Task 3: Modeling of tumor and dose delivery process during treatment via numerical simulations
Last update: 27/09/2016
The post-doc fellow Eric Garrido has started working in April 2014 with Nicolas Boussion.
This task aimed at developping two complementary tools:
1. A numerical model of the state-of-the-art TrueBeam (VARIAN) linear accelerator for external radiotherapy, as it has recently been acquired by the University Hospital of Brest.
2. A numerical model of tumor at two different scales, the voxel level, and the population of cells level. This model aims at simulating a tumor through its cells’ population characteristics with parameters input from both morphological and functional imaging and radiobiology.
Once these two simulation tools will be operational, we will have the ability to compute highly realistic simulations of personalized external radiotherapy treatments by introducing real multimodal images of a patient within the simulation framework combining the LINAC model and the multiscale tumor model.
1- Modeling of the LINAC Varian Truebeam
Modeling of the LINAC has been carried out within the GATE (Geant4 Application for Tomography Emission) framework, including the moving multi-leaf collimators and on-board real-time portal imaging. The most important development was the modeling and simulation of virtual source (phasespace) for the “patient-independent” part of the LINAC model to significantly reduce both the required disk storage space and computation times associated with it. Both dose delivery and integrated real-time portal imaging can be now simulated.
A technical note on the development on the LINAC source modeling is being written and scheduled to be submitted to Phys Med Biol in the next few months (Eric Garrido, Julien Bert, Nicolas Boussion, Dimitris Visvikis. Source Modelling of Varian TrueBeam for 6X, 6FF, 10X and 10FFF beams).
2- Tumor cells model
In the context of tumor growth modeling, we have developed a new multi-scale approach in order to simulate cancer evolution during treatment by radiotherapy. In this model, input data are FDG PET images and basic biological information. More precisely, we created a new scale (referred to as mesoscopic) in order to design a bridge between the usual microscopic (individual cells) and macroscopic (tumor) scales. The mesoscopic point of view can be seen as a population-based approach in which cells that have similar characteristics or behaviors are grouped into populations. This is very straightforward especially when considering PET images where voxels may contain up to 1 million cells.
As an illustration, we can consider the cells inside a voxel that are at the same phase of the cellular cycle, G1 for example. Instead of considering thousands of such cells in the simulation process, we define a single population of cells at the G1 state. In a given voxel the full cellular cycle may be segmented into 20 populations using the approach, leading to 20 parameters only. As a consequence, the management of each individual cell in the voxel is no longer necessary. In the same time, the microscopic information corresponding to the cellular cycle is preserved.
As an application, we have simulated the effect of oxygen on tumor evolution during treatment by radiotherapy. The process was evaluated using three FDG PET images from a series of 17 patients with rectal cancer treated with radiotherapy. The first image was acquired just before treatment begins (at staging/planning) whereas the second and third images were acquired at day 8 and at the end of treatment respectively.
As a preliminary step, the first two images were used to tune our model. As a second step the model was used to simulate the last image which was compared to the actual patient image. Satisfactory correlations between simulated and real images were found for the whole set of patients. Full results will be presented in a paper that currently in preparation (Apeke Sena, Gaubert Laurent, Boussion Nicolas, Visvikis Dimitris, Rodin Vincent, Redou Pascal. Multi-Scale Modelling and Oxygen Impact Simulation on Temporal Evolution of a Tumor: Application on Rectum Cancer During Treatment by Radiotherapy).
The next step will be the actual combination of an anthropomorphic phantom or patients multimodal images with the tumor evolution model within the simulation framework of GATE and the LINAC model.