Cross‐shaped windows transformer with self‐supervised pretraining for clinically significant prostate cancer detection in bi‐parametric MRI
This study proposes CSwin UNet, a new transformer-based model for end-to-end detection of csPCa, enhanced by self-supervised pretraining to enhance network generalizability. We employ an automatic weighted loss (AWL) to unify pretext tasks, improving representation learning. Evaluated on two multi-institutional public datasets, our method surpasses existing methods in detection metrics and demonstrates good generalization to external data.
Authors
- Jacob Wynne
- Richard L. J. Qiu (ORCID: https://orcid.org/0000-0002-7877-1900)
- Shaoyan Pan (ORCID: https://orcid.org/0009-0007-1040-0189)
- Pretesh R. Patel
- Hui Mao (ORCID: https://orcid.org/0000-0002-0147-6022)
- Tian Liu (ORCID: https://orcid.org/0000-0003-1295-7339)
- Jing Wang (ORCID: https://orcid.org/0000-0002-8491-4146)
- Yuheng Li (ORCID: https://orcid.org/0009-0001-0249-8952)
- Xiaofeng Yang
- Justin Roper
- Ashesh B. Jani
Institutions
- Georgia Institute of Technology (US)
- Emory University (US)
- The Wallace H. Coulter Department of Biomedical Engineering (US)
- Icahn School of Medicine at Mount Sinai (US)
Publication Details
- Journal
- Medical Physics
- Published
- 2024-11-26
- DOI
- https://doi.org/10.1002/mp.17546
- Citations
- 10
- Primary Topic
- Advanced MRI Techniques and Applications
- Type
- article
- Field-Weighted Citation Impact
- 3.64
Funders
- National Institutes of Health