Application of Recurrent CNN on Low Contrast Calcium Fluorescence Image Sequence

カテゴリ: 部門大会

論文No: TC1-28

グループ名: 【C】平成30年電気学会電子・情報・システム部門大会プログラム

発行日: 2018/09/05

タイトル(英語): Application of Recurrent CNN on Low Contrast Calcium Fluorescence Image Sequence

著者名: Moiloa Pelonomi(東北大学),本間 経康(東北大学),小山内 実(東北大学)

著者名(英語): Pelonomi Moiloa|Noriyasu Homma|Makoto Osanai

キーワード: Deep learning|Segmentation|Calcium Fluorescence Imaging|Recurrent Convolutional Neurual Network

要約(日本語): Regions of interest (ROI) need to be determined in order to obtain valuable data from low contrast cellular fluorescence image sequences. This process is exceptionally time consuming when done manually. Previous work has shown that current semi and fully automated segmentation methods do not offer viable alternatives to the manual approach. In this study, a convolutional neural network (CNN) which has proven successful in alternate medical imaging segmentation applications is expanded upon. A recurrent neural network (RNN) architecture is incorporated into the traditionally spatially focused CNN architecture in the form of a Recurrent CNN in order to attempt to exploit both the temporal and spacial features of the low contrast calcium fluorescence image sequence segmentation problem.

PDFファイルサイズ: 853 Kバイト