A SYSTEM ARCHITECTURE FOR THE BING
Brain Image Network Grid
Micael Pedrosa, Luís Alves
Institute of Electronics and Telematics Engeneering of Aveiro (IEETA), Campus de Santiago, Aveiro, Portugal
Ilídio Oliveira, José Maria Fernandes, João Paulo Silva Cunha
IEETA/Dept. of Electronics, Telecommunications and Informatics, University of Aveiro
Campus de Santiago, Aveiro, Portugal
Keywords: Medical Data, Brain Imaging, Grid-Enabled Scientific Repositories, Workflows, Semantic Data Structure.
Abstract: This paper presents a detailed architecture model for the Brain Imaging Network Grid (BING) that will be
the main IT infrastructure of the recently created Portuguese Brain Imaging Consortium. The proposed
architecture follows a service oriented philosophy and is designed to empower medical data sharing and
processing, specifically brain images. Allowing the use of computationally intensive methods like feature
extraction and retrieving of structured information, this system will take advantage of Grid computing new
paradigm. In BING context, Grid infrastructure is the right option to provide the ability to seamless
aggregate distributed computational power, extensive storage resources and high-bandwidth networking.
The goal is to develop a system that simultaneously can provide basic data services, allow collaborative
research between geographically distributed partners (e.g. analysis processes, workflows) and make use of
the Grid computational power.
1 INTRODUCTION
There is a long tradition in the Portuguese scientific
community of R&D in neurosciences where Brain
Imaging (BI) is a sub-area. BI is in the frontier
between neurology, engineering and physics.
Multimodal medical imaging techniques, such as
Magnetic Resonance Imaging (MRI and fMRI) and
Spectroscopy (MRS), Single Photon/Positron
Emitting Tomography (SPECT/PET) among others,
are emergent medical research tools that can provide
valuable information for diagnosis of brain diseases.
High-resolution electroencephalogram (HR-EEG)
and techniques for synchronizing and fuse its
analysis results and several imaging techniques are
also part of BI. Most of the BI related areas, from
medical, engineering or physics are subjects of
research in many R&D groups within the Portuguese
scientific system.
This strong critical mass in BI of the Portuguese
research community paved the way to a “bottom-up”
process, organized initially around a group of young
researchers in BI working in Portugal and abroad, to
propose the creation of a BI centre and network to
join all community efforts. From this process
emerged the consortium of the universities of
Aveiro, Coimbra, Minho and Porto which is now a
reality, already funded in 81.3% for the first 5 years
of operation of a BI centre, located in Coimbra
(where neuroimaging equipment will be installed
such as 3 Tesla MRI equipment), and a national IT
infrastructure that is open to the participation of
other national institutions.
The BI needs IT infrastructure to support distributed
collaborations between different neuroimaging
research member groups. In a previous paper (Cunha
et al., 2007) the authors addressed the national wide
deployment of the BI network at a organizational
and physical deployment level (Figure 2): it will be
constituted by a data provider (BI centre), located at
the University of Coimbra, two integrated data
processing and storage provider nodes, located at the
Universities of Aveiro and Porto, and a basic and
clinical neuroscience data access client node at the
University of Minho.
276
Pedrosa M., Alves L., C. Oliveira I., Fernandes J. and Silva Cunha J. (2009).
A SYSTEM ARCHITECTURE FOR THE BING - Brain Image Network Grid.
In Proceedings of the International Conference on Health Informatics, pages 276-281
DOI: 10.5220/0001549202760281
Copyright
c
 SciTePress

2 A
G
From t
h
Image
N
option t
o
distribut
e
resource
al., 200
7
level of
s
and ac
c
levels.
A
the use
handle t
h
storage
a
analysis
2004) (
A
Grid co
m
commu
n
scientifi
c
together
resource
scientifi
c
virtual e
n
The ap
p
generic
a
vision a
d
Medical
which i
m
the defi
n
and int
e
2006).
Figure 1 - Brai
n
G
RID-E
N
h
e beginning
N
etwork Gri
o
provide th
e
e
d computati
s and high-
ba
7
). In additio
s
ecurity, bot
h
c
ess (Virtual
A
t the same ti
m
of Grid in
h
e demandin
g
a
nd commun
i
workflows
(
A
mendolia, 2
0
m
puting is b
e
n
ity as an ess
e
c
challenges
distribute
d
s, storage s
p
c
instruments
)
n
vironment.
p
lication of
G
a
lly labeled a
s
d
vocates the
u
Research,
H
m
plies the av
a
n
ition and ado
e
roperability
n
Imaging Net
w
N
ABLED
B
a Grid inf
r
d (BING) s
e
ability to s
e
onal power,
a
ndwidth net
w
n, Grids als
o
h
at identity (
d
Organizatio
m
e successful
medical ima
g
requiremen
t
i
cation, and t
o
(
Breton, 200
5
0
05).
e
ing promote
d
e
ntial technol
o
due to its
d
capabilit
i
p
ace, inform
a
)
and use the
G
rid in life
s
HealthGrid
s
u
se of Grid
I
H
ealthcare a
n
a
ilability of
G
ption of inter
n
mechanisms
w
ork Grid.
B
ING
r
astructure,
B
eemed the
r
e
amless aggr
e
extensive sto
w
orking (Cun
h
o
ensure a pr
d
igital certific
a
ns manage
m
examples ex
i
ge processin
g
t
s of large im
o
enable co
m
5
) (Montagn
a
d
in the scie
n
o
gy to solve
l
ability to
b
i
es (comp
u
a
tion sources
m in a integ
r
sciences ca
n
s
. The Health
G
I
nfrastructure
s
n
d Life Scie
n
G
rid Services
n
ational stan
d
(Breton et
B
rain
r
ight
e
gate
rage
h
a et
oper
a
tes)
m
ent)
i
st in
g
to
ages
m
plex
a
t J.,
n
tific
l
arge
b
ring
u
ting
and
r
ated
n
be
G
rid
s
for
n
ces,
and
d
ards
al.,
(Ai
suc
h
an
d
co
m
req
u
an
d
sha
r
2.
1
BI
N
sev
to
h
req
u
typ
e
Go
o
p
ro
(T
a
ap
p
res
u
de
v
20
0
use
r
res
u
Fr
o
act
o
Th
e
sys
t
de
v
the
Th
e
des
p
ro
sys
t
rep
o
Th
e
res
e
stu
d
det
a
ins
t
out
p
cas
e
ass
o
inp
u
Da
t
p
ro
v
an
d
spe
Inf
o
cla
s
At a hig
h
sha and La
m
h
medical da
t
d
promote
m
putational
s
u
ired comput
a
d
education, e
n
r
e and explor
e
1
BING
U
N
G infrastru
c
eral research
h
igh speciali
z
u
irements (e.
g
e
s specificitie
o
d examples
cesses that
m
a
ylor, 1997),
p
lications or
f
u
lts. Althou
g
v
eloped by
i
0
1), when po
s
r
level as in
m
u
lts.
o
m the analys
o
rs and use c
a
e
informatics
t
e
m
, that
v
elopment of
d
semantic par
t
e
digi
t
al sign
a
cribe and r
e
cessing meth
t
em with h
e
o
rt outputs to
e
clinical exp
e
e
arch seman
d
ies, image
d
a
ched from
t
ance, validat
i
p
ut from sig
n
e
s, identify f
r
o
ciate patien
t
u
ts and result
s
t
a security
a
v
iding securi
t
d
discard of
cific actor ty
p
o
rmation an
d
s
sification an
d
h
er level, se
m
pros, 2006)
t
a analysis pr
o
the decou
p
s
ervices. Gr
i
a
tional and d
a
n
abling the c
o
e
clinical cas
e
U
se Cases
c
ture will n
profiles fro
m
z
ed research
u
g
. computatio
n
s, complex p
r
can be asso
c
m
ay be requi
r
for data
v
f
or interpreta
t
g
h some of
i
nformatics
r
s
sible the us
a
m
edical valid
a
is of some s
c
a
ses were ide
n
enginee
r
w
h
provides e
x
d
ata interpret
e
t
of the syste
m
a
l processing
r
e
use algorit
h
ods (that ma
y
e
lp of infor
m
clinical pers
o
e
rt whose res
tic where
c
d
ata and repo
r
the BING
i
on of study
r
n
al processing
r
om reports s
i
t
informatio
n
s
.
a
ccess base
d
t
y situations
l
non releva
n
p
e.
d
work share
d
organizatio
n
mantic-enabl
e
can suppor
t
o
tocols and
w
p
ling betwe
e
i
d will leve
a
ta resource t
o
o
mmunities t
o
e
s.
eed to acc
o
m
simple data
p
u
ser with ver
y
n
al power, se
v
r
ocessing wor
k
c
iated to digi
t
r
ed for imag
e
v
isualization
t
ion on case
s
these use
r
esearchers (
P
a
ge should b
a
tion of work
f
c
enarios the
f
n
tified:
h
o requires
a
x
tension po
e
rs and addin
g
m
.
r
esearcher tha
t
h
ms, test ne
w
y
be integrat
e
m
atics engin
e
o
nal for evalu
a
earch is at t
h
c
oncepts lik
e
r
ts should be
engine det
a
r
esults and c
o
, comparativ
e
ignificant co
n
n
to a rang
e
d
on sema
n
l
ike patient a
n
n
t informati
o
between ac
t
n
.
e
d Grids
services
w
orkflows
e
n Grid
rage
t
he
o
research
o
publish,
o
mmodate
p
roviders
y
specific
v
eral data
k
flows).
t
al signal
e
analysis
in client
s
analysis
must be
P
oliakov,
e kept at
f
lows and
f
ollowing
a
generic
ints for
g
them to
t
needs to
w
image
e
d in the
e
ers) and
a
tion.
h
e clinical
e
patient
logically
a
ils. For
o
nsequen
t
e
query of
n
clusions,
e
of data
n
tic rules
n
onymity
o
n for a
t
ors, data
A SYSTEM ARCHITECTURE FOR THE BING - Brain Image Network Grid
277

2.2
B
In this
c
oriented
to speci
f
the int
e
framew
o
p
rovidi
n
where
definitio
tear of
s
p
rocessi
n
standar
d
F
The init
i
natural
c
regardle
s
need su
sharing
(acquisi
t
is essen
t
really c
o
the leve
r
able to
p
rovidi
n
b
ioengi
n
integrat
e
neuroim
a
The co
n
other i
Biomed
i
et al., 2
0
easily a
n
B
ING Serv
i
c
ontext it wa
s
solution wh
e
f
ic needs of e
a
e
grity and
o
rk that from
n
g data stora
g
we envisag
e
n of the basi
c
s
ervices for a
n
g power for
d
access to het
e
F
igure 3 - BIN
G
i
al focus on
d
c
onsequence
s
s of their s
k
ch service.
I
and transp
a
t
ions, referen
c
t
ial to ensure
o
nnect the dis
t
r
age for coo
p
contribute
n
g the raw n
n
eering analy
s
e
d computati
o
a
ging researc
h
n
cept of the
B
nternational
i
cal Informat
i
0
05)) and will
n
d foster othe
r
i
ces
s
natural to
a
e
re services c
a
a
ch user with
o
expansibility
the initial s
t
g
e - to the
e
that BIN
G
c
services w
i
vast sort of
complex sig
n
e
rogeneous d
a
G
Architecture
d
ata storage
I
of the fact
t
k
ills and int
e
I
t is essenti
a
a
rent access
c
e models, et
c
that the dat
a
t
ributed resea
r
p
eration wher
e
with specifi
c
euroimaging
s
is methods,
u
o
nal methods
h
purposes.
B
ING is str
o
experiences
i
cs Research
be complete
l
r
nodes from
a
a
ssume a ser
v
a
n be custo
m
o
ut comprom
i
of the B
I
t
age – centr
e
future evolu
t
G
, through
i
ll be able of
f
clients, provi
n
al analysis a
n
a
ta storages.
Overview.
I
T facilities i
s
t
hat BING u
e
nded usage,
a
l to enable
to digital
c
.). This appr
o
a
IT network
r
ch centers a
n
e
each memb
e
c
skills na
m
data, develo
p
u
sing the diff
e
to their cli
n
o
ngly inspire
d
(such as
Network (G
r
l
y opened to
g
a
ny other res
e
v
ice-
m
ized
i
sing
I
NG
e
d in
t
ions
the
f
er a
ding
n
d a
s
the
sers,
will
safe
data
o
ach
will
n
d be
e
r is
m
ely:
p
ing
e
rent
n
ical
d
on
the
r
ethe
g
row
e
arch
gro
u
thi
s
gro
u
net
w
of
tru
s
co
m
kn
o
p
ra
c
eq
u
3
BI
N
ap
p
illu
wo
r
arc
h
car
e
dat
a
3.
1
Cli
e
Int
e
sta
n
in
t
tha
t
ha
n
3.
2
Th
e
en
g
ind
e
p
o
w
p
ro
seq
u
wo
r
p
ro
t
3.2
Th
e
de
fi
W
U
tha
t
co
m
im
p
mi
n
(ta
s
Th
e
co
n
u
ps that may
s
particular a
s
u
nd to deplo
y
w
ork. It add
r
Virtual Org
a
s
ted organi
z
m
putational r
e
o
wledge (dat
a
c
tices and
u
ipments (suc
h
BING A
R
N
G architect
u
p
roach in th
r
strated Fig
u
r
kflow engi
n
h
itecture wil
l
e
portals, cu
a
analysis an
d
1
BING I
n
e
nts can con
t
e
rface, a tie
r
n
dard SOAP/
W
t
o the Virtua
t
addresses
a
n
dles security
2
BING
W
e
central par
t
g
ine. The g
o
e
pendent of
d
w
er provider
s
cessing ope
r
u
ences of
p
r
kflows eith
e
t
ocols or as s
i
.1 Wokflo
w
e
basic co
n
fi
nition is the
W
U
can be tas
k
t
can be cr
e
m
patible p
r
p
lementation
n
imal semant
i
s
k, data or not
e
interface c
o
n
cept that,
join this init
s
pect Grid t
e
y
such a colla
b
r
esses specifi
c
a
nizations, u
n
z
ations, will
i
e
sources (pr
o
a
and anal
y
also spec
i
h
as advance
d
R
CHITE
C
u
re can be de
r
ee main fu
n
u
re 4: BI
N
n
e and Virtu
a
l
serve clie
n
stom applica
t
d
visualizatio
n
n
terface
t
ac
t
the inte
r
r
of Web
S
W
SDL. Its p
u
l File Syste
m
a
ll other par
t
access.
W
orkflow
E
t
of BING is
o
al here is
d
ata storage
s
that, besi
d
r
ations enab
l
p
rocessing o
p
e
r as neuroi
m
i
mple batch p
r
w
Unit
n
cept for
W
orkflow Un
k
s, data refer
e
e
ated, sent
a
r
ovider typ
e
but abstra
c
i
c associated
ification).
o
ncept is a n
a
besides be
i
t
iative in the
f
e
chnology is
b
orative and
c
c
ally the req
u
n
derstood as
i
ng to sh
a
o
cessors and
y
sis workflo
w
i
fic and
e
d
medical sca
n
C
TURE
scribed in a
t
n
ctional divi
N
G Interfac
e
a
l File Syst
e
n
t applicatio
n
t
ions, client
n
.
r
nal engine
v
S
ervices pro
v
u
rpose is to o
f
m
, a structur
e
t
s of the e
n
E
ngine
the BING
W
to supply
a
nature and p
d
es direct
a
l
e the defi
n
p
erations as
m
aging sema
n
r
ocessing inst
r
operations
w
n
it (WU).
e
nces or BI
N
a
nd received
e
. WU
a
c
t interfaces
depending o
n
a
tural object
–
i
ng indepe
n
f
uture. In
a natural
c
ontrolled
u
irements
teams of
a
re their
storage),
w
s), best
e
xpensive
n
ners ).
t
op-down
sions, as
e
, BING
e
m. This
n
s, health
tools fo
r
v
ia BING
v
ided by
f
fer a way
e
d system
n
gine and
W
orkflow
a
system
rocessing
a
ccess to
n
ition of
abstract
n
tic level
r
uctions.
w
orkflow
N
G events
by any
a
re not
with a
n
the type
–
oriented
n
dent of
HEALTHINF 2009 - International Conference on Health Informatics
278

implementation issues it provides mechanism for
extending and complementing the basic WU
concepts. By this way it is possible to accommodate
different software implementation solutions without
compromising the workflow definition – namely by
hiding specific state implementations and providing
only clear access functions with clear semantics.
Tasks, data and notifications are all WU that can be
used with any compatible provider type.
3.2.2 Workflows
The workflow concept in BING is not new and
maintain a high level description of software
architecture. A workflow can be seen as a succession
of data WU and processing WU, which in the end
will generate some output that will be stored in
BING – typically an image processing result.
To describe the workflow and maintain an high level
description, XML based solutions such as “Open
Symphony Workflow” (opensymphony, 2008) can
be a good solution as they are is flexible enough
while providing some semantic for workflow
validation through a Document Type Definitions
(DTD) (Shiyong et al., 2005).
With visual client tools using already available
operators and tasks extensions, medical researchers
are able to compose a workflow with no advanced
informatics knowledge.
3.3 Virtual File System
Virtual File System (VFS) is responsible for
providing an abstraction of the information related
services of the BING: organize instances of WUs,
enable access to the BING providers.
Providers can be data store servers (e.g. like FTP,
GSIFTP), or other processing hosts (e.g. location to
Grid Nodes). The VFS will be responsible for
keeping the mapping between the providers and the
associated VFS folders and WUs. In BING,
providers and WUs can have modality specific
semantic such EEG, fMRI, MEG, CT (Productions,
2001) which may facilitate the semantic verification
of workflows.
The VFS concept is similar to common file systems,
more like a semantic file system (Craig A. N.
Soules, 2004) (Prashanth Mohan, 2006) but in this
case, instead of files, it stores WUs and every folder
and WU reference are abstract structures mapped on
actual data/processing facilities in different
providers, local or remote.
Another important responsibility of the VFS is to
provide a secure and reliable environment where
users and permission access to both folders and
WUs is controlled. This implies that VFS must
preserve users and permissions configuration and
provide anonymity of data when required (e.g.
enable access to data without any patient
information).
Given the VFS abstract nature, it will be subject to
provider’s access policies. Nevertheless it will
enable a fine grained configuration in order to define
access and security policies at the BING level.
3.4 Persistence
The Processing and workflow Engine will need a
persistence mechanism in order to store both WU
information and maintain the VFS information. The
natural choice for a preliminary approach will be a
relational database that 1) enable fast development
for a specialized software engineer team, 2) are
recognised for their performance in query relational
and structured data. An example of a similar
problem can be found applied to semantic file
systems in (SELENG, 2007).
3.5 BING and Grid Middleware
The BING can be seen as part of the Grid
middleware where the Grid framework essentially
supply basic data and processing providers. While
maintaining a BING specific semantic, the Grid
Infrastructure can be used to access computing
power, an aggregation of distributed storage and
heterogeneous resources.
This solution does not compromise the
interoperability concern commonly referred on Grid
conceptions (eDiaMoND, 2008) (Foster et al.,
2001). It is important that the various stakeholders of
a Virtual Organization can transparently use all the
resources available (data, sensors, etc.) in
collaborative terms. Interoperability is achieved by
the use of a standards-based architecture, commonly
named as Grid middleware. The Grid middleware
dynamically shares, manages and controls physical
and logical resources geographically separated. In
some sense, BING relies on this functionalities for
IT infrastructure while providing services more
specific to BI support infrastructure, similarly to the
the Biomedical Informatics Research Network
(Grethe et al., 2005).
3.6 Grid Middleware Providers
One good candidate to be the BING Grid
middleware is the gLite (gLite, 2008). The gLite
A SYSTEM ARCHITECTURE FOR THE BING - Brain Image Network Grid
279

middleware has been introduced by EGEE project as
a result of contributions from many other projects
including LCG, European DataGrid and VDT
(gLite, 2007). This middleware is based on
GLOBUS2, Condor and many other services
developed in projects like those above. gLite
orchestrates multiple service layers such as Access,
Security, Information and Monitoring, Data, and Job
Management Services (gLite, 2007). It also
accomplishes the Open Grid Services Architecture
(OGSA) Standard introduced by the Open Grid
Forum (OGF) (Foster et al., 2002). This software is
Open licensed and portable.
The gLite job management is guaranteed by the use
of a gLite User Interface (UI) node that will operate
as a BING gateway to the Grid layer. This node
provides the proxy creation and, by the use of the
WMProxy (Workload Manager Proxy) Service,
access to the Workload Management System
(WMS). WMProxy accepts job submission requests
described with the Job Description Language (JDL)
(Pacini, 2006) and control requests such as job
cancellation, job status, job output retrieval, etc
(Maraschini A., 2006).
For data providers the gLite middleware supply
complementary data management services like the
Storage Resource Manager (SRM) and can support
different data access protocols and interfaces like
GSIFTP, RFIO and gsidcap allowing advanced
functionality such as dynamic space allocation and
file management on shared storage systems (gLite,
2007).
Some early results from our group (Andrade et al.,
2007b) support that gLite might be a good option.
On a vertical application (from user interface,
storage and processing) on Multi-voxel Non-linear
fMRI Analysis (Andrade et al., 2007a) it was
possible to have a positive assessment on the gLite
both on the processing power ( report a 7 fold
reduction in time consumption on EGEE Grid
environment) and on the application development.
4 DISCUSSION
This paper presents a detailed architecture model for
the Brain Imaging Grid (BING) that will be the main
IT infrastructure of the recently created Portuguese
Brain Imaging Consortium. The architecture was
designed with flexibility in mind providing
extension points for semantic information search and
query.
Currently a prototype is under development
integrating extensions to the basic WU concepts:
LocalGliteTask WU to execute gLite jobs,
LocalTask WU to map local processes, FtpData WU
to access to data in FTP and LocalData WU to map
local data storage. These are being used in BING
workflow engine development in order to support
simple workflows. Simultaneously we are
developing the basic processes to handle different
data formats, image modalities in order to be able to
support the complex brain image data model at the
VFS level.
ACKNOWLEDGEMENTS
This work was partially funded by the projects
BING (GRID/GRI/81833/2006) and Geres-Med
(GRID/GRI/81819/2006) of the F.C.T.
We would like also to thank our project colleagues
António Sousa Pereira (University of Aveiro),
Miguel Castelo-Branco (University of Coimbra),
Nuno Sousa (University of Minho), Aurélio
Campilho (University of Porto) and their institutions
for the support to the BING project.
REFERENCES
Aisha, N. & Lampros, K. S. (2006) Discovering
HealthGrid Services. Services Computing, 2006. SCC
'06. IEEE International Conference on.
Amendolia, S. R., et al. (2005) Deployment of a grid-
based medical imaging application, IOS Press.
Andrade, R., Oliveira, I., Fernandes, J. M. & Cunha, J. P.
(2007a) A Grid framework for non-linear brain fMRI
analysis. Stud Health Technol Inform, 126, 299-305.
Andrade, R., Oliveira, I., Fernandes, J. M. & Cunha, J. P.
(2007b) Multi-voxel Non-linear fMRI Analysis: A
Grid Computing Approach. IberGRID Santiago de
Compostela.
Breton, V., Blanquer, I., Hernandez, V., Legré, Y. &
Solomonides, T. (2006) Proposing a roadmap for
HealthGrids. HAL - CCSD.
Breton, V., Dean, K., et al. (2005) The Healthgrid White
Paper. Stud Health Technol Inform, 112, 249-321.
Craig A. N. Soules, G. R. G. (2004) Toward automatic
context-based attribute assignment for semantic file
systems. Pittsburgh, Carnegie Mellon University.
Cunha, J. P. S., Oliveira, I., Fernandes, J. M., Campilho,
A., Castelo-Branco, M., Sousa, N. & Sousa Pereira, A.
(2007) BING: The Portuguese Brain Imaging Network
GRID. IberGRID Santiago de Compostela.
Ediamond (2008) eDiaMoND – Diagnostic
Mammography National Database Project
Foster, I., Kesselman, C., Nick , J. & Tucke, S. (2002) The
Physiology of the Grid: An Open Services
Architecture for Distributed Systems Integration. .
HEALTHINF 2009 - International Conference on Health Informatics
280

Open Grid Service Infrastructure WG, Global Grid
Forum.
Foster, I., Kesselman, C. & Tuecke, S. (2001) The
Anathomy of the Grid. Intl J. Supercomputer
Applications, 15.
GLITE (2007) gLite User Guide.
GLITE (2008) gLite: Lightweight Middleware for Grid
Computing.
Grethe, J. S., Baru, C., Gupta, A., James, M., Ludaescher,
B., Martone, M. E., Papadopoulos, P. M., Peltier, S.
T., Rajasekar, A., Santini, S., Zaslavsky, I. N. &
Ellisman, M. H. (2005) Biomedical informatics
research network: building a national collaboratory to
hasten the derivation of new understanding and
treatment of disease. Stud Health Technol Inform, 112,
100-9.
Maraschini A., E. A. (2006) WMPROXY SERVICE. IN
INTEGRATION, J. M. E. A. (Ed.) EGEE User’s
Guide. EGEE.
Montagnat J., B. F., Benoit-Cattin H., Breton V., Brunie
L., Duque H., Legré Y., Magnin I. E., Maigne L.,
Miguet S., Pierson J. -M., Seitz L., Tweed T. (2004)
Medical Images Simulation, Storage, and Processing
on the European DataGrid Testbed. Journal of Grid
Computing, 2.
Opensymphony (2008) OSWorkflow.
Pacini, F. (2006) JOB DESCRIPTION LANGUAGE -
ATTRIBUTES SPECIFICATION (submission
through WMProxy Service). IN MIDDLEWARE, J.-.
(Ed.). EGEE.
Poliakov, A. V. A. A., Evan M and Corina, David P and
Ojemann, George A and Martin, Richard F and
Brinkley, James F (2001) Server-Based Approach to
Web Visualization of Integrated 3-D Medical Image
Data. Proceedings, American Medical Informatics
Association Fall Symposium.
Prashanth Mohan, R., Venkateswaran S and Dr. Arul
Siromoney (2006) Semantic File Retrieval in File
Systems using Virtual Directories. International
Conference on High Performance Computing (HiPC
2006). Bangalore, India
ProductionS, T. W. N. Y. A. D. G. (2001) The Secret Life
of the Brain.
Seleng, M.-L., Michal - Balogh, Zoltan - Hluchy, Ladislav
(2007) RDB2Onto: Approach for creating semantic
metadata from relational database data.
INFORMATICS 2007 Bratislava, Slovak Society for
Applied Cybernetics and Informatics.
Shiyong, L., Yezhou, S., Mustafa, A. & Farshad, F. (2005)
On the consistency of XML DTDs. Data Knowl. Eng.,
52, 231-247.
Taylor, J. G. (1997) Problems in the analysis of brain
imaging data. Digital Signal Processing Proceedings,
1997. DSP 97., 1997 13th International Conference
on.
A SYSTEM ARCHITECTURE FOR THE BING - Brain Image Network Grid
281