Skip navigation

DSpace

機構典藏 DSpace 系統致力於保存各式數位資料(如:文字、圖片、PDF)並使其易於取用。

點此認識 DSpace
DSpace logo
English
中文
  • 瀏覽論文
    • 校院系所
    • 出版年
    • 作者
    • 標題
    • 關鍵字
    • 指導教授
  • 搜尋 TDR
  • 授權 Q&A
    • 我的頁面
    • 接受 E-mail 通知
    • 編輯個人資料
  1. NTU Theses and Dissertations Repository
  2. 共同教育中心
  3. 智慧醫療與健康資訊碩士學位學程
請用此 Handle URI 來引用此文件: http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/94570
標題: 基於深度學習利用全玻片影像之卵巢透明细胞癌病理異質性反卷積研究
Deep Learning (DL)-Based Deconvolution of Histopathologic Heterogeneity in Ovarian Clear Cell Carcinoma (OCCC) Using Whole Slide Images (WSI)
作者: 陳智幸
Suphakorn Tanming
指導教授: 黃韻如
Ruby Yun-Ju Huang
共同指導教授: 林澤
Che Lin
關鍵字: Ovarian Clear Cell Carcinoma; deep learning; artificial intelligence; classification; histopathological pattern
Ovarian Clear Cell Carcinoma; deep learning; artificial intelligence; classification; histopathological pattern
出版年 : 2024
學位: 碩士
摘要: Ovarian cancer, despite having a lower incidence rate compared to other female cancers like breast cancer, remains one of the most lethal cancers for women globally, with around 314,000 new cases and 207,000 deaths each year. Often detected at the advanced stages due to the absence of early signs and symptoms, ovarian cancer carries an unfavorable prognosis and the highest mortality rate among women’s cancers. OCCC, a rare subtype, is particularly challenging, as it is frequently detected in the advanced stage, and has a high recurrent rate. Furthermore, significant geographical and racial disparities in incidence, add complexity to its diagnosis and management, with higher rates observed in East Asian countries
Histopathology, considered as a gold standard in medical diagnosis, involves the microscopic examination of tissue specimens. However, this process can be subjective, time-consuming, and susceptible to variability among different pathologists. Machine learning and deep learning applications have emerged as promising adjuncts to improve efficiency and accuracy, encompassing tissue segmentation, cell detection, feature extraction, and disease classification. Over the last decade, machine learning models have demonstrated their potential, particularly within digital pathology with regard to cancer diagnosis.
This study proposes the integration of computational digital histology to comprehensively assess tissue and cell morphologies to enhance a deeper understanding of the heterogeneity of disease. The utilization of deep convolution neural network (DCNN) techniques in 195 WSI of OCCC accurately achieved a high accuracy of 0.9885 for homogeneous pattern classification and an Area Under the Reciever Operating Characteristic Curve (AUROC) of 0.9995 on the internal test set. Additionally, a ResNet34 model trained on heterogeneous histotypes achieved an overall accuracy of 92.36%. These results highlight the potential that DCNN holds profound implications for the prognosis and guiding the selection of treatment options, ultimately elevating patient outcomes and advancing precision medicine in ovarian cancer care.
Ovarian cancer, despite having a lower incidence rate compared to other female cancers like breast cancer, remains one of the most lethal cancers for women globally, with around 314,000 new cases and 207,000 deaths each year. Often detected at the advanced stages due to the absence of early signs and symptoms, ovarian cancer carries an unfavorable prognosis and the highest mortality rate among women’s cancers. OCCC, a rare subtype, is particularly challenging, as it is frequently detected in the advanced stage, and has a high recurrent rate. Furthermore, significant geographical and racial disparities in incidence, add complexity to its diagnosis and management, with higher rates observed in East Asian countries
Histopathology, considered as a gold standard in medical diagnosis, involves the microscopic examination of tissue specimens. However, this process can be subjective, time-consuming, and susceptible to variability among different pathologists. Machine learning and deep learning applications have emerged as promising adjuncts to improve efficiency and accuracy, encompassing tissue segmentation, cell detection, feature extraction, and disease classification. Over the last decade, machine learning models have demonstrated their potential, particularly within digital pathology with regard to cancer diagnosis.
This study proposes the integration of computational digital histology to comprehensively assess tissue and cell morphologies to enhance a deeper understanding of the heterogeneity of disease. The utilization of deep convolution neural network (DCNN) techniques in 195 WSI of OCCC accurately achieved a high accuracy of 0.9885 for homogeneous pattern classification and an Area Under the Reciever Operating Characteristic Curve (AUROC) of 0.9995 on the internal test set. Additionally, a ResNet34 model trained on heterogeneous histotypes achieved an overall accuracy of 92.36%. These results highlight the potential that DCNN holds profound implications for the prognosis and guiding the selection of treatment options, ultimately elevating patient outcomes and advancing precision medicine in ovarian cancer care.
URI: http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/94570
DOI: 10.6342/NTU202403192
全文授權: 同意授權(限校園內公開)
電子全文公開日期: 2026-08-12
顯示於系所單位:智慧醫療與健康資訊碩士學位學程

文件中的檔案:
檔案 大小格式 
ntu-112-2.pdf
授權僅限NTU校內IP使用(校園外請利用VPN校外連線服務)
11.05 MBAdobe PDF
顯示文件完整紀錄


系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。

社群連結
聯絡資訊
10617臺北市大安區羅斯福路四段1號
No.1 Sec.4, Roosevelt Rd., Taipei, Taiwan, R.O.C. 106
Tel: (02)33662353
Email: ntuetds@ntu.edu.tw
意見箱
相關連結
館藏目錄
國內圖書館整合查詢 MetaCat
臺大學術典藏 NTU Scholars
臺大圖書館數位典藏館
本站聲明
© NTU Library All Rights Reserved