學術產出-國科會研究計畫
文章檢視/開啟
書目匯出
-
題名 基於病理影像的人工智慧腫瘤突變及型態預測
Pathological Image-Based Artificial Intelligence Tumor Mutation and Morphological Prediction作者 羅崇銘;梁哲維;翁睿謙 貢獻者 圖檔所 關鍵詞 深度學習; 間質腫瘤; 試劑光譜
deep learning; mesenchymal neoplasia; reagent spectrum日期 2021-06 上傳時間 30-十月-2024 11:45:24 (UTC+8) 摘要 間質腫瘤為罕見疾病，但於特定部位卻常發生，例如腸胃道之間質腫瘤(GIST)。此類疾病若合併有高風險因子，則一般考慮以標靶治療為主。但腫瘤突變型式須先確立，方能進行有效之標靶治療，然需檢驗的基因標的物往往不只一個。目前的檢驗流程，從病理科收取檢體至獲得突變結果約需一週以上的時間。為加速檢驗流程的時效性，並降低病患等待結果的焦慮感，本研究案擬在檢驗流程中，以新世代定序方法及人工智慧來應用，期使能加速檢驗流程，縮短所需時間。在此過程中，運用多線反應為一關鍵步驟，而加大多線反應的可讀性，則是當前研究的主題。透過協同性的物理或化學技術(如試劑反應、螢光反應或光譜反應等)，可以得到比目前單一作用更加明顯極清楚的訊號。加強可判別的訊號，亦即代表著試驗敏感性的提升及所需檢體量的下降。並可依此建立自動化深度學習模型，若此研究案得以成功，則不僅可加速檢驗的流程，也可以拓展未來微量技術的應用，深化腫瘤檢驗醫學的進步。
Mesenchymal neoplasia is rare; however, at specific site it may be quitecommon, such as in gastrointestinal tract (e.q., gastrointestinal stromal tumor, GIST). When high-risk features are present, these diseases are mainly treated by targeted therapy. However, to achieve effective treatment, mutational analysis of the target genes must be applied first, and there are usually multiple gene targets need to be tested. Current turn-around time, from receiving the pathology specimen to confirming the mutational status of the target genes, takes about 1 week to accomplish. To accelerate the test time and to reduce the waiting anxiety of the patients, in this project, we plan to incorporate artificial intelligence andnew-generation sequencing techniques in the pathological processes, trying to reduce the time needed on laboratory procedures. Among these, one of the major topics is how to enhance the reaction signals. Through synergetic physical and chemical techniques (either via reagent, by fluorescence, or spectrum distribution), a stronger and clearer signal can be obtained, compared with the current signal produced by a single reaction. Enhance the readability of the reaction signals means the increase of test sensitivity and the reduction in specimen amount that needed. The corresponding automatic analysis model based on deep learning can be established. The success of this project not only will lead into a quicker way in tumor gene testing, but also provide an opportunity in advancing future nano-techniques, which in turn benefits the whole oncologic society.關聯 科技部, MOST109-2622-E004-001-CC3, 109.06-110.05 資料類型 report dc.contributor 圖檔所 dc.creator (作者) 羅崇銘;梁哲維;翁睿謙 dc.date (日期) 2021-06 dc.date.accessioned 30-十月-2024 11:45:24 (UTC+8) - dc.date.available 30-十月-2024 11:45:24 (UTC+8) - dc.date.issued (上傳時間) 30-十月-2024 11:45:24 (UTC+8) - dc.identifier.uri (URI) https://nccur.lib.nccu.edu.tw/handle/140.119/154168 - dc.description.abstract (摘要) 間質腫瘤為罕見疾病，但於特定部位卻常發生，例如腸胃道之間質腫瘤(GIST)。此類疾病若合併有高風險因子，則一般考慮以標靶治療為主。但腫瘤突變型式須先確立，方能進行有效之標靶治療，然需檢驗的基因標的物往往不只一個。目前的檢驗流程，從病理科收取檢體至獲得突變結果約需一週以上的時間。為加速檢驗流程的時效性，並降低病患等待結果的焦慮感，本研究案擬在檢驗流程中，以新世代定序方法及人工智慧來應用，期使能加速檢驗流程，縮短所需時間。在此過程中，運用多線反應為一關鍵步驟，而加大多線反應的可讀性，則是當前研究的主題。透過協同性的物理或化學技術(如試劑反應、螢光反應或光譜反應等)，可以得到比目前單一作用更加明顯極清楚的訊號。加強可判別的訊號，亦即代表著試驗敏感性的提升及所需檢體量的下降。並可依此建立自動化深度學習模型，若此研究案得以成功，則不僅可加速檢驗的流程，也可以拓展未來微量技術的應用，深化腫瘤檢驗醫學的進步。 dc.description.abstract (摘要) Mesenchymal neoplasia is rare; however, at specific site it may be quitecommon, such as in gastrointestinal tract (e.q., gastrointestinal stromal tumor, GIST). When high-risk features are present, these diseases are mainly treated by targeted therapy. However, to achieve effective treatment, mutational analysis of the target genes must be applied first, and there are usually multiple gene targets need to be tested. Current turn-around time, from receiving the pathology specimen to confirming the mutational status of the target genes, takes about 1 week to accomplish. To accelerate the test time and to reduce the waiting anxiety of the patients, in this project, we plan to incorporate artificial intelligence andnew-generation sequencing techniques in the pathological processes, trying to reduce the time needed on laboratory procedures. Among these, one of the major topics is how to enhance the reaction signals. Through synergetic physical and chemical techniques (either via reagent, by fluorescence, or spectrum distribution), a stronger and clearer signal can be obtained, compared with the current signal produced by a single reaction. Enhance the readability of the reaction signals means the increase of test sensitivity and the reduction in specimen amount that needed. The corresponding automatic analysis model based on deep learning can be established. The success of this project not only will lead into a quicker way in tumor gene testing, but also provide an opportunity in advancing future nano-techniques, which in turn benefits the whole oncologic society. dc.format.extent 116 bytes - dc.format.mimetype text/html - dc.relation (關聯) 科技部, MOST109-2622-E004-001-CC3, 109.06-110.05 dc.subject (關鍵詞) 深度學習; 間質腫瘤; 試劑光譜 dc.subject (關鍵詞) deep learning; mesenchymal neoplasia; reagent spectrum dc.title (題名) 基於病理影像的人工智慧腫瘤突變及型態預測 dc.title (題名) Pathological Image-Based Artificial Intelligence Tumor Mutation and Morphological Prediction dc.type (資料類型) report
