Segmentation of Cell Membrane and Nucleus using Branches with Different Roles in Deep Neural Network

Tomokazu Murata, Kazuhiro Hotta, Ayako Imanishi, Michiyuki Matsuda, Kenta Terai

2018

Abstract

We propose a segmentation method of cell membrane and nucleus by integrating branches with different roles in a deep neural network. When we use the U-net for segmentation of cell membrane and nucleus, the accuracy is not sufficient. It may be difficult to classify multi-classes by only one network. Thus, we designed a deep network with multiple branches that have different roles. We give each branch a role which segments only cell membrane or nucleus or background, and probability map is generated at each branch. Finally, the generated probability maps by three branches are fed into the convolution layer to improve the accuracy. The final convolutional layer calculates the posterior probability by integrating the probability maps of three branches. Experimental results show that our method improved the segmentation accuracy in comparison with the U-net.

Paper Citation

in Harvard Style

Murata T., Hotta K., Imanishi A., Matsuda M. and Terai K. (2018). Segmentation of Cell Membrane and Nucleus using Branches with Different Roles in Deep Neural Network. In Proceedings of the 11th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2018) - Volume 4: BIOSIGNALS; ISBN 978-989-758-279-0, SciTePress, pages 256-261. DOI: 10.5220/0006717002560261

in Bibtex Style

@conference{biosignals18,
author={Tomokazu Murata and Kazuhiro Hotta and Ayako Imanishi and Michiyuki Matsuda and Kenta Terai},
title={Segmentation of Cell Membrane and Nucleus using Branches with Different Roles in Deep Neural Network},
booktitle={Proceedings of the 11th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2018) - Volume 4: BIOSIGNALS},
year={2018},
pages={256-261},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0006717002560261},
isbn={978-989-758-279-0},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 11th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2018) - Volume 4: BIOSIGNALS
TI - Segmentation of Cell Membrane and Nucleus using Branches with Different Roles in Deep Neural Network
SN - 978-989-758-279-0
AU - Murata T.
AU - Hotta K.
AU - Imanishi A.
AU - Matsuda M.
AU - Terai K.
PY - 2018
SP - 256
EP - 261
DO - 10.5220/0006717002560261
PB - SciTePress