-
Notifications
You must be signed in to change notification settings - Fork 5
/
main.tex
149 lines (103 loc) · 7.15 KB
/
main.tex
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
\documentclass[11pt]{report}
% packages
\usepackage{baththesis}
\usepackage{amssymb} %for Blackboard bold etc
\usepackage{amsmath}
\usepackage{amsfonts}
\usepackage{graphicx} %for including eps graphics
\usepackage{float}
\usepackage[english]{babel}
\usepackage{placeins}
\usepackage{doi}
\usepackage{hyperref}
\usepackage{tikz}
\usepackage{wrapfig}
\usepackage{titlecaps}
\usepackage{sectsty}% http://ctan.org/pkg/sectsty
\usepackage{epstopdf}
%\usepackage[printwatermark]{xwatermark}
%\newsavebox\mybox
%\savebox\mybox{\tikz[color=red,opacity=0.3]\node{CONFIDENTIAL};}
%\newwatermark*[
% allpages,
% angle=45,
% scale=6,
% xpos=-20,
% ypos=15
%]{\usebox\mybox}
%
\usepackage{algpseudocode,algorithm,algorithmicx}
\algrenewcommand\algorithmicrequire{\textbf{Launch:}}
%%%
\algrenewcommand\algorithmicensure{\textbf{End Kernel}}
\makeatletter
\def\ALG@special@indent{%
\ifdim\ALG@thistlm=0pt\relax
\hskip-\leftmargin
\else
\hskip\ALG@thistlm
\fi
}
\newcommand{\Launch}[1]{\item[]\noindent\ALG@special@indent \textbf{Launch:}\ #1}
\newcommand{\EndKernel}{\item[]\noindent\ALG@special@indent \textbf{End Kernel}}
\makeatother
\newcommand{\tb}{{}}
\usepackage{subfigure}
\usepackage{indentfirst}
\usepackage{booktabs} % for much better looking tables
\usepackage{array} % for better arrays (eg matrices) in maths
\usepackage{paralist} % very flexible & customisable lists (eg. enumerate/itemize, etc.)
\usepackage{verbatim} % adds environment for commenting out blocks of text & for better verb
\usepackage{matlab-prettifier}
\renewcommand{\lstlistingname}{Code Snippet}
\usepackage{multicol}
\usepackage{caption}
\usepackage{makecell}
\newcommand{\argmin}{\mathop{\mathrm{arg\,min}}}
\newcommand{\argmax}{\mathop{\mathrm{arg\,max}}}
\makeatletter
\newcommand{\vast}{\bBigg@{4}}
\newcommand{\Vast}{\bBigg@{5}}
\makeatother
% front matter
\title{ Iterative Reconstruction and Motion compensation in Computed Tomography on GPUs}
\author{Ander Biguri}
\degree{Doctor of Philosophy}
\department{Department of Electrical and Electronic Engineering}
\degreemonthyear{November 2017}
\norestrictions
\begin{document}
\maketitle
%\begin{abstract}
%bla
%\end{abstract}
\tableofcontents
\listoffigures
\listoftables
\renewcommand{\abstractname}{Acknowledgements}
\begin{abstract}
First and foremost, enormous thanks to my supervisors: Manuchehr Soleimani for guiding me through the labyrinthic process a PhD is. Manjit Dosanjh for being such a strong leader and showing me the impact and seriousness of medical research. Steven Hancock for having that unstoppable will for doing science. Thousands thanks to the three of you, its been a honour meeting and working with you, .
I would like to thank everyone that has been involved somehow in the work presented in this thesis: Andr\'as De\'ak for the help in the maths of total variation methods and Manasavee Lohvithee for her further work in these same methods and all the help on implementing them. Thanks to everyone that has directly or indirectly contributed before and after release to the TIGRE Toolbox, either via conversations or by reports, for all the effort put onto this. Thanks specially to R.B., for all the improvements proposed in the CUDA part. Thanks to Nvidia for the donation of the GPU used in this work. I am also grateful to everyone that shared data for this thesis, The Christie Hospital, and Michelle Darrow and Imanol Luengo from Diamond Light Source. Sam Loescher and Reuben Lindroos: you did fantastic job with the python version of TIGRE, I hope you the best. And all you ``internet friends'' that have helped with all the complex programming issues that constantly appeared during the whole process, thanks to all of you, your help is as important any other.
Finally, I'd like to thanks to everyone that has pushed me and helped me to get to this moment. Myriad thanks to my family, especially my parents Joseba and Judith the never ending support. I wish everyone had the parents I have, you are the best. Thanks to all my friends, the single reason the though times disappear is because you have all been there to make them good! Special thanks to Dr (!) Imanol Luengo for being there to talk about science and convince me to do a PhD. Finally, infinite thanks to Irantzu Perez, for being there for me in all the good and bad times. This whole thing has been possible because of you, truly. There are hundreds of people that I would like to name to thank for all the professional and personal help provided, you are too many, and I have too little space, so I am stopping this now. Truly, thanks to everyone.
\end{abstract}
\renewcommand{\abstractname}{Abstract}
\begin{abstract}
Computed tomography (CT), and especially cone-beam computed tomography (CBCT) has a wide range of applications. This thesis focuses on CBCT for image-guided radiation therapy (IGRT), particularly for lung cancer treatment. In lung IGRT the tumour moves due to respiration, not only making it hard to target with the radiation beam, but also blurring the images acquired for daily treatment tuning. Generating high quality images without motion artefacts is essential for \textcolor{blue}{radiation} and hadron therapy. In this thesis, motion modelling ideas from CERN's phase space tomography are modified and adapted to lung CBCT. The CERN method includes a knowledge of the motion in the basic building blocks of the image reconstruction and uses all the acquired data to reconstruct a single static image at any chosen moment within the acquisition timespan. In order to use this method, and in general improve the reconstructed image quality of CBCT, iterative algorithms are explored with a focus on fast reconstruction using GPUs. The work presented here lead to the publication of the TIGRE Toolbox, a fast, easy-to-use MATLAB-CUDA toolbox for the reconstruction of CBCT images at state-of-the-art speeds with an extensive variety of iterative algorithms. This thesis presents the mathematics, GPU techniques and different applications of TIGRE and its algorithms, strengthening the idea already stated that iterative algorithms can significantly improve image quality in CBCT. A motion compensation method is developed together with a fast GPU implementation and its robustness is tested numerically by simulating the expected clinical errors in the data. The method is very robust and provides high-quality static images using data from disparate moments in time, offering the prospect of videos of patients breathing at no extra cost in radiation dose.
\end{abstract}
\chapterfont{\titlecap}
\sectionfont{\titlecap}
\subsectionfont{\titlecap}
\subsubsectionfont{\titlecap}
\Addlcwords{the on by in of and at from via an vs}
%\chapter{X-ray imaging in medicine}
\input{Introduction/intro.tex}
\input{StateOfArt/stateofart.tex}
\input{RecAlgorithms/RecAlgorithms.tex}\label{ch:rec}
\input{GPUmethods/GPUmethods_SH.tex}\label{ch:GPU}
\input{applications/Chapplications.tex}\label{ch:apllications}
\input{MotionCorrection/motioncorrection.tex}
\input{accuracyMC/accuracyMC.tex}\label{ch:motion analisis}
\input{conclusions/conclusions.tex}\label{ch:conclusions}
\bibliographystyle{abbrv}
\bibliography{RecAlgorithms,motioncorrection,intro,GPUmethods,motioncorrection2}
\end{document}