Development of Deep Learning-based Nuclear Detection Algorithm for Mouse Embryo

2019 Fiscal Year Final Research Report

• Project/Area Number: 16H04731
• Research Category: Grant-in-Aid for Scientific Research (B)
• Allocation Type: Single-year Grants
• Section: 一般
• Research Field: System genome science
• Research Institution: Keio University
• Principal Investigator: Funahashi Akira 慶應義塾大学, 理工学部(矢上), 准教授 (70324548)
• Project Period (FY): 2016-04-01 – 2020-03-31
• Keywords: 画像解析 / 機械学習 / 深層学習 / 発生・分化

Outline of Final Research Achievements

Using deep learning, we developed an image processing algorithm to identify nuclei from the 4-dimensional fluorescence microscopic images of mouse embryo development. The existing analysis of cell dynamics in mouse embryogenesis has a problem that the accuracy of nuclear identification after 16 cell stages is very low for the 4D microscopy images. In this research, we developed a nuclear identification algorithm using deep learning, which has recently attracted attention as a powerful technique in image analysis. Our algorithm performed accurate nuclear identification up to the 50 cell stage to obtain quantitative criteria that can evaluate the quality of embryos.

Free Research Field

定量生物学

Academic Significance and Societal Importance of the Research Achievements

学術的意義として、当研究課題で開発したQCANetは、極体を除いた細胞核のセグメンテーションを行うことに成功し、初期マウス発生過程における胚ごとの違いを定量的に比較することが可能であることが示された点が挙げられる。
社会的意義としては、今後QCANetを用いて初期胚を定量的に評価するための最適な指標の探索を行うことで、経験的に定められた指標に代わる、産仔作出能との関連性が高い「胚の質を評価し得る指標」の確立が期待される点が挙げられる。