Deep-learning prediction of breast cancer hormone receptor status from CEM: a preliminary study
A ResNet-18, optimized for class imbalance through weighted training and with calibrated probabilities, predicted HR positivity on CEM with 91.9% accuracy and AUROC 0.81 in an independent test cohort using an F1-tuned threshold. Balanced accuracy (0.70) and MCC (0.60) demonstrate maintained discrimination despite an approximate 85% class imbalance (HR-positive cases). Patient-level splitting was employed to ensure robust evaluation. Limitations related to the dataset's scope and specific imaging protocols may influence broader generalizability.
Authors
- Alessandro Carriero (ORCID: https://orcid.org/0000-0001-7481-7169)
- Renzo Boldorini (ORCID: https://orcid.org/0000-0003-1183-2737)
- Anna Colarieti (ORCID: https://orcid.org/0000-0003-4133-2012)
- Tullio Meloni (ORCID: https://orcid.org/0000-0002-0604-8924)
- Anna Maria Rampi
- Marco Albera (ORCID: https://orcid.org/0009-0009-1054-2552)
Institutions
- Università degli Studi del Piemonte Orientale “Amedeo Avogadro” (IT)
- Azienda Ospedaliero Universitaria Maggiore della Carita (IT)
- Azienda Ospedaliera Ospedale Maggiore (IT)
- Piedmont University (US)
Publication Details
- Journal
- European Radiology Experimental
- Published
- 2025-12-02
- DOI
- https://doi.org/10.1186/s41747-025-00653-3
- Primary Topic
- Breast Cancer Treatment Studies
- Type
- article
- Field-Weighted Citation Impact
- 0.00