DreamOn: a data augmentation strategy to narrow the robustness gap between expert radiologists and deep learning classifiers
We show that REM-dream-inspired conditional GAN-based data augmentation is a promising approach to improving deep learning model robustness against noise perturbations in medical imaging. Additionally, we highlight a gap in robustness between deep learning models and human experts, emphasizing the imperative for ongoing developments in AI to match human diagnostic expertise.
Authors
- Alexander Pöllinger (ORCID: https://orcid.org/0009-0008-9783-5712)
- Lukas S. Huber (ORCID: https://orcid.org/0000-0002-7755-6926)
- Amith Kamath (ORCID: https://orcid.org/0000-0002-7243-8883)
- Aurélie Pahud de Mortanges (ORCID: https://orcid.org/0000-0002-5410-1080)
- Walter Senn (ORCID: https://orcid.org/0000-0003-3622-0497)
- Verena Carola Obmann (ORCID: https://orcid.org/0000-0001-6572-9968)
- Luc Lerch
- Mauricio Reyes
- Florian Dammann
Institutions
- University of Bern (CH)
- University Hospital of Bern (CH)
- University of Tübingen (DE)
Publication Details
- Journal
- Frontiers in Radiology
- Published
- 2024-12-19
- DOI
- https://doi.org/10.3389/fradi.2024.1420545
- Citations
- 3
- Primary Topic
- AI in cancer detection
- Type
- article
- Field-Weighted Citation Impact
- 1.04
Funders
- National Science Foundation
- Schweizerischer Nationalfonds zur Förderung der Wissenschaftlichen Forschung