位。[30] P.-H. Feng, Y.-T. Lin and C.-M. Lo, "A machine learning texture model for classifying lung cancer subtypes using preliminary bronchoscopic findings," Med. Phys., vol. 45, p. 5509-5514. doi: 10.1002/mp.13241.[31] P. H. Feng, T. T. Chen, Y. T. Lin, S. Y. Chiang and C. M. Lo, "Classification of lung cancer subtypes based on autofluorescence bronchoscopic pattern recognition: A preliminary study," Comput. Methods Programs Biomed., vol. 163, p. 33-38. doi: 10.1016/j.cmpb.2018.05.016.[32] Y. Li, et al., "Comparison of the autofluorescence bronchoscope and the white light bronchoscope in airway examination," Chin. J. Cancer, vol. 29, p. 1018-22. doi: 10.5732/cjc.010.10298.[33] A. K. Banerjee. Detection of early lung cancer: Autofluorescence bronchoscopy and investigational modalities - UpToDate. Available: https://www.uptodate.com/contents/detection-of-early-lung-cancer-autofluorescence-bronchoscopy-and-investigational-modalities[34] F. J. F. Herth, R. Eberhardt, D. Anantham, D. Gompelmann, M. W. Zakaria and A. Ernst, "Narrow-Band Imaging Bronchoscopy Increases the Specificity of Bronchoscopic Early Lung Cancer Detection," J. Thorac. Oncol., vol. 4, p. 1060-1065. doi: 10.1097/JTO.0b013e3181b24100.[35] A. Y. Altonbary, H. Hakim and A. M. El-Shamy, "Diagnostic Efficacy of Endoscopic Ultrasound Elastography in Differentiating Solid Pancreatic Lesions: A Single-Center Experience," Clin. Endosc., vol. 52, p. 360-364. doi: 10.5946/ce.2018.160.[36] S. Mittal, A. Mohan, V. Hadda and K. Madan, "Endobronchial ultrasound elastography in mediastinal lymphadenopathy: Report of two cases and systematic review of literature," Lung India, vol. 36, p. 149-153. doi: 10.4103/lungindia.lungindia_349_17.[37] H.-Y. He, M. Huang, J. Zhu, H. Ma and X.-D. Lyu, "Endobronchial Ultrasound Elastography for Diagnosing Mediastinal and Hilar Lymph Nodes," Chin. Med. J. (Engl.), vol. 128, p. 2720-2725. doi: 10.4103/0366-6999.167296.[38] Nakajima, T., et al., "Elastography for Predicting and Localizing Nodal Metastases during Endobronchial Ultrasound," Respiration, vol. 90, pp. 499-506. doi: 10.1159/000441798.[39] G. A. Silvestri, et al., "Methods for Staging Non-small Cell Lung Cancer: Diagnosis and Management of Lung Cancer, 3rd ed: American College of Chest Physicians Evidence-Based Clinical Practice Guidelines," Chest, vol. 143. doi: 10.1378/chest.12-2355.[40] 三軍總醫院（2021）。肺癌的診斷與處置。取自https://wwwv.tsgh.ndmctsgh.edu.tw/unit/10041/14484[41] 連啟惇（2012）。早期肺癌檢查新利器--窄頻染色及自體螢光支氣管鏡。取自 http://www.kmuh.org.tw/www/kmcj/data/10101/18.htm zh_TW dc.identifier.doi (DOI) 10.6814/NCCU202101399 en_US
