Robustness of Transformer-Based Fluence Map Prediction Under Clinically Realistic Perturbations
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2026
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866910184760147968 |
|---|---|
| author | Mgboh, Ujunwa Sultan, Rafi Ibn Kim, Joshua Thind, Kundan Zhu, Dongxiao |
| author_facet | Mgboh, Ujunwa Sultan, Rafi Ibn Kim, Joshua Thind, Kundan Zhu, Dongxiao |
| contents | Learning-based fluence map prediction offers a fast alternative to iterative inverse planning in intensity-modulated radiation therapy (IMRT), but its robustness under realistic distribution shifts remains unclear. We study a two-stage transformer pipeline that maps anatomy (CT and contours) to dose and then to beamlet fluence maps. We compare fluence-stage transformer backbones with hierarchical, global, and hybrid attention, trained with a physics-informed loss enforcing energy consistency. Robustness is evaluated under geometric perturbations, radiometric noise, reduced training data, and domain shifts using a prostate IMRT dataset, with additional evaluation of the dose stage on public datasets. Results show smooth degradation under moderate perturbations but sharp failures under severe rotations and noise. Hierarchical transformers (e.g., SwinUNETR) exhibit slower growth in upper-quartile energy error, indicating improved robustness. We further show that SSIM alone fails to capture clinically relevant errors, highlighting the need for physics-informed evaluation. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_00904 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Robustness of Transformer-Based Fluence Map Prediction Under Clinically Realistic Perturbations Mgboh, Ujunwa Sultan, Rafi Ibn Kim, Joshua Thind, Kundan Zhu, Dongxiao Computer Vision and Pattern Recognition Learning-based fluence map prediction offers a fast alternative to iterative inverse planning in intensity-modulated radiation therapy (IMRT), but its robustness under realistic distribution shifts remains unclear. We study a two-stage transformer pipeline that maps anatomy (CT and contours) to dose and then to beamlet fluence maps. We compare fluence-stage transformer backbones with hierarchical, global, and hybrid attention, trained with a physics-informed loss enforcing energy consistency. Robustness is evaluated under geometric perturbations, radiometric noise, reduced training data, and domain shifts using a prostate IMRT dataset, with additional evaluation of the dose stage on public datasets. Results show smooth degradation under moderate perturbations but sharp failures under severe rotations and noise. Hierarchical transformers (e.g., SwinUNETR) exhibit slower growth in upper-quartile energy error, indicating improved robustness. We further show that SSIM alone fails to capture clinically relevant errors, highlighting the need for physics-informed evaluation. |
| title | Robustness of Transformer-Based Fluence Map Prediction Under Clinically Realistic Perturbations |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2605.00904 |