Beam's eye view to fluence maps 3D network for ultra fast VMAT radiotherapy planning

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.

To accelerate VMAT treatment planning by quickly predicting fluence maps from a 3D dose map. The predicted fluence maps can be quickly leaf sequenced because the network was trained to take into account the machine constraints.

We developed a 3D network which we trained in a supervised way using a combination of $L_1$ and $L_2$ losses, and radiation therapy (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 preprocess 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 and SSIM), as well as in the 3D dose domain using the dose-volume histogram (DVH) on a test set. The network inference, which does not include the data loading and processing, is less than 20 ms. 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.

We developed a novel deep learning approach for ultrafast VMAT planning by predicting all the fluence maps of a VMAT arc in one single network inference. The small difference of the DVH validate this approach for ultrafast VMAT planning.