TY - GEN
T1 - SE-U-Net
T2 - 13th Asian Conference on Intelligent Information and Database Systems, ACIIDS 2021
AU - Jiang, Lin Yi
AU - Kuo, Cheng Ju
AU - Tang-Hsuan, O.
AU - Hung, Min Hsiung
AU - Chen, Chao Chun
N1 - Funding Information:
This work was supported by Ministry of Science and Technology (MOST) of Taiwan under Grants MOST 109-2221-E-006-199, 108-2221-E-034-015-MY2, and 109-2218-E-006-007. This work was financially supported by the “Intelligent Manufacturing Research Center” (iMRC) in NCKU from The Featured Areas Research Center Program within the framework of the Higher Education Sprout Project by the Ministry of Education in Taiwan.
Publisher Copyright:
© 2021, Springer Nature Switzerland AG.
PY - 2021
Y1 - 2021
N2 - We proposed a context segmentation method for medical images via two deep networks, aiming at providing segmentation contexts and achieving better segmentation quality. The context in this work means the object labels for segmentation. The key idea of our proposed scheme is to develop mechanisms to elegantly transform object detection labels into the segmentation network structure, so that two deep networks can collaboratively operate with loosely-coupled manner. For achieving this, the scalable data transformation mechanisms between two deep networks need to be invented, including representation of contexts obtained from the first deep network and context importion to the second one. The experimental results reveal that the proposed scheme indeed performs superior segmentation quality.
AB - We proposed a context segmentation method for medical images via two deep networks, aiming at providing segmentation contexts and achieving better segmentation quality. The context in this work means the object labels for segmentation. The key idea of our proposed scheme is to develop mechanisms to elegantly transform object detection labels into the segmentation network structure, so that two deep networks can collaboratively operate with loosely-coupled manner. For achieving this, the scalable data transformation mechanisms between two deep networks need to be invented, including representation of contexts obtained from the first deep network and context importion to the second one. The experimental results reveal that the proposed scheme indeed performs superior segmentation quality.
UR - http://www.scopus.com/inward/record.url?scp=85104814793&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85104814793&partnerID=8YFLogxK
U2 - 10.1007/978-3-030-73280-6_54
DO - 10.1007/978-3-030-73280-6_54
M3 - Conference contribution
AN - SCOPUS:85104814793
SN - 9783030732790
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 678
EP - 691
BT - Intelligent Information and Database Systems - 13th Asian Conference, ACIIDS 2021, Proceedings
A2 - Nguyen, Ngoc Thanh
A2 - Chittayasothorn, Suphamit
A2 - Niyato, Dusit
A2 - Trawiński, Bogdan
PB - Springer Science and Business Media Deutschland GmbH
Y2 - 7 April 2021 through 10 April 2021
ER -