Why does my medical AI look at pictures of birds? Exploring the efficacy of transfer learning across domain boundaries
We conclude that intra-domain pretraining generally outperforms cross-domain pretraining, but that very narrow domain definitions apply. Put simply, pretraining on CT images instead of natural images yields an advantage when fine-tuning on CT images, and only on CT images. We further conclude that ImageNet pretraining remains a strong baseline, as well as the best choice for pretraining if only insufficient data from the target domain is available. Finally, we publish our pretrained models and pretraining guidelines as a baseline for future research.
Authors
- Jan Egger (ORCID: https://orcid.org/0000-0002-5225-1982)
- René Hosch (ORCID: https://orcid.org/0000-0003-1760-2342)
- Frederic Jonske (ORCID: https://orcid.org/0000-0001-8622-6673)
- Johannes Haubold (ORCID: https://orcid.org/0000-0003-4843-5911)
- Enrico Nasca
- Michael Kamp (ORCID: https://orcid.org/0000-0001-6231-0694)
- Constantin Seibold (ORCID: https://orcid.org/0000-0001-6042-8437)
- Felix Nensa (ORCID: https://orcid.org/0000-0002-5811-7100)
- Janis Evers
- Moon Kim
Institutions
- RWI – Leibniz Institute for Economic Research (DE)
- TU Dortmund University (DE)
- Essen University Hospital (DE)
- University of Duisburg-Essen (DE)
Publication Details
- Journal
- Computer Methods and Programs in Biomedicine
- Published
- 2025-02-01
- DOI
- https://doi.org/10.1016/j.cmpb.2025.108634
- Citations
- 3
- Primary Topic
- Radiomics and Machine Learning in Medical Imaging
- Type
- article
- Field-Weighted Citation Impact
- 4.08