Re-thinking Mammography Transfer Learning: The Dataset-Informed Transfer Learning (DITL) Framework for Breast Cancer Screening and Lesion Diagnosis
International Conference on Computer Vision and Image Processing / 2025 / conference
Bhandary Panambur A, Bayer S, Maier A
Scientific summary
DITL investigates how source-dataset selection influences mammography transfer learning for breast cancer screening and lesion diagnosis.
Abstract
TODO: Add the CVIP abstract when the proceedings page is public.
Why it matters
Dataset-informed model transfer can reduce brittle generalization in mammography tasks where target labels are expensive and domain shifts are common.
Contribution
The work frames transfer learning as a dataset-informed decision rather than a generic model reuse step.
Method overview
TODO: Add the exact DITL selection and evaluation protocol after public release.
Key findings
- In print according to the existing repository content.
- TODO: Add verified metrics after proceedings publication.
Citation
Bhandary Panambur A, Bayer S, Maier A. (2025). Re-thinking Mammography Transfer Learning: The Dataset-Informed Transfer Learning (DITL) Framework for Breast Cancer Screening and Lesion Diagnosis. International Conference on Computer Vision and Image Processing.
TODO: Add BibTeX.
