RobMedNAS: searching robust neural network architectures for medical image synthesis
Investigating U-Net model robustness in medical image synthesis against adversarial perturbations, this study introduces RobMedNAS, a neural architecture search strategy for identifying resilient U-Net configurations. Through retrospective analysis of synthesized CT from MRI data, employing Dice coefficient and mean absolute error metrics across critical anatomical areas, the study evaluates traditional U-Net models and RobMedNAS-optimized models under adversarial attacks. Findings demonstrate RobMedNAS's efficacy in enhancing U-Net resilience without compromising on accuracy, proposing a novel pathway for robust medical image processing.
Authors
- Richard J. Bruce (ORCID: https://orcid.org/0000-0003-2004-8864)
- John W. Garrett (ORCID: https://orcid.org/0000-0002-8152-736X)
- Tanmayee Joshi
- Varun Jog (ORCID: https://orcid.org/0000-0003-4159-0900)
- Po‐Ling Loh (ORCID: https://orcid.org/0000-0002-6514-7834)
- Meltem Uyanik (ORCID: https://orcid.org/0000-0002-7567-0303)
- Weijie Chen (ORCID: https://orcid.org/0000-0002-1608-0899)
- Jinnian Zhang (ORCID: https://orcid.org/0000-0003-4352-7681)
- Alan B. McMillan (ORCID: https://orcid.org/0000-0003-4502-6522)
- Xiaomin Zhang (ORCID: https://orcid.org/0000-0002-5678-7637)
Institutions
- University of Wisconsin–Madison (US)
- University of Cambridge (GB)
- Madison Group (United States) (US)
- London Mathematical Laboratory (GB)
- Institute of Mathematical Statistics (US)
Publication Details
- Journal
- Biomedical Physics & Engineering Express
- Published
- 2024-08-13
- DOI
- https://doi.org/10.1088/2057-1976/ad6e87
- Primary Topic
- Advanced Neural Network Applications
- Type
- article
- Field-Weighted Citation Impact
- 0.00
Funders
- National Science Foundation
- UK Research and Innovation
- National Institutes of Health
- Engineering and Physical Sciences Research Council
- Science and Technology Facilities Council
- U.S. National Library of Medicine