TY - GEN
T1 - Domain generalization via feature disentanglement with reconstruction for pathology image segmentation
AU - Lin, Yu Hsuan
AU - Tsai, Hung-Wen
AU - Shen, Meng-Ru
AU - Chung, Pau Choo
N1 - Funding Information:
ACKNOWLEDGMENT The authors thank the National Center for High-performance Computing (NCHC) of the National Applied Research Laboratories (NARLabs) of Taiwan for providing the computational and storage resources used in the present study. This work was supported in part by the National Science and Technology Council, Taiwan under Grant NSTC 111-2634-F-006-012.
Funding Information:
The authors thank the National Center for Highperformance Computing (NCHC) of the National Applied Research Laboratories (NARLabs) of Taiwan for providing the computational and storage resources used in the present study. This work was supported in part by the National Science and Technology Council, Taiwan under Grant NSTC 111-2634-F-006-012.
Publisher Copyright:
© 2023 IEEE.
PY - 2023
Y1 - 2023
N2 - In pathology, the learned model may suffer from performance degradation due to stain variations between the training and testing datasets. To address this challenge, this paper proposes a feature disentanglement approach to learn the domain-invariant features to achieve domain generalization. The adaptive instance normalization (AdaIN)-based reconstruction is introduced to preserve the important semantic information. The generalization ability of the proposed method is further improved by using a contrastive loss function based on color augmentation to attract the domain-invariant features and repel the domain-specific features in the feature disentanglement process. The proposed method is evaluated on liver tumor and liver lipid droplet segmentation tasks. The results demonstrate that the proposed method can be applied to unseen datasets scanned by different scanners without significant performance degradation.
AB - In pathology, the learned model may suffer from performance degradation due to stain variations between the training and testing datasets. To address this challenge, this paper proposes a feature disentanglement approach to learn the domain-invariant features to achieve domain generalization. The adaptive instance normalization (AdaIN)-based reconstruction is introduced to preserve the important semantic information. The generalization ability of the proposed method is further improved by using a contrastive loss function based on color augmentation to attract the domain-invariant features and repel the domain-specific features in the feature disentanglement process. The proposed method is evaluated on liver tumor and liver lipid droplet segmentation tasks. The results demonstrate that the proposed method can be applied to unseen datasets scanned by different scanners without significant performance degradation.
UR - http://www.scopus.com/inward/record.url?scp=85168676210&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85168676210&partnerID=8YFLogxK
U2 - 10.1109/CAI54212.2023.00073
DO - 10.1109/CAI54212.2023.00073
M3 - Conference contribution
AN - SCOPUS:85168676210
T3 - Proceedings - 2023 IEEE Conference on Artificial Intelligence, CAI 2023
SP - 152
EP - 153
BT - Proceedings - 2023 IEEE Conference on Artificial Intelligence, CAI 2023
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2023 IEEE Conference on Artificial Intelligence, CAI 2023
Y2 - 5 June 2023 through 6 June 2023
ER -