FluenceFormer: Transformer-Driven Multi-Beam Fluence Map Regression for Radiotherapy Planning

Ujunwa Mgboh, Rafi Ibn Sultan, Joshua Kim, Kundan Thind, Dongxiao Zhu
Proceedings of The 9th International Conference on Medical Imaging with Deep Learning, PMLR 315:685-700, 2026.

Abstract

Fluence map prediction is central to automated radiotherapy planning but remains an ill-posed inverse problem due to the complex relationship between volumetric anatomy and beam-intensity modulation. Convolutional methods in prior work often struggle to capture long-range dependencies, which can lead to structurally inconsistent or physically unrealizable plans. We introduce FluenceFormer, a backbone-agnostic transformer framework for direct, geometry-aware fluence regression. The model uses a unified two-stage design: Stage 1 predicts a global dose prior from anatomical inputs, and Stage 2 conditions this prior on explicit beam geometry to regress physically calibrated fluence maps. Central to the approach is the Fluence-Aware Regression (FAR) loss, a physics-informed objective that integrates voxel-level fidelity, gradient smoothness, structural consistency, and beam-wise energy conservation. We evaluate the generality of the framework across multiple transformer backbones, including Swin UNETR, UNETR, nnFormer, and MedFormer, using a prostate IMRT dataset. FluenceFormer with Swin UNETR achieves the strongest performance among the evaluated models and improves over existing benchmark CNN and single-stage methods, reducing Energy Error to $\mathbf{4.5%}$ and yielding statistically significant gains in structural fidelity ($p < 0.05$).

Cite this Paper

BibTeX
@InProceedings{pmlr-v315-mgboh26a, title = {FluenceFormer: Transformer-Driven Multi-Beam Fluence Map Regression for Radiotherapy Planning}, author = {Mgboh, Ujunwa and Sultan, Rafi Ibn and Kim, Joshua and Thind, Kundan and Zhu, Dongxiao}, booktitle = {Proceedings of The 9th International Conference on Medical Imaging with Deep Learning}, pages = {685--700}, year = {2026}, editor = {Huo, Yuankai and Gao, Mingchen and Kuo, Chang-Fu and Jin, Yueming and Deng, Ruining}, volume = {315}, series = {Proceedings of Machine Learning Research}, month = {08--10 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v315/main/assets/mgboh26a/mgboh26a.pdf}, url = {https://proceedings.mlr.press/v315/mgboh26a.html}, abstract = {Fluence map prediction is central to automated radiotherapy planning but remains an ill-posed inverse problem due to the complex relationship between volumetric anatomy and beam-intensity modulation. Convolutional methods in prior work often struggle to capture long-range dependencies, which can lead to structurally inconsistent or physically unrealizable plans. We introduce FluenceFormer, a backbone-agnostic transformer framework for direct, geometry-aware fluence regression. The model uses a unified two-stage design: Stage 1 predicts a global dose prior from anatomical inputs, and Stage 2 conditions this prior on explicit beam geometry to regress physically calibrated fluence maps. Central to the approach is the Fluence-Aware Regression (FAR) loss, a physics-informed objective that integrates voxel-level fidelity, gradient smoothness, structural consistency, and beam-wise energy conservation. We evaluate the generality of the framework across multiple transformer backbones, including Swin UNETR, UNETR, nnFormer, and MedFormer, using a prostate IMRT dataset. FluenceFormer with Swin UNETR achieves the strongest performance among the evaluated models and improves over existing benchmark CNN and single-stage methods, reducing Energy Error to $\mathbf{4.5%}$ and yielding statistically significant gains in structural fidelity ($p < 0.05$).} }
Endnote
%0 Conference Paper %T FluenceFormer: Transformer-Driven Multi-Beam Fluence Map Regression for Radiotherapy Planning %A Ujunwa Mgboh %A Rafi Ibn Sultan %A Joshua Kim %A Kundan Thind %A Dongxiao Zhu %B Proceedings of The 9th International Conference on Medical Imaging with Deep Learning %C Proceedings of Machine Learning Research %D 2026 %E Yuankai Huo %E Mingchen Gao %E Chang-Fu Kuo %E Yueming Jin %E Ruining Deng %F pmlr-v315-mgboh26a %I PMLR %P 685--700 %U https://proceedings.mlr.press/v315/mgboh26a.html %V 315 %X Fluence map prediction is central to automated radiotherapy planning but remains an ill-posed inverse problem due to the complex relationship between volumetric anatomy and beam-intensity modulation. Convolutional methods in prior work often struggle to capture long-range dependencies, which can lead to structurally inconsistent or physically unrealizable plans. We introduce FluenceFormer, a backbone-agnostic transformer framework for direct, geometry-aware fluence regression. The model uses a unified two-stage design: Stage 1 predicts a global dose prior from anatomical inputs, and Stage 2 conditions this prior on explicit beam geometry to regress physically calibrated fluence maps. Central to the approach is the Fluence-Aware Regression (FAR) loss, a physics-informed objective that integrates voxel-level fidelity, gradient smoothness, structural consistency, and beam-wise energy conservation. We evaluate the generality of the framework across multiple transformer backbones, including Swin UNETR, UNETR, nnFormer, and MedFormer, using a prostate IMRT dataset. FluenceFormer with Swin UNETR achieves the strongest performance among the evaluated models and improves over existing benchmark CNN and single-stage methods, reducing Energy Error to $\mathbf{4.5%}$ and yielding statistically significant gains in structural fidelity ($p < 0.05$).
APA
Mgboh, U., Sultan, R.I., Kim, J., Thind, K. & Zhu, D.. (2026). FluenceFormer: Transformer-Driven Multi-Beam Fluence Map Regression for Radiotherapy Planning. Proceedings of The 9th International Conference on Medical Imaging with Deep Learning, in Proceedings of Machine Learning Research 315:685-700 Available from https://proceedings.mlr.press/v315/mgboh26a.html.

Related Material