A Beam's Eye View to Fluence Maps 3D Network for Ultra Fast VMAT Radiotherapy Planning

TL;DR

This work addresses the slow process of VMAT fluence-map generation by predicting all 180 fluence maps directly from BEV-transformed 3D dose maps using a 3D ConvNeXt-based network (3D MedNeXt). By expanding the training data with over 2000 Eclipse-generated plans and aligning gantry rotation with a BEV representation, the method achieves substantial gains in image-based metrics (PSNR/SSIM) and maintains closely matched DVHs, with inference times under 20 ms. The study demonstrates that dataset scale and a 3D, end-to-end approach significantly outperforms traditional 2D/3D U-Nets, providing a potential ultra-fast initialization or module for VMAT optimization. This approach lays the groundwork for faster, more scalable inverse planning across prostate VMAT and possibly other anatomies, with future work aiming at multi-arc and broader data integration.

Abstract

Volumetric Modulated Arc Therapy (VMAT) revolutionizes cancer treatment by precisely delivering radiation while sparing healthy tissues. Fluence maps generation, crucial in VMAT planning, traditionally involves complex and iterative, and thus time consuming processes. These fluence maps are subsequently leveraged for leaf-sequence. The deep-learning approach presented in this article aims to expedite this by directly predicting fluence maps from patient data. We developed a 3D network which we trained in a supervised way using a combination of L1 and L2 losses, and RT plans generated by Eclipse and from the REQUITE dataset, taking the RT dose map as input and the fluence maps computed from the corresponding RT plans as target. Our network predicts jointly the 180 fluence maps corresponding to the 180 control points (CP) of single arc VMAT plans. In order to help the network, we pre-process the input dose by computing the projections of the 3D dose map to the beam's eye view (BEV) of the 180 CPs, in the same coordinate system as the fluence maps. We generated over 2000 VMAT plans using Eclipse to scale up the dataset size. Additionally, we evaluated various network architectures and analyzed the impact of increasing the dataset size. We are measuring the performance in the 2D fluence maps domain using image metrics (PSNR, SSIM), as well as in the 3D dose domain using the dose-volume histogram (DVH) on a validation dataset. The network inference, which does not include the data loading and processing, is less than 20ms. Using our proposed 3D network architecture as well as increasing the dataset size using Eclipse improved the fluence map reconstruction performance by approximately 8 dB in PSNR compared to a U-Net architecture trained on the original REQUITE dataset. The resulting DVHs are very close to the one of the input target dose.

Figures (4)

• Figure 1: BEV transform of the 3D dose map.
• Figure 2: Our 3D network (3D MedNeXt) takes the BEV transform of the dose map as input, and output a 3D tensor where each slice is the fluence map prediction of one of the 180 control points.
• Figure 3: Example from the validation set, showing the fluence maps of the 180 control points of a VMAT plan predicted by our 3D MedNeXt network compared to the corresponding target fluence maps.
• Figure 4: DVHs of our 3D MedNeXt network. We show four examples from our validation dataset. The plain lines are the DVHs calculated from the target fluence maps and the dashed lines are the DVHs calculated from the fluence maps predicted by our network.