Geometrically focused training and evaluation of organs‐at‐risk segmentation via deep learning
We developed and evaluated a new geometric metric, wDSC, as a better indicator of D2cm<sup>3</sup> accuracy, which has the potential to become a surrogate for dosimetric accuracy in cervical brachytherapy. The model with DP loss showed non-statistically significant improvements in geometric and dosimetric performance. This work also holds the potential to be used for precise OARs delineation in adaptive radiotherapy.
Authors
- Benjamin Haibe‐Kains (ORCID: https://orcid.org/0000-0002-7684-0079)
- Jennifer Croke (ORCID: https://orcid.org/0000-0002-7090-6313)
- Kathy Han (ORCID: https://orcid.org/0000-0001-8439-3929)
- Alexandra Rink (ORCID: https://orcid.org/0000-0003-2790-669X)
- Ruiyan Ni (ORCID: https://orcid.org/0000-0002-2040-4155)
- Elizabeth Chuk (ORCID: https://orcid.org/0000-0001-7403-1633)
- Anthony Fyles (ORCID: https://orcid.org/0000-0002-3633-2483)
- Jelena Luković (ORCID: https://orcid.org/0000-0002-0232-6106)
- Michael Milosevic
Institutions
- University Health Network (CA)
- University of Toronto (CA)
- Princess Margaret Cancer Centre (CA)
- Vector Institute (CA)
Publication Details
- Journal
- Medical Physics
- Published
- 2025-04-25
- DOI
- https://doi.org/10.1002/mp.17840
- Citations
- 1
- Primary Topic
- Radiomics and Machine Learning in Medical Imaging
- Type
- article
- Field-Weighted Citation Impact
- 1.36