Debian Codegraph @

texmf-dist/doc/latex/xebaposter/poster.tex

raw
1 \documentclass[portrait]{xebaposter}
2
3 %\usepackage[vlined]{algorithm2e}
4 \usepackage{times}
5 \usepackage{calc}
6 \usepackage{url}
7 \usepackage{amsmath}
8 \usepackage{amssymb}
9 \usepackage{relsize}
10 \usepackage{multirow}
11 \usepackage{booktabs}
12 \usepackage{graphicx}
13 \usepackage{multicol}
14 %\usepackage[T1]{fontenc}
15 \usepackage{ae}
16 \usepackage{wrapfig}
17 \graphicspath{{images/}}
18
19 %\usepackage{geometry}
20 %\geometry{papersize={90cm,170cm},verbose=ture,reset}
21 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
22 %%%% Some math symbols used in the text
23 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
24 % Format
25 % \newcommand{\RotUP}[1]{\begin{sideways}#1\end{sideways}}
26
27
28 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
29 % Multicol Settings
30 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
31 % \setlength{\columnsep}{0.7em}
32 % \setlength{\columnseprule}{0mm}
33
34
35 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
36 % Save space in lists. Use this after the opening of the list
37 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
38 \newcommand{\compresslist}{%
39 \setlength{\itemsep}{1pt}%
40 \setlength{\parskip}{0pt}%
41 \setlength{\parsep}{0pt}%
42 }
43
44 %% Begin of Document
45 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
46 \begin{document}
47
48 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
49 %% Here starts the poster
50 %%---------------------------------------------------------------------------
51 %% Format it to your taste with the options
52 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
53 \definecolor{silver}{cmyk}{0,0,0,0.3}
54 \definecolor{yellow}{cmyk}{0,0,0.9,0.0}
55 \definecolor{reddishyellow}{cmyk}{0,0.22,1.0,0.0}
56 \definecolor{black}{cmyk}{0,0,0.0,1.0}
57 \definecolor{darkYellow}{cmyk}{0,0,1.0,0.5}
58 \definecolor{darkSilver}{cmyk}{0,0,0,0.1}
59
60 \definecolor{lightyellow}{cmyk}{0,0,0.3,0.0}
61 \definecolor{lighteryellow}{cmyk}{0,0,0.1,0.0}
62 \definecolor{lighteryellow}{cmyk}{0,0,0.1,0.0}
63 \definecolor{lightestyellow}{cmyk}{0,0,0.05,0.0}
64
65 \begin{poster}%
66 % Poster Options
67 {
68 eyecatcher=false,
69 % Color style
70 bgColorOne=darkSilver,
71 bgColorTwo=lighteryellow,
72 borderColor=reddishyellow,
73 headerColorOne=yellow,
74 headerColorTwo=reddishyellow,
75 headerFontColor=black,
76 boxColorOne=lightyellow,
77 boxColorTwo=darkSilver,
78 % Format of textbox
79 textborder=faded,
80 % Format of text header
81 headerborder=open,
82 headerheight=0.1\textheight,
83 headershape=rectangle,
84 headershade=shadetb,
85 % headerfont=\Large, %Sans Serif
86 boxshade=shadeLR,
87 % background=plain,
88 linewidth=2pt,
89 % grid=true,
90 }
91 % Eye Catcher
92 {
93 \includegraphics[width=0.08\linewidth]{logo}
94 % \includegraphics[width=0.08\linewidth]{track_frame_00450_06}
95 % \includegraphics[width=0.08\linewidth]{track_frame_04999_06}
96 }
97 % Title
98 {Parallel Nonlinear Analysis of Weighted Brain's Gray and White Matter Images for Alzheimer's Dementia Diagnosis}
99 % Authors
100 {\large Seiied-Mohammad-Javad~Razavian, Meysam Torabi, and Kio Kim \\%[1em]
101 {\normalsize\texttt{[email protected], [email protected], [email protected]}}}
102 % University logo
103 {
104 \begin{tabular}{r}
105 \includegraphics[height=0.07 \textheight]{shariflogo}\\
106 \end{tabular}
107 }
108
109 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
110 %%% Now define the boxes that make up the poster
111 %%%---------------------------------------------------------------------------
112 %%% Each box has a name and can be placed absolutely or relatively.
113 %%% The only inconvenience is that you can only specify a relative position
114 %%% towards an already declared box. So if you have a box attached to the
115 %%% bottom, one to the top and a third one which should be inbetween, you
116 %%% have to specify the top and bottom boxes before you specify the middle
117 %%% box.
118 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
119
120 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
121 % \headerbox{Abstract}{name=abstract,column=0,row=0,span=1}{
122 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
123 %We are proposing a novel approach using T1-weighted and T2-weighted Magnetic Resonance Images (MRI's) of brain. Since T1-weighted images and T2-weighted images have inherent physical differences, we extracted some specific features from each. Then the variations of the relevant eigenvalues of the extracted features were tracked to pick up the most informative ones. The final features were assigned to two parallel systems to be nonlinearly categorized. The dataset includes 60 T1-weighted images and 60 T2-weighted images of normal and abnormal cases.
124 % }
125
126 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
127 \headerbox{Introduction}{name=introduction,column=0,row=0}{
128 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
129 %AD disease destroys the nervous system and defects the temporal lobe and parietal lobe of the brain.
130
131 %\begin{wrapfigure}[10]{r}{.72\textwidth}
132 %\centering
133 %{\includegraphics[scale=.9]{Maghz}}
134 %\end{wrapfigure}
135 We introduce a novel method for Alzheimer's dementia diagnosis.
136 Our focus is on analysis of MRI because of its high resolution in brain tissues.% which includes normal and abnormal cases.
137 We believe that the tissue analysis can be a reliable tool to diagnose various diseases. The novelty of this work is using both T1 and T2 images simultaneously after they are optimally weighted.
138 }
139
140 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
141 \headerbox{Phase II: core processing}{name=phase2,column=1,span=1}{
142 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
143 We extracted features which can capture the properties of the brain tissue using Gray-Level Co-occurrence Matrix (GLCM) which helps scan the tissue in order to give a sense of how the relationship between different pixels of an image is.% \cite{BeDeTo},
144 %\textbf{Neighborhood definition and calculating GLC matrix}%
145
146 \begin{wrapfigure}{r}{.6\textwidth}
147 \centering
148 \vspace{-10pt}
149 {\includegraphics[scale=.5]{fig3}}
150 \vspace{-20pt}
151 %\caption{\scriptsize Right figure shows the definition of neighborhood for the central point in T1 images while the left figure shows the definition of the neighborhood for T2. As it can be obviously seen in the figure, the neighborhood defined for T1 covers more detail.}
152 \label{fig3}
153 \end{wrapfigure}
154
155 Considering the fact that AD defects the white and gray regions of brain more than its black and marginal regions, and also since T1-weighted has more medical data of white and gray regions than T2-weighted images, we believe that T1 images have more data than T2 images, thus we define neighborhood of T1 and T2 differently.
156
157 %\textbf{Extraction of statistical features from GLC matrices}
158 The desired Statistical features are as follows:
159 1) Energy,
160 2) Contrast,
161 3) Homogeneity, and
162 4) Correlation
163
164 %To calculate correlation, two following parameters initially are calculated:
165 %in which elements of GLC matrix are shown with the symbol C(i, j) and the number of gray-levels with N.
166
167 Correlation in the GLC matrix:
168 \begin{align*}
169 f_1 &= \frac{HXY-HXY1}{max\{HX,HY\}}\\
170 f_2 &= \sqrt{1-\exp\{-2(HXY2-HXY)\}}
171 \end{align*}
172 %in which, $HX$ and $HY$ are entropy of $C_x$ and $C_y$ respectively, and $HXY$, $HXY1$ and $HXY2$ are defined as follows, respectively:
173 {\scriptsize
174 \centerline{\begin{minipage}{.95\textwidth}
175 C(i, j) is an element of GLC matrix and N is number of gray-levels.
176 \end{minipage}}
177 \vspace{-.5mm}
178 $$\begin{array}{l}
179 C_x(i) =\sum_{j=1}^N C(i,j), \quad C_y(i)=\sum_{i=1}^N C(i,j)\\
180 HXY = -\sum_{i=1}^N\sum_{j=1}^N C(i,j) \log\{C(i,j)\}\\
181 HXY1 = -\sum_{i=1}^N\sum_{j=1}^N C(i,j) \log\{C_x(i)\times C_y(j)\}\\
182 HXY2 = -\sum_{i=1}^N\sum_{j=1}^N C_x(i)\times C_y(j) \log\{C_x(i)\times C_y(j)\}\\
183 \end{array}$$
184 }
185
186 4*16 features from T1 and 4*8 features from T2 are extracted.
187 }
188
189 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
190 % \headerbox{Phase III: Feature Analysis}{name=phase3,column=2,span=1,row=0}{
191 \headerbox{Feature Extraction}{name=phase3,column=2,span=1,row=0}{
192 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
193 We reduce two vectors, one for T1 images and another one for T2 images that using PCA. % for extracting the main components.
194 %Our simulation results are demonstrated in Figure~\ref{fig4}, showing the last 20 eigenvalues of both T1 images and T2 images. Other eigenvalues are not displayed, since, as the figure implies, they are almost zero. Thus,
195 We chose 19 first eigenvalues with the largest values for T1 image, and also 17 eigenvalues with the largest values for T2 images.
196
197 %\centering
198 \centerline{\includegraphics[scale=1.1,height=7cm]{fig4}}
199 %\caption{Trends of changes in eigenvalues of two input vectors.}
200 %\label{fig4}
201 %\end{figure}
202 }
203
204 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
205 % \headerbox{Phase IV: Discrimination }{name=phase4,column=2,span=1,below=phase3}{
206 \headerbox{Discrimination }{name=phase4,column=2,span=1,below=phase3}{
207 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
208 Two feed-forward neural networks are tuned and used with 19 and 17 inputs for analysis of T1 and T2 images respectively. Both networks have a hidden layer with 10 nodes equipped with Sigmoid transfer function.
209
210 We multiply the first neural network's output by weighting factor 0.63, and also multiply the second neural network's output by weighting factor 0.37, that were initialized and determined during the training of the classifier system.
211
212
213 }
214
215 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
216 \headerbox{Results }{name=results,column=1,span=2,below=phase2}{
217 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
218 \begin{multicols}{2}
219 In order to give a more clear measure of the performance of the classifier system, consisting of two parallel neural networks, the right-top figure illustrates an ROC graph.
220
221 However, higher accuracies in separation of training data can cause a better result in our database, but this increases the number of layers and nodes of the neural networks which may lead to overtraining of the system and thus, it may reduce the classification accuracy of test images. Therefore, we believe that the response shown here is an optimal one.
222
223 %We aware of similar works of other researchers (e.g. \cite{WaRoCoSpBaGaDoChSe,FrFo}) in which they obtained better results, but We try to increase the speed of process to have a real-time classifier system.
224
225 In the future research, we try to, in addition
226 to improving the diagnosis of AD, produce initially independent features in
227 order to increase the speed of process to have a real-time
228 classifier system.
229
230 \centerline{\includegraphics[scale=.7]{fig6}}
231 %\caption{An ROC graph corresponding to classifier's responses shows the accuracy of the proposed AD diagnosis system.}
232 \centerline{\includegraphics[scale=.7]{fig5}}
233 %\caption{The final result of the AD diagnosis system.}
234
235 \end{multicols}
236 }
237
238
239 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
240 % \headerbox{Related Works}{name=relatedworks,column=0,below=introduction}{
241 % %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
242 %A review of the works done during the last 20 years :
243 %\begin{itemize}
244 %\compresslist
245 % \item brain atrophy %\cite{WaRoCoSpBaGaDoChSe}
246 % \item number and size of senile plaques in the brain %\cite{HaHoDe}
247 % \item deformation of the brain structure %\cite{MaPe}
248 % \item bio-signal processing %\cite{HaPeGaMo}.
249 % \item medical image processing and analysis
250 %\begin{itemize}
251 % \compresslist
252 % \item microscopic methods
253 % \item macroscopic methods
254 % \end{itemize}
255 %\end{itemize}
256 %{\scriptsize (see paper for references.)}
257 %}
258
259 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
260 \headerbox{Brain Images Dataset}{name=imagedataset,column=0,span=1,below=introduction}{
261 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
262 \begin{itemize}
263 \compresslist
264 \item 120 MRI's (including T1 and T2 images) for different cross-sections of the brain
265 \item 52 images belong to the normal control group while the rest belongs to AD patients.
266 \item Half of the images is T1 and the other half is T2.
267 \item 60 percent of the images is used for training, and the rest is reserved to test and control the algorithm.
268 \item Images dataset: The Whole Brain Atlas. Available on \url{http://www.med.harvard.edu/AANLIB/home.html}
269 \end{itemize}
270 }
271
272 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
273 \headerbox{Disease Diagnosis Method}{name=ddm,column=0,span=1,below=imagedataset}{
274 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
275 \centerline{\includegraphics[scale=.7,width=.9\textwidth,height=6cm]{fig1}}
276 }
277
278 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
279 \headerbox{Phase I: preprocessing}{name=phase1,column=0,span=1,below=ddm}{
280 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
281 Removing irrelevant information:% that does not help the algorithm detect AD:
282 \begin{enumerate}
283 \compresslist
284 \item marginal parts of the images.
285 \item background.
286 \item other parts of brain, e.g. parts of eyes.
287 \end{enumerate}
288 %\subsubsection*{Registration}
289 We registered each image on its corresponding image as precisely as possible using 10 landmarks. The landmarks are located and placed on those points of the brain which are geometrically and anatomically more important. %\cite{YuXiBeLiZhGuDeJuWo}.
290 %\cite{ToMoVaRaDeRaTaFa}.
291 %(Figure~\ref{fig2}).
292 Registration includes %two stages of interpolation:
293 1) pixel position interpolation and 2) gray-level interpolation. For performing interpolation process, radial basis function $r^2 \log r^2$ is used. %\cite{ToMoVaRaDeRaTaFa}.
294 The affine terms have not been used in the proposed algorithm, since the results were already satisfactory.
295
296 \centerline{\includegraphics[scale=.34]{fig2}}
297 \centerline{\begin{minipage}{.8\textwidth}
298 \scriptsize Two samples of cross-sections of brain which are marked with 10 landmarks.
299 \end{minipage}}
300 %{\scriptsize Two samples of cross-sections of brain which are marked with 10 landmarks.}
301 %\label{fig2}
302
303 }
304
305 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
306 % \headerbox{Funding}{name=funding,column=1,span=2,above=bottom}{
307 % %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
308 % This work is entirely done at Sharif University of Technology, Tehran, Iran.
309 % }
310
311 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
312 \headerbox{References}{name=references,column=1,span=2,below=results}{
313 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
314 \smaller
315
316 \bibliographystyle{ieee}
317 \renewcommand{\section}[2]{\vskip 0.05em}
318 \begin{thebibliography}{1}\itemsep=-0.01em
319 \setlength{\baselineskip}{0.4em}
320
321 % \bibitem{amberg07:nicp}
322 % B.~Amberg, A.~Blake, T.~Vetter
323 % \newblock On Compositional Image Alignment with an Application to Activce Appearance Models
324 % \newblock In {\em CVPR'09}, 2009.
325 \bibitem{}
326 M. Torabi, H. Moradzadeh, R. Vaziri, S. M. J. Razavian, R. Dehestani
327 Ardekani, M. Rahmandoost, A. Taalimi, and E. Fatemizadeh, ``Development
328 of Alzheimer's Disease Recognition Using Semiautomatic
329 Analysis of Statistical Parameters Based on Frequency Characteristics
330 of Medical Images,'' in \emph{Proc. of IEEE Int. Conf. Signal Proc. \&
331 Communication}, Nov 2007, pp. 868--871.
332 \bibitem{}
333 M. Torabi, R. Dehestani Ardekani, and E. Fatemizadeh, ``Discriminationnation
334 between Alzheimer's Disease and Control Group in MR-Images Based
335 on Texture Analysis Using Artificial Neural Network,''
336 in \emph{Proc. of Int. Conf. Bio. \& Pharm. Eng.}, Dec 2006, pp. 79--83.
337
338 \end{thebibliography}
339 }
340
341 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
342 \headerbox{Acknowledgment}{name=ack,column=1,span=2,below=references}{
343 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
344 The main concept of this work is inspired by the ideas of the philosopher Majid Jafari-Tabar.
345 }
346
347 \end{poster}%
348 %
349 \end{document}