Annotation-efficient, patch-based, explainable deep learning using curriculum method for breast cancer detection in screening mammography
Increasing numbers of mammograms for radiologists to interpret pose a logistical challenge. We trained a DL model leveraging curriculum learning with mixed annotations for mammography. The DL model outperformed the baseline model with image-level annotations using only 20% of the strong labels. The study addresses the challenge of requiring extensive datasets and strong supervision for DL efficacy. The model demonstrated improved explainability through Grad-CAM, verified by a higher ground truth overlap ratio. He proposed approach also yielded robust performance on external testing data.
Authors
- Toygar Tanyel (ORCID: https://orcid.org/0000-0002-2421-6880)
- Mustafa Ege Şeker (ORCID: https://orcid.org/0000-0001-7664-5786)
- Deniz Alış (ORCID: https://orcid.org/0000-0002-7045-1793)
- Esma Çerekçi
- Ercan Karaarslan (ORCID: https://orcid.org/0000-0002-4581-4273)
- Nurper Denizoglu
- Emine Meltem (ORCID: https://orcid.org/0000-0003-3927-321X)
- İlkay Öksüz (ORCID: https://orcid.org/0000-0001-6478-0534)
- Ozden Camurdan
Institutions
- Acıbadem University (TR)
- Sağlık Bilimleri Üniversitesi (TR)
- Kent Hastanesi (TR)
- Şişli Etfal Eğitim ve Araştırma Hastanesi (TR)
- Acıbadem Adana Hospital (TR)
- Istanbul Technical University (TR)
Publication Details
- Journal
- Insights into Imaging
- Published
- 2025-03-19
- DOI
- https://doi.org/10.1186/s13244-025-01922-w
- Citations
- 4
- Primary Topic
- AI in cancer detection
- Type
- article
- Field-Weighted Citation Impact
- 9.92