2018 Fiscal Year Final Research Report
Automated diagnosis for colonoscopy using deep learning
Project/Area Number |
17K15971
|
Research Category |
Grant-in-Aid for Young Scientists (B)
|
Allocation Type | Multi-year Fund |
Research Field |
Gastroenterology
|
Research Institution | Showa University |
Principal Investigator |
|
Research Collaborator |
Mori Kensaku
Itoh Hayato
Oda Masahiro
|
Project Period (FY) |
2017-04-01 – 2019-03-31
|
Keywords | 人工知能 / 内視鏡 / 大腸癌 |
Outline of Final Research Achievements |
The aim of this research was to develop computer-aided detection(CADe) and computer-aided diagnosis(CADx) system for colonoscopy and evaluate its diagnostic performance.1.We developed CADx system based on ultra-magnifying endoscopy for differentiating colonic neoplasms and non-neoplasms. Support vector machine which is a traditional machine learning method, was applied for the CADx and achieved 97.4% accuracy. 2.The CADe system that works on conventional endoscopy, was developed using 3-dimensional convolution neural network. We prepared fully annotated 1.8 million frame of colonoscopy videos for machine learning.The CADe system achieved 90% sensitivity for colorectal lesion based on video based analysis.(Misawa M, et al. Gastroenterology 2018)
|
Free Research Field |
消化器内科学
|
Academic Significance and Societal Importance of the Research Achievements |
大腸内視鏡におけるリアルタイム病変検出・診断は、近年重要視され高精度化が要求されている。これは病変の見落としを防ぐことで、大腸癌罹患を予防し、加えて治療不要な非腫瘍性ポリープを確実に診断することで、かかる治療・病理検査を省略できるためである。本研究では大腸内視鏡における、病変の発見、病変の診断を人工知能で支援し、どのようなレベルの医師であっっても均てん化した医療が提供できる可能性を示した。これにより、本邦がん罹患数1位のがん種である大腸がんを抑制することが期待される。
|