DIMADOSE

Radiotherapy is a cornerstone in the treatment of cancers. Image-guided and adaptive radiotherapy (IGART) is now considered as the optimal strategy to deliver a high conformal dose to the tumor. The recent introduction of MR-linac will potentially revolutionize IGART, by taking advantage of dynamic and high contrast MR images acquired all throughout the treatment. MR can be used to generate good quality synthetic CT images using deep learning methods. However, dose calculation on these images is not yet deployed, due to a lack of tools for validating the artificially created CT data. Indeed, patient anatomy variations and therefore treatment plan adjustments are performed using the initial CT, which provide the density information. Treatment planning would be more accurate if the density variations could be assessed daily and accounted for to calculate the cumulative dose. Such cumulative dose calculation is not implemented clinically, as the uncertainties involved are poorly assessed. Since very recently, MRI has been used to generate good quality synthetic X-ray scanner images (required for dose distribution calculations) using deep learning methods. However, dose calculation on these images is not yet deployed, due to a lack of tools to validate these artificially created data. Variations in patient anatomy, and therefore adjustments to the treatment plan, are made on the basis of initial tissue density data and CT images. Moreover, no cumulative dose calculations are carried out clinically, as the uncertainties affecting this calculation are poorly assessed. Our project aims to optimize the use of MR-linacs, which are currently far from being exploited to their full potential, to provide patients with the most personalized treatment possible. To achieve this strategy, three innovations are proposed: 1. Optimization of RT treatments by proposing original cumulative dose-guided ART tools based on digital twins. 2. Designing original methods for improving the quality assurance of this complex MRI-based RT strategy, including sCT generation step. 3. Simplify the standard MR-linac workflow by deleting the pretreatment planning step including CT scan and MRI acquisition as well as a pretreatment dose calculation. The main deliverables of the project will be the methodolgy used to simulate the digital twins, a prototype software integrating the digital twin, quality control tools of the MRI and sCT image quality and subsequent cumulated dose.