Projects per year
Abstract
Robust cardiac image segmentation is still an open challenge due to the inability of the existing methods to achieve satisfactory performance on unseen data of dfferent domains. Since the acquisition and annotation of medical data are costly and time-consuming, recent work focuses on domain adaptation and generalization to bridge the gap between data from different populations and scanners. In this paper, we propose two data augmentation methods that focus on improving the domain adaptation and generalization abilities of state-to-the-art cardiac segmentation models. In particular, our \Resolution Augmentation"
method generates more diverse data by rescaling images to different resolutions within a range spanning different scanner protocols. Subsequently, our \Factor-based Augmentation" method generates more diverse data by projecting the original samples onto disentangled latent spaces, and combining the learned anatomy and modality factors from different domains. Our extensive experiments demonstrate the importance of effcient adaptation between seen and unseen domains, as well as model generalization ability, to robust cardiac image segmentation.
method generates more diverse data by rescaling images to different resolutions within a range spanning different scanner protocols. Subsequently, our \Factor-based Augmentation" method generates more diverse data by projecting the original samples onto disentangled latent spaces, and combining the learned anatomy and modality factors from different domains. Our extensive experiments demonstrate the importance of effcient adaptation between seen and unseen domains, as well as model generalization ability, to robust cardiac image segmentation.
Original language | English |
---|---|
Title of host publication | STACOM: International Workshop on Statistical Atlases and Computational Models of the Heart |
Subtitle of host publication | Statistical Atlases and Computational Models of the Heart. M&Ms and EMIDEC Challenges |
Publisher | Springer |
Pages | 187-195 |
ISBN (Electronic) | 978-3-030-68107-4 |
ISBN (Print) | 978-3-030-68106-7 |
DOIs | |
Publication status | E-pub ahead of print - 29 Jan 2021 |
Event | 23rd International Conference on Medical Image Computing and Computer Assisted Intervention - Lima, Peru Duration: 4 Oct 2020 → 8 Oct 2020 Conference number: 23 https://www.miccai2020.org/en/ |
Publication series
Name | Lecture Notes in Computer Science |
---|---|
Publisher | Springer |
Volume | 12592 |
ISSN (Print) | 0302-9743 |
ISSN (Electronic) | 1611-3349 |
Conference
Conference | 23rd International Conference on Medical Image Computing and Computer Assisted Intervention |
---|---|
Abbreviated title | MICCAI 2020 |
Country/Territory | Peru |
City | Lima |
Period | 4/10/20 → 8/10/20 |
Internet address |
Keywords / Materials (for Non-textual outputs)
- Cardiac image segmentation
- Data augmentation
- Disentangled factors mixing
- Domain adaptation
- Domain generalization
Fingerprint
Dive into the research topics of 'Disentangled Representations for Domain-generalized Cardiac Segmentation'. Together they form a unique fingerprint.Projects
- 1 Finished