ESTRO 2024 - Abstract Book

S3620

Physics - Dose prediction, optimisation and applications of photon and electron planning

ESTRO 2024

Conclusion:

The PatchGAN model produced similar results compared to the Cascade U-Net, which may indicate that greater model adjustments may be needed to increase the influence of the discriminator during model training. Additionally, this may also indicate that GANs may not be suited for improving coarse details compared to other high performance models. Future work will continue to investigate GAN predictions with metrics that are unaffected by volume averaging, can quantify and compare the entropy of predictions, and evaluate the performance of GANs with ablated input.

Keywords: GAN, Dose Prediction, Head and Neck

References:

1. Liu S, Zhang J, Li T, Yan H, Liu J. Technical Note: A cascade 3D U-Net for dose prediction in radiotherapy. Med Phys. 2021;48(9):5574-5582. doi:10.1002/mp.15034

2. Radford A, Metz L, Chintala S. Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks. arXiv.org. Published November 19, 2015. Accessed October 24, 2023. https://arxiv.org/abs/1511.06434v2

2014

Digital Poster

Online adaptive VMAT radiotherapy of bone metastases based on a diagnostic CT and a standard linac

Rune S Thing 1,2 , Charlotte Kristiansen 1 , Christine V Madsen 1 , Martin Berg 1,2 , Lars U Fokdal 1,2

1 University Hospital of Southern Denmark, Department of Oncology, Vejle, Denmark. 2 University Hospital of Southern Denmark, Radiotherapy Research Team, Vejle, Denmark

Purpose/Objective:

The introduction of dedicated machines for online adaptive radiotherapy (oART) has renewed the focus on reducing treatment and waiting times in palliative radiotherapy (RT). Several studies have investigated the delivery of palliative RT without a dedicated planning CT scan (pCT), and perform target delineation and treatment planning directly on an existing diagnostic CT-scan (dCT). These studies have either used the dCT-based treatment plan after Cone Beam CT (CBCT) based image verification of the anatomy (typically in simple, opposing field treatment setups), or required online target and plan adaptation to be performed on a dedicated treatment machine (CBCT or MRI-based oART linac platform). In this abstract, we demonstrate the feasibility of performing CBCT-based oART using a conventional C-arm linac, with target delineation and VMAT pre-planning on a dCT image, followed by adaptation of the treatment plan to the anatomy on the pre-treatment CBCT. This serves as an example of the workflow in the RAPID-care phase II-protocol.

Material/Methods: