Utility of Same-Modality, Cross-Domain Transfer Learning for Malignant Bone Tumor Detection on Radiographs: A Multi-Faceted Performance Comparison with a Scratch-Trained Model
Transfer learning may not improve overall AUC but can enhance practical performance at clinically crucial thresholds. By maintaining high detection rates while reducing false positives, the transfer learning model offers superior clinical utility. Same-modality cross-domain transfer learning is an efficient strategy for developing robust AI systems for rare diseases, supporting tools more readily acceptable in real-world screening workflows.
Authors
- Toshifumi Ozaki (ORCID: https://orcid.org/0000-0003-1732-9307)
- Naoaki Kahara
- Shusa Ohshika
- Shinji Miwa (ORCID: https://orcid.org/0000-0002-5962-8287)
- Shuhei Osaki (ORCID: https://orcid.org/0000-0002-0498-5017)
- Toshiyuki Kunisada (ORCID: https://orcid.org/0000-0002-1775-850X)
- Hiroya Kondo (ORCID: https://orcid.org/0000-0002-4642-5448)
- Shunji Nishimura
- Yujiro Otsuka (ORCID: https://orcid.org/0000-0002-6677-0804)
- Tomohiro Fujiwara (ORCID: https://orcid.org/0000-0003-4257-9180)
- Joe Hasei (ORCID: https://orcid.org/0000-0002-4028-0786)
- Aki Yoshida (ORCID: https://orcid.org/0000-0002-4790-933X)
- Hironari Tamiya (ORCID: https://orcid.org/0000-0003-0455-8047)
- Koichi Takeuchi
- Yusuke Nakamura
- Ryuichi Nakahara
- K. Ikuta
Institutions
- Kanazawa University (JP)
- Okayama University of Science (JP)
- Hirosaki University (JP)
- Okayama University (JP)
- Juntendo University (JP)
- Mizushima Central Hospital (JP)
- Kindai University Hospital (JP)
- Osaka International Cancer Institute (JP)
- Nagoya University (JP)
Publication Details
- Journal
- Cancers
- Published
- 2025-09-27
- DOI
- https://doi.org/10.3390/cancers17193144
- Citations
- 1
- Primary Topic
- Sarcoma Diagnosis and Treatment
- Type
- article
- Field-Weighted Citation Impact
- 1.52