On particle imaging with application to particle radiotherapy

Authors: Collins Fekete, Charles-Antoine
Advisor: Beaulieu, Luc; Seco, Joao
Abstract: The goal of this thesis is to develop methodology and knowledge in charged particle imaging for application in hadron radiotherapy. First, the various existing algorithm to estimate the path of a charged particle crossing a medium have been studied as a function of their efficiency and accuracy. To find an optimal solution for those two constraints, a phenomenological model has been developed that predict the most likely particle path in a medium. It was subsequently grounded in a solid physical background and extended to every ion up to carbon. Furthermore, prior-knowledge techniques were introduced to obtain the highest accuracy in the path estimate prediction for any ions. With these techniques in hand, we then approached the problem of tomographic reconstruction of charged particle radiographies. The first step of the work was to introduce the aforementioned path estimate method into a conventional charged particle reconstruction algorithm such as the algebraic reconstruction technique. This process requires a large calculation time that prevents an efficient reconstruction in a clinical work-flow, and suffer from convergence problems that leave the images with a high-noise level. Thus, it was decided to develop our own tomographic reconstruction algorithm in which the main difference resided in the optimization of individual projections. In our algorithm, the object was discretized into voxels and the average relative stopping power through voxel columns defined from the source to the detector pixels is optimized such that it maximizes the likelihood of the proton energy loss. The length spent by individual protons in each column is calculated through the path estimate. In this way, the spatial resolution of individual radiographies is optimized. The new radiographies can then be fed into a conventional X-ray tomographic algorithm, such as FDK, for a high resolution pCT reconstruction. The tomographic reconstruction requires a large number of projections and each can be individually long to acquire. This might cause problem into a clinical context where the beam time is costly and limited. There is a demand for efficiency in the procedure, which requires optimization of the algorithms. In this context, the next part of the thesis consisted on developing a method to utilize a subset of proton radiographies to retrieve stopping power parameters specific to the patient. This was done because a fewer number of radiographies can be acquired rapidly prior to the treatment. We studied the possibility of combining those subset of proton radiographies (usually either a single radiography or a pair) with single-energy X-ray tomographic images acquired prior for diagnostic. A new algorithm was develop to combine these two types of images and evaluated against various anthropomorphic phantoms that represents three body sites, the lung, the pelvis and the head. It has been shown that with a limited number of radiographies, it is possible to retrieve stopping power specific to the patient with an RMS error to the ground truth below 1%. The last part of the work was the experimental validation of the various algorithms developed. In collaboration with the (Deutsches Krebsforschungszentrum, Heidelberg), the HIT (Heavy Ion Therapy facility, Heidelberg) and the pCT collaboration (Loma Linda University, University of California San Francisco, University of California Santa Cruz, University Baylor), we designed an experiment to acquire charged particle tomographic images. To do so, we used the HIT’s synchrotron to produce a collimated beam of charged particle combined with the pCT detector to detect the particle before and after having crossed a pre-determined medium. This study allowed us to evaluate the noise, the spatial resolution and the precision achievable with charged particle imaging tomography.
Document Type: Thèse de doctorat
Issue Date: 2017
Open Access Date: 24 April 2018
Permalink: http://hdl.handle.net/20.500.11794/28130
Grantor: Université Laval
Collection:Thèses et mémoires

Files in this item:
SizeFormat 
33827.pdf22.31 MBAdobe PDFView/Open
All documents in CorpusUL are protected by Copyright Act of Canada.