-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathmain.tex
192 lines (159 loc) · 9.28 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
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
\documentclass[portrait, color=UCLburgundy, margin=1cm]{uclposter}
\linespread{1.0}
\input{commands.tex}
\usepackage{bm}
\usepackage{algorithm}
\usepackage{algorithmic}
\usepackage{caption}
\usepackage{blindtext}
\usepackage{siunitx}
\input{glossary}
\usepackage[style=ieee, maxbibnames=1, minbibnames=1, maxcitenames=1, mincitenames=1, backend=biber, defernumbers=false]{biblatex}
\addbibresource{./Biblio.bib}
\AtEveryBibitem{\clearfield{month}}
\AtEveryBibitem{\clearfield{day}}
\AtEveryBibitem{\clearfield{volume}}
\AtEveryBibitem{\clearfield{issue}}
\AtEveryBibitem{\clearfield{pages}}
\AtEveryBibitem{\clearfield{number}}
\AtEveryBibitem{\clearfield{title}}
\AtEveryBibitem{\clearfield{isbn}}
\AtEveryBibitem{\clearfield{keywords}}
\AtEveryBibitem{\clearfield{issn}}
\AtEveryBibitem{\clearfield{journal}}
\usepackage{fontspec}
\setmainfont[Ligatures=TeX]{LexendDeca-Regular.ttf}
\begin{document}
\title{PET/CT Motion Correction Exploiting Motion Models Fit \newline~on Coarsely Gated Data Applied to Finely Gated Data}
\author[12*]{Alexander~C.~Whitehead}
\author[3]{Kuan-Hao~Su}
\author[3]{Scott~D.~Wollenweber}
\author[2]{\newline~Jamie~R.~McClelland}
\author[12]{Kris~Thielemans}
\affil[1]{INM, UCL}
\affil[2]{CMIC, UCL}
\affil[3]{GE Healthcare}
\affil[*]{[email protected]}
\maketitle
\begin{multicols}{2}
\section*{Introduction}
\begin{highlightbox}[UCLlightgreen]
\begin{itemize}
\item \glss{MM} parameterise \glss{DVF} in terms of a \gls{SS}.
\item \glss{MM} can obtain \glss{DVF} for unseen data~\cite{McClelland2013}.
\item Previous work~\cite{Whitehead2021ComparisonMap} indicated that \glss{MM} and \acrshort{TOF} increase resolution.
\item This work:
\begin{itemize}
\item Incorporates \acrshort{MLACF}~\cite{Nuyts2012ML-reconstructionFactors}.
\item Incorporates a diffeomorphic velocity field parameterised registration.
\item Fits the \gls{MM} on coarsely gated data and applies it to finely gated data.
\item Uses more realistic simulation and count levels.
\item Differentiates itself by using two \glss{SS}, and group-wise registration.
\end{itemize}
\end{itemize}
\end{highlightbox}
\subsection*{\underline{\textbf{Evaluation}}}
\begin{itemize}
\item \gls{MC} was also applied to data in the same way, but using high noise high temporal/gate resolution, or noiseless data, for the \gls{MM} fitting.
\item Data also reconstructed without \gls{MC}, using either a sum of all \acrshort{Mu-Map} or the end inhalation \acrshort{Mu-Map}.
\item Volumes without \gls{MC} registered to the position of the \acrshort{Mu-Map}.
\item \glss{DVF} generated by each method were also applied to noiseless data for visual analysis.
\item Comparisons used included: A visual analysis, \acrshort{SSIM} to the ground truth~\cite{Wang2009MeanMeasures}, a profile over the lesion, \acrshort{SUV}\textsubscript{max} and \acrshort{SUV}\textsubscript{peak}.
\end{itemize}
\end{multicols}
\begin{figure}[H]
\centering
\includegraphics[width=1.0\linewidth]{visual_analysis.png}
\begin{highlightbox}[UCLlightblue]
\captionsetup{singlelinecheck=false, justification=centering}
\caption{First row reconstructions with \acrshort{AC} and \gls{MC}, second row noiseless data. Colour map ranges consistent for all images in each column.}
\end{highlightbox}
\label{fig:visual_analysis}
\end{figure}
\begin{multicols}{2}
\section*{Methods}
\subsection*{\underline{\textbf{\acrshort{XCAT} Volume Generation}}}
\begin{itemize}
\item \acrshort{XCAT} generated 480 volumes using a 240s respiratory trace.
\item \acrshort{FOV} including the base of the lungs with a 20mm diameter lesion.
\end{itemize}
\subsection*{\underline{\textbf{\acrshort{PET} Acquisition Simulation}}}
\begin{itemize}
\item Simulated corresponding to a \acrshort{GE} Discovery 710.
\item Pseudo-randoms and scatter were added.
\item Noise was simulated, such that data matched an acquisition over 240s. The count rate was selected to match that of research scans.
\item A respiratory \gls{SS} was generated using \acrshort{PCA}~\cite{Thielemans2011}.
\item Gated into 4 respiratory bins using the \gls{SS} and its gradient, each bin was a quadrant centred on the maximum or minimum of the displacement or gradient.
\end{itemize}
\subsection*{\underline{\textbf{\acrshort{MLACF} Image Reconstruction}}}
\begin{itemize}
\item \acrshort{MLACF} (7 full iterations, 24 subsets) for the activity update, and 9 iterations for the attenuation update~\cite{Nuyts2012ML-reconstructionFactors}.
\item Initialised using 1 iteration of \acrshort{MLEM}, with breath hold \acrshort{CT} for \acrshort{AC}.
\item Normalised between iterations and epsilon added.
\item Quadratic prior on emission image.
\end{itemize}
\subsection*{\underline{\textbf{Registration}}}
\begin{itemize}
\item Pre-processing including; replication of end-slices, smoothing, and\newline~standardisation.
\item Between each iteration, resampled volume was registered to the \acrshort{Mu-Map} and \glss{DVF} composed together.
\end{itemize}
\subsection*{\underline{\textbf{\acrlong{MM} Estimation}}}
\begin{itemize}
\item \gls{MM} fit using weighted \acrlong{LR} between registration \glss{DVF} and 2 \glss{SS}.
\item \gls{MM} fit between each iteration.
\end{itemize}
\subsection*{\underline{\textbf{Image Reconstruction with \acrshort{AC}}}}
\begin{itemize}
\item Re-gated into 30 respiratory bins using displacement gating, \gls{SS} and its gradient (10 amplitude and 3 gradient bins).
\item \acrshort{Mu-Map} determined using the inverse of the \glss{DVF} from the \gls{MM}.
\item Re-reconstructed with \acrshort{AC} using \acrshort{OSEM} (2 full iterations, 24 subsets).
\item Volumes post-filtered with a Gaussian smoothing, (\acrshort{FWHM} of 6.4mm in transverse plane and a normal Z-filter).
\end{itemize}
\begin{figure}[H]
\centering
\includegraphics[width=1.0\linewidth]{profile.png}
\begin{highlightbox}[UCLlightblue]
\captionsetup{singlelinecheck=false, justification=centering}
\caption{Profile across the lesion.}
\end{highlightbox}
\label{fig:profile}
\end{figure}
\begin{table}[H]
\centering
\begin{highlightbox}[UCLlightblue]
\captionsetup{singlelinecheck=false, justification=centering}
\caption{Comparison of \acrshort{SUV}\textsubscript{max} and \acrshort{SUV}\textsubscript{peak}.}
\end{highlightbox}
\vspace{1.0cm}
\resizebox*{1.0\linewidth}{!}
{
\begin{tabular}{||c|cc||}
\hline
\textbf{\acrshort{SUV}} & \textbf{Max} & \textbf{Peak} \\
\hline
\textbf{Ground Truth} & 8.76 & 7.96 \\
\hline
\textbf{4 Gate \gls{MM}} & 8.04 & 6.18 \\
\textbf{30 Gate \gls{MM}} & 1.77 & 1.32 \\
\hline
\textbf{4 Gate Noiseless \gls{MM}} & 8.05 & 6.24 \\
\textbf{30 Gate Noiseless \gls{MM}} & 7.96 & 5.99 \\
\hline
\textbf{Ungated, Static \acrshort{CT}} & 6.61 & 5.08 \\
\textbf{Ungated, \acrlong{AV-CCT}} & 5.65 & 4.44 \\
\hline
\end{tabular}
}
\label{tab:suv}
\end{table}
\section*{Conclusion}
\begin{highlightbox}[UCLlightgreen]
\begin{itemize}
\item A low number of gates for \gls{MM} fitting has minimal impact at low noise and improves \gls{MC} when there is a high level of noise in the gates
\item In the future, work will focus on evaluating the method on patient data.
\end{itemize}
\end{highlightbox}
\AtNextBibliography{\small}
\printbibliography
\end{multicols}
\end{document}