Please use this identifier to cite or link to this item:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/92393
Title: | 醫療大數據價值創新在胃癌防治之應用 Utilizing Medical Big Data for Value Creation on Gastric Cancer Prevention |
Authors: | 李宜家 Yi-Chia Lee |
Advisor: | 余峻瑜 Jiun-Yu Yu |
Keyword: | 胃癌,防治策略,醫療大數據,供應鏈,資料治理,人工智慧,遠距醫療, Gastric cancer,prevention strategy,medical big data,value chain,data governance,artificial intelligence,telemedicine, |
Publication Year : | 2023 |
Degree: | 碩士 |
Abstract: | 降低癌症死亡率為癌症防治的目標,癌症死亡率為發生率與存活率的綜合效應,要降低胃癌死亡率,需要由降低胃癌發生率及提高其存活率來著手,前者可針對減少風險因子及切除癌前病變,後者可藉由早期發現早期治療、以及晚期癌症的先進治療。本論文的第一部分乃闡述台灣的胃癌防治計畫,基於幽門螺旋桿菌感染為胃癌的主要風險因子,首先在高風險社區開始進行篩檢及治療幽門桿菌,後續逐漸擴展到一般民眾,並且發展數位化登錄系統、利用供應鏈思維以及品質管理指標以達到組織性篩檢的目標。然而,幽門螺旋桿菌的長期感染仍可能產生黏膜不可逆的破壞,清除幽門螺旋桿菌並無法百分之百預防胃癌,成功除菌之後,如何了解民眾的胃相,如何有效率標示高風險者定期接受胃鏡檢查,仍為臨床的痛點,過去預測胃癌風險需透過消化專科醫師進行胃鏡切片,並傳送檢體請病理科醫師判讀,需要大量人力與時間成本,也缺乏即時性,實務上需要有更有效率的做法。
論文的第二部分闡述了醫療大數據的興起與臨床應用,醫療與科技一直是臺灣兩大重點產業,除了兩方面各自的快速發展之外,醫療結合科技更是新興的趨勢,醫療為一個知識、技術、勞力高度密集的產業,在長期醫護人力短缺且過度勞累之情況下,如何將慣例收集的醫療大數據,使用數位科技技術運用在提升臨床照護的水平上,讓數據創造價值,為目前臨床醫學高度關注的重點,本論文以大型醫療體系的大數據資料庫出發,基於提升海量資料的可近性以及增進臨床研究的科學量能的使命,在科技發展、資料管理、以及流程設計三大創新面向,使用數位化與整合性的兩大策略,透過垂直整合申請流程、水平整合體系內單位、架構產學研合作平台的方法,達到創造數據價值的目標,也分析此產業的內外部環境,在封閉式創新或是開放式創新、科技政策或是商業模式之間的選擇上進行討論。 根據前兩部分的內涵,論文的第三部分則提出一個創新的實例,在胃癌防治的情境中,利用醫療大數據資料來開發深度學習模型,模擬內視鏡醫師檢查與病理科醫師判讀的過程,開發出能診斷胃相的人工智慧系統以解決前述的臨床痛點,協助第一線醫師標示高風險民眾,並實際落地使用胃癌防治系統之中。此技術包括兩個模組,第一個模組為深度學習模組,包括三部分,第一部分將胃鏡影像區分為胃部與非胃部分,第二部分將胃部影像分為幽門、體部和胃竇部,第三部分使用增強圖像對比後,評估胃癌前病變如萎縮性胃炎和腸上皮化生的嚴重程度以及幽門螺旋桿菌是否有感染;第二個模組為雲端運算,包括軟體即服務以及平台即服務的方法,我們將偏鄉醫院的胃鏡影像,傳輸至行動醫學影像存檔與通信系統中,將胃鏡影像傳輸至醫學中心人工智慧運算中心進行多階段深度學習模型運算,結果迅速傳輸回位於當地的行動裝置上,並顯示解釋熱度圖提供偏鄉第一線醫師做參考,提供一個端到端的服務,可以準確預測胃相,提供便捷的遠距醫療服務,也讓有限的醫療資源,能夠精確地運用,最後針對此系統進行價值主張與商業模式分析。 綜合三大部分,本論文在傳統胃癌防治的情境上,藉由數位化與數位轉型的世界趨勢,賦與新興技術的元素,讓初段預防與次段預防的模式能夠契合,實際使用醫療大數據的平台,建立出一個可以在社區實際應用的系統,不僅可以有效對民眾進行風險分級,也可以讓既有的數據資料創造出額外的價值。 Reducing the cancer mortality rate is the goal of cancer prevention and control. The cancer mortality rate is the combined effect of incidence and survival rates. To reduce the mortality rate of gastric cancer, efforts should be made to reduce the incidence of gastric cancer and improve its survival rate. The former can be achieved by targeting risk factors and removing precancerous lesions, while the latter can be accomplished through early detection and treatment as well as advanced treatments for late-stage cancer. The first part of this thesis elaborated on Taiwan's gastric cancer prevention and control program. Based on Helicobacter pylori infection as the primary risk factor for gastric cancer, screening and treatment for H. pylori were initiated in high-risk communities and gradually expanded to the general population. A digital registration system was developed, using the supply chain thinking and quality management indicators to achieve the goal of organized screening. However, eradicating H. pylori cannot entirely prevent gastric cancer. After successful eradication, understanding individuals' gastric mucosal conditions and arranging endoscopic surveillance for the high-risk individuals remain the clinical challenges. Long-term H. pylori infection can still result in irreversible mucosal damage. In the past, predicting gastric cancer risk required gastroenterologists to perform gastric endoscopy, take biopsies, and send the samples to pathologists for interpretation, which involved significant manpower and time costs and lacked real-time feedback. In practice, there is a need for more efficient approaches. The second part of the thesis discussed the rise of big data in healthcare and its clinical applications. Healthcare and technology have been two major focus industries in Taiwan. Besides their rapid individual development, the emerging trend is the integration of technology into healthcare. Healthcare is a knowledge-intensive, highly technical, and labor-intensive industry. In the face of long-term shortages of healthcare professionals and excessive workloads, utilizing healthcare big data collected conventionally with the help of technological models for clinical care, thus creating value from data, is a significant focus in clinical medicine. This part started with the background of a large healthcare system's extensive medical big data. With the mission of improving the accessibility of massive data and enhancing the scientific capability of clinical research, it focused on three major innovation aspects: technological development, data management, and process design. Using digitalization and integration as two major strategies, it aimed to create data value through vertical integration application processes, horizontal integration of internal units, and the establishment of an industry-academia-research collaboration platform. This part also analyzed the internal and external environment of this industry and discussed the choice between closed innovation and open innovation, as well as the selection between technology policies and business models. Based on the contents of the first two sections, the third part of the thesis introduced an innovative case. Within the context of gastric cancer prevention using healthcare big data, a deep learning model was developed, which simulated the process of endoscopists' examinations and pathologists' interpretations. The goal was to create an artificial intelligence system capable of diagnosing gastric conditions to address the aforementioned clinical challenges. This system assisted frontline doctors in identifying high-risk individuals and was integrated into the gastric cancer prevention system. This technology consisted of two modules. First, the deep learning module consisted of three artificial intelligence models. The first model distinguished gastric and non-gastric regions in endoscopic images. The second model further divided the gastric region into the antrum, corpus, and fundus. The third model assessed the severity of atrophic gastritis and intestinal metaplasia and the presence of H. pylori infecton after enhancing image contrast. The second module (the cloud computing module) utilized the software-as-a-service and platform-as-a-service approaches. Endoscopic images from rural hospitals were transmitted to a mobile Picture Archiving and Communication System. The upper endoscopic images were then sent to the artificial intelligence computing center at the medical center using the three-stage deep learning model. The results were swiftly transmitted back to local mobile devices, displaying heatmaps for physician reference in the rural hospital, thus providing an end-to-end service. An end-to-end service was developed using endoscopic images through deep learning, a mobile platform, and cloud computing. By accurately predicting the severity of precancerous gastric lesions, this system enabled convenient telemedicine services and ensured the precise allocation of limited healthcare resources. The value proposition analysis and business model analysis for this system were finally conducted. Collectively, the thesis incorporates elements of emerging technologies into traditional gastric cancer prevention, allowing the primary prevention and secondary prevention models to align. By establishing a system using the platform of medical big data that can be practically applied in the community, it not only enables effective risk stratification for the populations at risk but also creates additional value from the medical big data. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/92393 |
DOI: | 10.6342/NTU202304534 |
Fulltext Rights: | 未授權 |
Appears in Collections: | 商學組 |
Files in This Item:
File | Size | Format | |
---|---|---|---|
ntu-112-1.pdf Restricted Access | 9.89 MB | Adobe PDF |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.