• Search Research Projects
  • Search Researchers
  • How to Use
  1. Back to project page

2022 Fiscal Year Final Research Report

Domain adaptation using curriculum learning.for biomedical image analysis

Research Project

  • PDF
Project/Area Number 21K19829
Research Category

Grant-in-Aid for Challenging Research (Exploratory)

Allocation TypeMulti-year Fund
Review Section Medium-sized Section 62:Applied informatics and related fields
Research InstitutionKyushu University

Principal Investigator

Bise Ryoma  九州大学, システム情報科学研究院, 准教授 (00644270)

Project Period (FY) 2021-07-09 – 2023-03-31
Keywords機械学習 / ドメイン適応 / 深層学習 / 弱教師学習 / 半教師学習
Outline of Final Research Achievements

In this research project, we developed methods to address the domain shift problem in bio-medical image analysis. For example, we proposed a curriculum-based approach for learning cell shapes in cell detection tasks, gradually expanding the domain. This method aimed to overcome the challenge of models trained on a specific dataset (source domain) not performing well on datasets captured under different conditions (target domain). As a result, many paper were accepted in a top journal and conferences, e.g., MedIA (IF: 13.828), MICCAI (h5-index: 78), ISBI2023 (h5-index: 55), and WACV2022 (h5-index: 76).

Free Research Field

コンピュータビジョン

Academic Significance and Societal Importance of the Research Achievements

バイオ医療画像解析分野において,ある特定のデータセットに対して教師データを作成さえすれば,実環境における異なるドメインにおいて新たに教師データの作成をすることなく,対象物体の認識が可能となることは究極の課題の一つである.さらに言えば,既に存在する教師ありの公開データを元に,多様な実環境のデータでの学習が教師なしで可能となれば,実利用のハードルが格段に下がり,多くの医学及び生物学の研究で活用されることが期待される.本研究課題は,この課題に取り組むものであり,大いに意義がある.

URL: 

Published: 2024-01-30  

Information User Guide FAQ News Terms of Use Attribution of KAKENHI

Powered by NII kakenhi