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

2019 Fiscal Year Final Research Report

Artificial intelligence pathological diagnosis by hyperspectral nonlinear Raman scattering imaging

Research Project

  • PDF
Project/Area Number 17H02793
Research Category

Grant-in-Aid for Scientific Research (B)

Allocation TypeSingle-year Grants
Section一般
Research Field Optical engineering, Photon science
Research InstitutionHokkaido University

Principal Investigator

Hashimoto Mamoru  北海道大学, 情報科学研究院, 教授 (70237949)

Co-Investigator(Kenkyū-buntansha) 高松 哲郎  京都府立医科大学, 医学(系)研究科(研究院), 教授 (40154900)
加藤 祐次  北海道大学, 情報科学研究院, 助教 (50261582)
三宅 淳  大阪大学, 国際医工情報センター, 特任教授 (70344174)
新岡 宏彦  大阪大学, データビリティフロンティア機構, 特任准教授(常勤) (70552074)
Project Period (FY) 2017-04-01 – 2020-03-31
Keywords深層学習 / 非線形ラマン散乱 / ハイパースペクトルイメージング
Outline of Final Research Achievements

By reducing the noise of observation data and improvement of observation time using deep learning, we succeeded in the imaging rate improvement of 1.6–1.2 image/min. to 12.5-4.0 image/min. In tissue classification by machine learning of nonlinear Raman scattering images, pre-training with fluorescence images significantly improved the segmentation ability. Besides, we have developed a new microscope for acquiring a large number of hyperspectral nonlinear Raman scattering images and succeeded in increasing the speed 14 times and reducing the excitation light peak irradiance 1/12 compared to the conventional method. It was shown that hyperspectral images of cultured cells were classified by deep learning, and that unsupervised learning could classify cells with different culture conditions.

Free Research Field

生体光計測

Academic Significance and Societal Importance of the Research Achievements

近年の病理医不足の解消には,一人の病理医が処理できる病理診断数の向上が課題であり,病理検査の効率化,自動化を行なうと共に,自動的に病理診断およびそのスクリーニングを行なうシステムの開発が重要である.従来にない新しいイメージング手法と人工知能の組み合わせにより,イメージング速度の向上や分別能力の向上が示された.医師の負担削減によりより多くの人々の健康な生活のサポートが可能となる.

URL: 

Published: 2021-02-19  

Information User Guide FAQ News Terms of Use Attribution of KAKENHI

Powered by NII kakenhi