UMinho 2019
Rogé io Gomes Lopes Mo ei a
Building an imaging-based esea ch pla o m o expe imen s wi h b ain connec i i y da a
No embe 2019
Rogé io Gomes Lopes Mo ei a
Building an imaging-based esea ch
pla o m o expe imen s wi h
B ain connec i i y da a
Ana Filipa de Oli e ia Ramos Deep Lea ning Applied o Medical Imaging
i
Rogé io Gomes Lopes Mo ei a
Building an imaging-based esea ch
pla o m o expe imen s wi h b ain
connec i i y da a
Mas e ’s Disse a ion
In eg a ed Mas e ’s in In o ma ics Enginee ing
Disse a ion o ien ed by
Vic o Manuel Rod igues Al es
Nicolás F ancisco Lo i
No embe 2019

i
Acknowledgemen s
I would like o hank my supe iso , Vic o Al es o his guidance, a ailabili y, sha ing and encou agemen .
His ad ices and assis an we e uly impo an o me. I would also like o hank Nicolás Lo i, my co-
supe iso , who augh me so much abou neu oimaging.
Finally, o my pa en s, my b o he and Jo ge o all he suppo and ad ice in my decisions. To my cou se
colleagues, Samuel, Gus a o and Diogo who suppo ed me g ea ly and we e always willing o help me.
To Luis, my eache who uly changed he way I saw he wo ld. I’m o e e hank ul o his insigh s and
mind opening. And inally, o Madalena, o being wi h me all he ime in good and bad and suppo ing
me h ough my academic jou ney. Thank you o all you lo e.
ii
DECLARATION
Name: Rogé io Gomes Lopes Mo ei a
Disse a ion Ti le: Building an imaging-based esea ch pla o m o expe imen s wi h b ain connec i i y
da a
Men o s: Vic o Manuel Rod igues Al es, Nicolás F ancisco Lo i
Conclusion Yea : 2019
Mas e Designa ion: Mes ado In eg ado em Engenha ia In o má ica
Mas e B anch: In o má ica Médica
I decla e ha I g an o he Uni e si y o Minho and i s agen s a non-exclusi e license o ile and make
a ailable h ough i s eposi o y, in he condi ions indica ed below, my disse a ion, as a whole o pa ially,
in digi al suppo .
I decla e ha I au ho ize he Uni e si y o Minho o ile mo e han one copy o he disse a ion and, wi hou
al e ing i s con en s, o con e he disse a ion o any o ma o suppo , o he pu pose o p ese a ion
and access.
Fu he mo e, I e ain all copy igh s ela ed o he disse a ion and he igh o use i in u u e wo ks.
I au ho ize he pa ial ep oduc ion o his disse a ion o he pu pose o in es iga ion by means o a
w i en decla a ion o he in e es ed pe son o en i y.
This is an academic wo k ha can be used by hi d pa ies i in e na ionally accep ed ules and good
p ac ice wi h ega d o copy igh and ela ed igh s a e espec ed.
Thus, he p esen wo k can be used unde he e ms o he license indica ed below.
In case he use needs pe mission o be able o make use o he wo k in condi ions no o eseen in he
indica ed licensing, he should con ac he au ho h ough he Reposi o iUM o he Uni e si y o Minho.
A ibuição-NãoCome cial-SemDe i ações
CC BY-NC-ND
h ps://c ea i ecommons.o g/licenses/by-nc-nd/4.0/
Uni e sidade do Minho, ____/____/______
Signa u e: ___________________________________
iii
STATEMENT OF INTEGRITY
I he eby decla e ha ing conduc ed his academic wo k wi h in eg i y. I con i m ha I ha e no used
plagia ism o any o m o undue use o in o ma ion o alsi ica ion o esul s along he p ocess leading o
i s elabo a ion.
I u he decla e ha I ha e ully acknowledged he Code o E hical Conduc o he Uni e si y o Minho.
Uni e sidade do Minho, ____/____/______
Signa u e: ________________________________________
i
ABSTRACT
Wi hin he pas decade, no only socie ies in gene al bu also medicine and heal hca e, in pa icula , ha e
changed emendously. In la ge pa because o he apid dissemina ion o compu e s and digi al
communica ions which lead o he appea ance o new medical disciplines, such as Medical In o ma ics.
Nowadays, one o he mos p ominen ield in Medical In o ma ics is Medical Imaging, as i is implied, i
is a collec ion o me hodologies and echniques used in o de o isually and spa ially ep esen pa s o
he b ain o diagnos ic and esea ch pu poses.
In he esea ch ecosys em, Neu oimaging is an inc easing popula ield, wi h applica ions in neu ology
and psychia y. Howe e , due o he di icul ies o handle Neu oimaging da a, since da a has i s own
speci ici ies, esea che s ha e encoun e ed p oblems o co ec ly handling his da a. This can be a c ucial
issue specially wi h la ge olumes o Neu oimaging da a and all he esea ch ma e ials associa ed.
This wo k aims o a chi ec and build a esea ch pla o m o co ec ly a chi e Medical Imaging da a and
all he associa ed esea ch ma e ials, whe e esea che s can exchange imaging da a and collabo a e in
Neu oimaging esea ch p ojec s. The pla o m o e s a co ec way o collec and s o e all imaging da a,
a chi ing all o pa ien exams wi h he co esponden in o ma ion, making a ailable he co esponden
in o ma ion o esea che s in a con iden ial, secu e and e icien way.
The wo main ou comes o his wo k a e an a chi ec u e o a pla o m ha manages all imaging da a and
associa ed esea ch ma e ials, plus an open-sou ce Py hon package o easily in e ac wi h ha pla o m.
• Keywo ds: Da a Sha ing, Medical Imaging, Neu oin o ma ics, Neu oimaging, XNAT.
xi
ABBREVIATIONS AND ACRONYMS
A
AI
A i icial In elligence
API
Applica ion P og amming In e ace
B
BIRN
Biomedical In o ma ics Resea ch Ne wo k
C
CNDA
Cen al Neu oimaging Da a A chi e
CSV
Comma-sepa a ed alues
CT
Compu ed Tomog aphy
D
DICOM
Digi al Imaging and Communica ions in Medicine
DBMS
Da abase Managemen Sys em
DL
Deep Lea ning
F
MRI
Func ional magne ic esonance imaging
G
GUI
G aphical Use In e ace
H
HDD
Ha d Disk D i e
HTTP
Hype ex T ans e P o ocol
HCP
Human Connec ome P ojec
J
JAAT
Join Anonymiza ion and A chi e T ansmission
L
Loni IDA
Loni Image Da a A chi e

xii
LXC
Linux Con aine s
M
ML
Machine Lea ning
MRI
Magne ic Resonance Imaging
MR
Magne ic Resonance
MVCC
Mul i e sion concu ency con ol
N
NI TI
Neu oimaging In o ma ics Technology Ini ia i e
NM
Nuclea Medicine
NIAC
Neu oimaging In o ma ics and Analysis Cen e
NGINX
Engine-X
NA-MIC
Na ional Alliance o Medical Image Compu ing
O
OS
Ope a ing Sys em
P
PET
Posi on-Emission Tomog aphy
PITR
Poin -in-Time-Reco e y
R
REST
Rep esen a ional S a e T ans e
S
SFTP
Secu e File T ans e P o ocol
SCP
Se ice Class P o ide
SPECT
Single-pho on emission compu ed omog aphy
SQL
S uc u ed Que y Language
SSD
Solid-S a e D i e
U
UI
Use In e ace
UMINHO
Uni e si y o Minho
xiii
US
Uni ed S a ed
V
VM
Vi ual Machine
X
XDAT
eX ensible Da a A chi e Toolki
XML
eX ensible Ma kup Language
XNAT
eX ensible Neu oimaging A chi e Toolki
XSD
XML Schema De ini ion
XTK
X-Ray Toolki
Y
YAML
YAML Ain' Ma kup Language
xi
GLOSSARY
Da ase
A collec ion o examples. Each example con ains one o mo e ea u es
and a label (i using supe ised aining).
DICOM
DICOM (Digi al Imaging and Communica ions in Medicine) is a s anda d
o handling, s o ing, p in ing, and ansmi ing in o ma ion in Medical
Imaging. I includes a ile o ma de ini ion and a ne wo k
communica ions p o ocol.
Magne ic Resonance
Imaging (MRI)
Non-in asi e echnique used disease de ec ion and ea men
moni o ing. MRI scanne s a e pa icula ly well sui ed o image he so
issues o he body e.g., b ain, spinal co d and ne es, muscles and
ligamen s.
Medical Imaging
In o ma ics
A discipline ha esul s om he combina ion o medical images and
biomedical in o ma ics. Se es o enhance he imp o emen and
knowledge o clinical ca e. On he one hand, Medical Imaging allows he
s udy o a disease s a e
in i o
. On he o he hand, biomedical
in o ma ics is in ol ed in he de elopmen o compu e science
echniques o c ea ing and managing medical da a.
Ni y
A simple, minimalis ic o ma which has been widely adop ed in
Neu oimaging esea ch, allowing scien is s o mix and ma ch image
p ocessing and analysis ools de eloped by di e en eams.
XNAT
An open sou ce imaging in o ma ics so wa e pla o m designed o
acili a e common managemen and handling o Neu oimaging and
associa ed da a.
INTRODUCTION
1
INTRODUCTION
INTRODUCTION
16
1.1
CONTEXT AND MOTIVATION
Medical in o ma ics is a ecen medical discipline when compa ing wi h o he medical disciplines ha
ha e been a ound o cen u ies [1]. Wi hin he pas decade socie ies in gene al, and medicine and
heal hca e in pa icula , ha e emendously changed, in la ge because o he apid dissemina ion o
compu e s and digi al communica ions [2]. This allowed new medical disciplines o appea , such as
Medical In o ma ics, which is he in e sec ion o compu e in o ma ics and heal hca e [3]. This new
discipline deals wi h he esou ces, de ices, and me hods equi ed o op imize he acquisi ion, s o age,
e ie al, and use o in o ma ion in heal hca e sys ems.
One o he mos p ominen ield in Medical In o ma ics is Medical Imaging, as i is implied, i is a collec ion
o me hodologies and echniques used in o de o isually and spa ially ep esen pa s o he human
body o diagnos ic and esea ch pu poses. This ield is so impo an nowadays ha i is ha d o imagine
a medical diagnos ic wi hou he use o imaging ools since hey allow doc o s o ha e a clea pic u e o
he p oblem. In esea ch, Medical Imaging has gained conside able a en ion due o i s widesp ead u ili y
in a mul i ude o clinical applica ions and he conside able imp o emen s in he echniques [4].
One o he mos p ominen ields in Medical Imaging is he ep esen a ion/s udy o he b ain, known as
Neu oimaging. This ield is a se o imaging echniques ha ei he di ec ly o indi ec ly image he s uc u e
o unc ion o he ne ous sys em, which includes he b ain [5]. Neu oimaging comp ises se e al di e en
modali ies such as Magne ic Resonance Imaging (MRI), Posi on Emission Tomog aphy (PET), Compu ed
Tomog aphy (CT) and unc ional MRI ( MRI) [6]. These echniques ha e majo clinical applica ions in
neu ology, se ing as a c ucial pa in diagnosis and, mo e ecen ly, in psychia y, wi h special ocus on
psychia ic diso de s by helping o unde s and he in e ac ions be ween he b ain and he body. All o
hese make Neu oimaging an inc easing popula ield o esea ch, especially in he las 20 yea s [7].
In conjunc ion wi h his, in he las ew yea s has been a g owing in e es in including Machine Lea ning
(ML) [8] and Deep Lea ning (DL) [9] in he analysis o Neu oimaging da a.
Machine Lea ning add esses he ques ion o how o make compu e s imp o e au oma ically hough
expe ience. I is one o mos apidly g owing echnical ields, joining a eas like compu e science, s a is ics
and da a science [10].
Deep Lea ning is one o he mos ecognized ML echniques, which allows compu a ional models
composed o mul iple laye s o lea n ep esen a ions o da a wi h mul iple le els o abs ac ion. These
me hods ha e d ama ically imp o ed ha d compu e p oblems like speech ecogni ion, isual objec

INTRODUCTION
17
ecogni ion and objec de ec ion. Deep Lea ning disco e s a s uc u e in la ge da a se s by using an
algo i hm o indica e how a machine should change i s in e nal pa ame e s, used o compu e he
ep esen a ion in each laye om he ep esen a ion in he p e ious laye [11].
A la ge numbe o s udies show ha ML and DL can be used o ex ac new exci ing in o ma ion om
Neu oimaging da a [12]. This ield o esea ch has e ealed i sel o be e y complex om he beginning
as i equi es analysis o complex and mul i a ia e da a [13]. Da a wi h bo h hese cha ac e is ics is
p ecisely wha leads o he de elopmen o compu e echniques such as ML and DL which allow o ain
classi ie s ha decode da a and ex ac he ele an in o ma ion. This would no be possible wi h he so-
called “ adi ional compu ing”, since he olumes o in o ma ion o decode a e ex emely la ge [13].
The ecen p og ess in ML has been d i en bo h by he de elopmen o new algo i hms and by he
a ailabili y o da a. Wi h his, esea che s ely mo e and mo e upon la ge da ase s ha some imes a e
no co ec ly labeled and a chi ed, which can esul in un eliable esul s. This is especially ue in
Neu oimaging since he da a is di icul o handle and is mos o he ime sliced in mul iple iles.
Fo all o hese easons, a esea ch pla o m o co ec ly s o e Medical Imaging da a plus all he associa ed
esea ch ma e ials is a c ucial pa o a la ge sough ecosys em whe e esea che s can exchange da a
and collabo a e in Neu oimaging esea ch p ojec s.
One o hese ecosys ems was de eloped h ough he collabo a ion be ween he Algo i mi Resea ch Cen e
and he Li e and Heal h Sciences Resea ch Ins i u e (ICVS), bo h a Uni e si y o Minho (UMinho) [14],
wi h he objec i e o map and sea ch o pa e ns on he human b ain (Fig 1.1).
Figu e 1.1 – UMinho Neu oimaging p ojec
This wo k has he pu pose o a chi ec a pla o m o a chi e Neu oimaging da a and build a p o o ype
wi h he planned a chi ec u e o be in eg a ed in he cu en sys em wo k low o he UMinho Neu oimaging
p ojec bu implemen ed in such a way ha i can easily be ans e ed o o he wo k low sys ems. The
de eloped pla o m has he objec i e o acili a e he dis ibu ion o Neu oimaging da a a ailable in bo h
INTRODUCTION
18
public da ase s and acqui ed locally a he ICVS MRI scanne . Wi h he wo k low used be o e, esea che s
mo ed Neu oimaging da a and he associa ed esea ch ma e ials ia ha d-d i es o pla o ms like
Redmine [15], which is no scalable as he da ase sizes inc ease.
The pla o ms ha handle Medical Imaging da a and in pa icula neu oimages ha e unique
cha ac e is ics, ha make i di e en om adi ional sha ing sys ems. Also, medical esea ch has igid
wo k lows ha can’ be changed easily. Commonly, pla o ms ha y o change hese wo k lows don’
succeed. The sys em o be buil mus seamlessly in eg a e in he ac ual wo k low, wi h only mino
adap a ions, bu also be in eg able in a global Neu oimaging da a sha ing ne wo k (Fig. 1.2).
Figu e 1.2 – Example o mul iple Medical Imaging pla o ms connec ed o sha e esea ch da a
O e he las yea s, ICVS and Algo i mi ha e been collabo a ing in s udies abou Neu oimaging and
nowadays all he ongoing esea ch in ol es a la ge numbe o dispe se esea che s, la ge da ase s and
di e en da a sou ces. Wi h his inc ease in popula i y, a new p oblem was encoun e ed. How o p ope ly
s o e, manage and make a ailable his da a and he esea ch ma e ials associa ed?
The me hods used un il now duplica e he space used and don’ ensu e p ope ly da a a chi ing, plus, as
esea ch inc easingly specializes in a na owe ange (Fig. 1.2) o applica ions i becomes ha de and
ha de o ob ain su icien da a i i is dispe sed ac oss esea che s and p ojec s.
INTRODUCTION
19
Figu e 1.3 - Compa ing scien i ic quali y e sus amoun o sha ed da a (image om [16])
Taking o example ano he esea ch ecosys em like he one a UMinho, he Cance Imaging A chi e [17]
has mo e han 25.000 pa ien s/subjec s, 73.000 s udies and 21 million images which ep esen s mo e
han 11 e aby es o da a, all a ailable o esea che s. Any esea che can easily selec da a om he
global da ase wi h he equi ed cha ac e is ics and use he da a wi hou he need o make a new
acquisi ion.
1.2
OBJECTIVES
The main objec i e o his wo k is o a chi ec u e an imaging-based esea ch pla o m o he
implemen a ion o expe imen s wi h Medical Imaging da a and associa ed esea ch ma e ials, wi h a
pa icula ocus on Neu oimaging. Also, as his ype o pla o m in ol es mul iple i e a ions, a p o o ype
o he wo k will be a ailable in ICVS and Algo i mi o manage all he Neu oimaging da a used he e.
Fu he mo e, as esea che s need a simple way o p og amma ically in e ac wi h he pla o m, he e is
also he de elopmen o bo h APIs and lib a ies o he in e ac ion wi h he de eloped pla o m.
1.3
DISSERTATION STRUCTURE
The wo ocuses o his wo k a e he XNAT@DI pla o m and he XNATUM package. The disse a ion is
s uc u ed as ollows. Sec ion 2 e iews published sys ems in he ield and discusses he addi ional
INTRODUCTION
20
equi emen s ha ha e mo i a ed he de elopmen o he XNAT@DI p ojec . Sec ion 3 is an o e iew o
he sys em’s ea u es, echnologies and he o e all a chi ec u e o he pla o m. Sec ion 4 p esen s
implemen a ion de ails o he pla o m. Sec ion 5 p esen s an o e iew and he implemen a ion o he
XNATUM package, including desc ip ions o ea u es, pe o mance and es ing. Finally, Sec ion 6 p esen s
conclusions and u u e wo k.
STATE OF THE ART
27
• Wo k as a ede a ed pla o m whe e p ojec owne s e ain au onomous con ol o e hei da a
and de ine p ojec -speci ic p ocedu es, bu also allow o ha e global p ojec s, which can be used
by any esea che ;
• Ha e he possibili y o connec op ional ga eways se e s wi hin each clinical ins i u ion;
• Includes ea u es like DICOM di ec upload, REST API o upload and download da a, quali y
con ol modules, web-based in e ace, powe ul and sophis ica ed sea ch, online image iewe
and a pipeline engine o au oma ing image p ocessing asks.

PLATFORM DESIGN AND ARCHITECTURE
3
PLATFORM DESIGN AND
ARCHITECTURE
29
The sys em o be concei ed, mus be a secu e da a pla o m ha simpli ies and au oma es he p ocess
o accessing and sha ing esea ch ma e ials o Medical Imaging wi h gene al applicabili y which can be
adap ed o bene i esea ch p ojec s ac oss a a ie y o ields a any esea ch acili y.
The main goal o he esea ch pla o m is o p o ide a lexible, clinical and esea che - iendly sys em o
sha ing Medical Imaging da a and esea ch ma e ials ac oss mul iple depa men s and esea ch cen e s.
The sys em aims o sa is y he demands o pa ien con iden iali y, da a secu i y and o b ing he bene i s
om a web pla o m o imaging esea ch; wi h he goal o encou aging mo e da a sha ing, bo h
s anda dizing and simpli ying he da a ans e and anonymiza ion p ocesses, and he p o iding o
cen alized s o age and backups which simpli y he sha ing o esea che -de i ed esul s. The pla o m
will achie e all o hese aims by building upon exis ing and well-es ablished echnologies, accessible and
usable o di e en ypes o use s, and wi h an ease-o -use app op ia e o hose ha a e mo e o less
amilia ized wi h his kind o da a and echnology.
3.1
PURPOSED SOLUTION
A gene al sys em diag am o he pla o m is shown in Fig.3.1. The pla o m consis s o a cen al se e
and op ional ga eway se e s o clinical ins i u ions and esea ch depa men s. The e o e, he al eady
exis ing echnologies used mus be c oss-pla o m compa ible, o e ing complexi y wi h ega d o he
equi ed ha dwa e and ope a ing sys em. The se ice mus p o ide a da a upload ea u e wi h au oma ed
anonymiza ion o pa ien in o ma ion and which in eg a es wi h so wa e al eady in use a he esea ch
labs like Ho os [40] o Osi iX [41].
30
Figu e 3.1 - Pla o m concep ual a chi ec u e
31
3.2
PLATFORM ARCHITECTURE
3.2.1
OVERVIEW
In consequence o he necessi y o esea ch cen e s o manage da a he e a e a lo o di e en da a
managemen sys ems o acili a e collabo a i e esea ch and da a sha ing in Neu oimaging such as Li y
[42], Cap e a [43], No a [44] and XNAT [45].
The anonymized esea ch and associa ed esea ch ma e ials will be hos ed on a dedica ed ins ance, wi h
his, use s can access and download da a wi hin he secu e websi e om a web in e ace access o any
b owse . The pla o m (Fig. 3.2) may also p o ide a se o p o ocols and se ices o be accessed
p og amma ically, o allow in eg a ion wi h hi d-pa y so wa e and ools, in a o m o a Rep esen a ional
S a e T ans e Applica ion P og amming In e ace (REST API).
Figu e 3.2 – Pla o m concep ual a chi ec u e ne wo k.
3.2.2
INTEGRATION WITH OTHER INSTANCES AND HOSPITAL SYSTEMS
The pla o m mus allow he connec ion o op ional ga eways o enable each ins i u ion and esea ch
depa men o p o ide image uploading mechanisms sui ed o hei needs. Each in e es ed ins i u ion can
p o ide a ga eway om hei in as uc u e o o he s. The e a e wo main ways his is done wi h he
32
cu en s anda ds: syncing p ojec s be ween ins ances, o sha ing aw imaging da a. The i s me hod
es ablishes a pe sis en da a sync o all o selec ed da a be ween ins ances, which allows he esea che
o upload he da a jus once and gua an ee ha he da a is au oma ically uploaded o he o he ins ances.
The second me hod allows, o example, o mo e la ge amoun s o DICOM da a om hospi al acili ies
and esea ch cen e s.
3.2.3
DATA SECURITY AND ACCESS CONTROL
All he da a s o ed on he pla o m o be buil is anonymized and no iden i iable. All da a exchange will
happen o e he In e ne in an enc yp ed connec ion wi h he se e ce i ica e ensu ing clien s can only
connec o he genuine se e . The se e i ewalls es ic access only o us ed clien s and limi he
maximum a emp s o ailed connec ions, blocking consecu i e ailed connec ions.
All da a access equi es a pe sonal accoun . A use can eques a use accoun o he p ima y In es iga o
o o a pla o m adminis a o .
3.2.4
WORKFLOW
The da ase s a e ei he coming om a esea che o di ec ly om a hospi al wi h an MRI machine. The e
a e wo main ways da a can be uploaded o he pla o m, ia REST API o ia he web in e ace. When
uploading da a, he mos s aigh o wa d p ocess is i s o upload all p ojec me ada a and hen he
associa ed subjec s and complemen a y in o ma ion which can be bo h made ia REST API o ia CSV
upload on he web in e ace. Then he anonymized sessions can be uploaded o hei co esponden
p ojec and subjec o o he p e-a chi e o be analyzed by a p ojec esea che be o e inclusion in he
co esponding p ojec .
Resea che s e ie e he anonymized da a using in eg a ed so wa e o simply by downloading needed
da a ia he web in e ace.

33
3.3
TECHNOLOGIES AND CONCEPTS
In his chap e a e desc ibed some o he echnologies ha will be used o build he pla o m as well as
well as a jus i ica ion in why a pa icula echnology was chosen in a o o o he s. The gene al p e e ence
was o use al eady well es ablished open-sou ce echnologies exis en in he ma ke o many yea s and
which ha e p o en hei wo h in he esea ch a ea.
3.3.1
VIRTUALIZATION SOFTWARE
The chosen i ualiza ion so wa e was Docke , a p og am c ea ed o isola e p ocesses om he sys em
on which i is unning. I was c ea ed o isola e an applica ion and he esou ces equi ed o un ha
applica ion om he ha dwa e on which i uns. Docke uses p e-buil Linux Con aine s (LXC) which a e
an ope a ing sys em le el i ualiza ion buil -in di ec ly on he Linux ke nel, wi hou he o e head o a ull-
ledged VM. XNAT applica ion depends in a ew di e en and dis inc applica ions which can cause
di e en ypes o e o s. Docke sol es his p oblem ia con aine iza ion, ensu ing compa ibili y ac oss
pla o ms and, he e o e, elimina ing "wo ks on my machine" issues in esea ch and so wa e de eloping.
Despi e appea ances, he mos common Docke con aine s a e no VMs. They ha e a key di e ence,
while VMs don' sha e he same Linux ke nel, in he Docke con aine s he con a y happens. Docke
con aine s sha e he same Linux ke nel wi h he ha dwa e hey a e unning, esul ing on much less boo
ime and esou ce usage han VM’s. This has made Docke pa icula ly a ac i e o indus y and has hus
seen a s eep ise in adop ion o he echnology. Docke will be used o con aine ize he pla o m on he
se e as well as simpli y and gua an ee he consis ency o he ins alla ion and se up o he sys em, since
some o he used hi d-pa y so wa e used is only a ailable o Linux-based sys ems.
Docke was chosen because i sa is ies he p ac ical equi emen s, and he accessibili y o Docke Hub
enables images o be quickly ound and deployed. “Vi ual machines like hose c ea ed in Vi ual Box [46]
o VMwa e [47] p o ide lo s o ange in e ms o ope a ing sys ems ha can be launched and allow na i e
access o he machine h ough a GUI. Howe e , hough hese a e g ea ea u es, hey a e no necessa y
o he pla o m.” [48]
34
Figu e 3.3 - Docke engine laye s (image om [49])
3.3.1.1
DOCKER MANAGING ENVIRONMENTS
The Docke API allows a lack o op ions o in e acing wi h Docke , he con aine s, and images om CLIs
o desk op applica ions and web-based managemen ools. These ools go om Ki ema ic [50], he de aul
GUI p og am ha ships wi h Docke o Shipya d [51], a web-based app ha p o ides an in e ace o
con aine s. Al hough, he chosen ool was Po aine [52] a UI managemen so wa e which allows o
manage Docke en i onmen s and simply deploy any docke con aine o un on he Docke engine. All
Docke esou ces, including con aine s, images, olumes and ne wo ks can be managed ia Po aine
[52]. This solu ion was chosen since i is web-based, which means ha can be accessible in any b owse
wi hou he need o ins all o he hi d-pa y so wa e, is ee and open-sou ce.
35
Figu e 3.4 - The in e ace o Po aine , he Docke managemen ool
3.3.2
DATA REPOSITORY MANAGEMENT SYSTEM
The chosen Da a Reposi o y Managemen Sys em was he eX ensible Neu oimaging A chi e Toolki
(XNAT) [45], one o he oldes and mo e s able open-sou ce da a eposi o ies a ailable. I is de eloped by
he Neu oin o ma ics Resea ch G oup wi h he pu pose o add essing and acili a ing da a managemen
challenges in Neu oimaging s udies. I consis s o an image eposi o y o s o e aw and pos -p ocessed
images, a da abase o s o e me ada a and non-imaging measu es and use in e ace ools o accessing,
sea ching and explo ing he da a. Use in e ace ools include a secu e web applica ion, command line
ools and desk op applica ions o o ganizing and uploading local da a. I suppo s ull DICOM wo k low
o all common imaging modali ies and an ex ensible sys em o modeling and s o ing clinical, beha io al,
and o he non-imaging da a. XNAT has au oma ed ools o cap u e da a om mul iple sou ces and keeps
hem in a secu e eposi o y and dis ibu es he da a o au ho ized use s. XNAT elies hea ily on XML and
XML Schema [53]. XNAT uses XML Schema De ini ion (XSD) o de ine da a ypes and o gene a e cus om
componen s, g aphical and local con en o i s P esen a ion, Applica ion and Da a ie s. Also, XML is
used o secu i y, inpu alida ion and que ies. The e o e, XNAT p o ides an ad hoc wo k low o
36
Neu oimaging da a acquisi ion and sani izing, ha consis s o au oma ed da a acquisi ion di ec ly in he
scanne and/o ia upload ollowed by a s ic quali y con ol p ocedu e, whe e we can include da a
anonymiza ion. Howe e , XNAT also has some p oblems like ine icien me ada a s o age. The XNAT da a-
model can be esumed in h ee main da a ypes: Expe imen s, Subjec s and P ojec s. Imaging da a om
he scanne s en e he wo k low using mechanisms like he Digi al Imaging and Communica ions in
Medicine (DICOM), Secu e File T ans e P o ocol (SFTP), o po able ha d media. Non-imaging da a such
as clinical assessmen s, subjec demog aphics and gene ic measu es a e passed ia web-based o ms,
cs uploads o XML (eX ensible Ma kup Language) [54]. XNAT is used by ins i u ions and da a sha ing
p ojec s a ound he wo ld including he Biomedical In o ma ics Resea ch Ne wo k (BIRN), he Na ional
Alliance o Medical Imaging Compu ing (NA-MIC) and he In o ma ics o In eg a ing Biology & he
Bedside (I2B2) [55].
Figu e 3.5 - XNAT a chi ec u e [56]
A da a model is an abs ac model ha o ganizes ce ain elemen s and s anda dizes how hey can ela e
o each o he . I explici ly de e mines he s uc u e o da a and o ha eason i is some imes e e ed o
as a da a s uc u e. XNAT se e is a pla o m buil o imaging esea ch pu poses, his unc ions can
BUILDING THE RESEARCH PLATFORM – XNAT@DI
43
Pe o mance
The main poin s ela ed wi h pe o mance ake in o accoun ha he wo mos common ac ions a e da a
upload and da a download. These wo ac ions can ha e a huge impac on he se e since i is necessa y
a pe sis en connec ion. XNAT@DI p io i izes low impac on ne wo k o e speed and la ency. As he se e
is on
edu oam
(Educa ion Roaming) ne wo k da a upload is bandwid h limi ed.
Faul Tole ance
To p e en da a loss, iles physically s o ed on he se e a e egula ly backed up o seconda y disks. I
he se e connec ion is no a ailable o da a upload ails, uploading is e-a emp ed mul iple imes wi h
an exponen ially inc easing delay [59].
Taking in o accoun all o hese discussed opics and wha was discussed p e iously, he se e
con igu a ion whe e he pla o m is hos ed has:
• OS: Ubun u 14.04 LTS (64 bi ) as he ope a i e sys em
• CPU: In el Xeon E5-1650
• RAM: 64Gb
• Seconda y Memo y: 2 Disks o 2TB and 1 disk o 512Gb
• GPU: NVIDIA P6000
o Cuda Pa allel-P ocessing Colo s 3840
o GPU 24 GB GDDR5X
o FP32 Pe o mance 12 TFLOPS
4.2
SHARING DATA BETWEEN INSTANCES
Sha ing da a be ween ins ances is c i ical in his kind o pla o m. I ’s e y common o ha e mul iple
depa men s o wo k on mul iple p ojec s ha can some imes sha e da a be ween hem. In XNAT he e
a e wo main ways o connec wo ins ances: using he XNAT P ojec Sync plugin o he Join
Anonymiza ion and A chi e T ansmission (JAAT).
The XNAT P ojec Sync plugin is mo e sui able o sha e da a be ween esea ch ins i u ions. This me hod
es ablishes a pe sis en da a sync o all o selec ed da a be ween XNAT ins ances, which allows he da a-

BUILDING THE RESEARCH PLATFORM – XNAT@DI
44
owning esea che o upload he da a jus once and gua an ees ha he da a is au oma ically uploaded
o he o he ins ances.
The JAAT me hod was de eloped o allow he mig a ion o la ge amoun s o DICOM da a. This me hod is
highly sui able o la ge amoun s o da a. Fo example, o ans e sessions be ween he hospi al whe e
he MRI machine is and he XNAT se e .
Figu e 4.2 - XNAT JAAT me hod explained
4.3
CONTAINERIZATION AND PLATFORM DEPLOY
As men ioned on he p e ious chap e s all se ices unning on he se e a e on op o Docke con aine s.
A Docke con aine is a p ocess/se ice ha uns di ec ly on he se e , sligh ly di e en han a egula
p ocess because he Docke daemon along wi h he Linux ke nel ensu es i uns in o al isola ion, which
allows o ha e mul iple se ices unning a he same ime on a single compu e , wi hou in e e ing wi ch
BUILDING THE RESEARCH PLATFORM – XNAT@DI
45
each o he [59]. The e we e al eady mul iple con aine s deployed on he se e , all managed wi h
Po aine [52]. XNAT would be only “one mo e se ice unning”, al hough o un i wi h he needed
ea u es (specially edundancy and scalabili y) some changes had o be made. As discussed on he
p e ious chap e s i is necessa y o un mul iple se ices a he same ime, including Tomca , Pos g es,
NGINX and, obli iously, he main XNAT con aine . To do his, we need a way o manage and deploy a
he same ime all hese con aine s. This can be made wi h Docke Compose [60], which allows o manage
hese ice e sions, he connec ions be ween con aine s, and e en which po s o expose in each
con aine . Compose is a ool buil -in di ec ly on he Docke engine o de ine and un mul iple-con aine
en i onmen s in a single ne wo k. To use he ool, an YAML [61] ile wi h he con aine s and se ices
speci ica ion was c ea ed. The YAML ile below boo s aps he deploymen o h ee con aine s:
BUILDING THE RESEARCH PLATFORM – XNAT@DI
46
1. e sion: '1'
2. se ices:
3. xna -web:
4. build:
5. con ex : ./xna
6. a gs:
7. XNAT_VER: '1.7.4.1'
8. SMTP_ENABLED: ' alse'
9. SMTP_HOSTNAME: ake. ake
10. SMTP_PORT:
11. AMTP_AUTH:
12. SMTP_USERNAME:
13. SMTP_PASSWORD:
14. XNAT_DATASOURCE_DRIVER: 'o g.pos g esql.D i e '
15. XNAT_DATASOURCE_URL: 'jdbc:pos g esql://xna -db/xna '
16. XNAT_DATASOURCE_USERNAME: 'xna '
17. XNAT_DATASOURCE_PASSWORD: 'xna '
18. XNAT_HIBERNATE_DIALECT: 'o g.hibe na e.dialec .Pos g eSQL9Dialec '
19. TOMCAT_XNAT_FOLDER: XNAT@DI
20. XNAT_ROOT: /da a/xna
21. XNAT_HOME: /da a/xna /home
22. po s:
23. - "8081:8080"
24. - "8000:8000"
25. - "8104:8104"
26. expose:
27. - "8080"
28. - "8104"
29. olumes:
30. - ./xna -da a/home/logs:/da a/xna /home/logs
31. - ./xna -da a/home/plugins:/da a/xna /home/plugins
32. - ./xna -da a/a chi e:/da a/xna /a chi e
33. - ./xna -da a/cache:/da a/xna /cache
34. - ./xna -da a/p ea chi e:/da a/xna /p ea chi e
35. - ./xna -da a/ p:/da a/xna / p
36. - ./xna -da a/build:/da a/xna /build
37. - ./xna -da a/pipeline:/da a/xna /pipeline
38. - / a / un/docke .sock:/ a / un/docke .sock
39. depends_on:
40. - xna -db
41. en i onmen :
42. - CATALINA_OPTS=-Xms128m -Xmx1024m -Dxna .home=/da a/xna /home -
agen lib:jdwp= anspo =d _socke ,se e =y,suspend=n,add ess=8000
43. - XNAT_HOME=/da a/xna /home
44.
45. xna -db:
46. build: ./pos g es
47. expose:
48. - "5432"
49. olumes:
50. - ./pos g es-da a:/ a /lib/pos g esql/da a
51.
52. xna -nginx:
53. build: ./nginx
54. po s:
55. - "80:80"
56. expose:
57. - "80"
58. links:
59. - xna -web
P og am 4.1 - Docke Compose con igu a ion ile
BUILDING THE RESEARCH PLATFORM – XNAT@DI
47
Tomca + XNAT
This con aine uses Tomca and XNAT web applica ion. Tomca is an open sou ce implemen a ion o he
Ja a Se le , Ja aSe e Pages, Ja a Exp ession Language and Ja a WebSocke echnologies [62] and is
used as p oxy o he XNAT con aine . The deploymen o hese applica ions also includes se ing sys em
esou ces and se ing he se ice use and pe missions so ha he Tomca p ocess can ead and w i e
om he XNAT a chi e and wo king olde s. As men ioned be o e, he se e has a main disk, which is a
Solid-S a e Disk, and addi ional Ha d Disk D i es (HDD). Taking in conside a ion ha :
1. SSDs a e much as e on ead and w i e ope a ions bu HDDs a e cheape ;
2. The applica ion iles a e accessed much mo e han he Medical Imaging iles.
All he ope a i e sys ems (OS) and Docke con aine s un on op o he SSD, bu iles a e s o ed in he
HDDs disks. This ensu es ha he pla o m is as , and he s o age is cheap o main ain. To do his some
symbolic links om inside he con aine o he HDD we e c ea ed:
This ensu es ha he XNAT con aine w i es all da a on he seconda y disks, bu all he necessa y olde s
emain in he con aine . The HDD disks con aining he olde s a e backed up o o he HDD disks a ached
o he se e . Also, as discussed on he p e ious chap e s, i was necessa y o c ea e a DICOM SCP
1. - ./xna -da a/home/logs:/da a/xna /home/logs
2. - ./xna -da a/home/plugins:/da a/xna /home/plugins
3. - ./xna -da a/a chi e:/da a/xna /a chi e
4. - ./xna -da a/cache:/da a/xna /cache
5. - ./xna -da a/p ea chi e:/da a/xna /p ea chi e
6. - ./xna -da a/ p:/da a/xna / p
7. - ./xna -da a/build:/da a/xna /build
8. - ./xna -da a/pipeline:/da a/xna /pipeline
9. - / a / un/docke .sock:/ a / un/docke .sock
P og am 4.2 - The symbolic links c ea ed
1. po s:
2. - "8081:8080"
3. - "8000:8000"
4. - "8104:8104"
5. expose:
6. - "8080"
7. - "8104"
P og am 4.4 - The exposed po s P og am 4.3 - The exposed po s on he XNAT con aine
BUILDING THE RESEARCH PLATFORM – XNAT@DI
48
ecei e [63]. Fo his, i was necessa y o expose a new po in which he ecei e will ge all he Medical
Imaging da a.
Pos g eSQL
Pos g eSQL [57] is an open-sou ce objec - ela ional da abase sys em ha ex ends he SQL language
combining ea u es o sa ely s o e and scale da a. Pos g eSQL has ea ned a s ong epu a ion o i s
a chi ec u e, eliabili y, da a in eg i y, obus ea u e se , ex ensibili y and he consis en ly inno a i e
solu ions. Some o he highligh s a e cus om da a ypes, da a in eg i y, scalabili y, concu ency,
pe o mance, secu i y and ex ensibili y. XNAT uses he Pos g eSQL da abase o s o e all o i s pe sis en
da a ha 's no s o ed in iles on he local s o age de ice like subjec in o ma ion o use accoun s.
NGINX
NGINX is a web se e ha uses a non- h eaded, e en -d i en a chi ec u e. I also does o he impo an
hings, like load balancing, HTTP caching o e e se p oxying. A NGINX p oxy con igu a ion equi es a
se e block which con igu es connec ions, and a loca ion block, which poin s o he speci ic Tomca
ins ance wan ed o he web o connec o. The ollowing con igu a ion se s up a lis ene on po 80 (h p).
This is hen associa ed wi h he Tomca unning on po 8080.
Finally, when s a ing o un he mul i-con aine en i onmen wi h
docke -compose up
, se e al di ec o ies
a e c ea ed o sa e he pe sis en da a:
• pos g es-da a - con ains he XNAT da abase;
• xna -da a/a chi e - con ains he XNAT a chi e;
• xna -da a/build - con ains he XNAT build space;
• xna -da a/home/logs - con ains he XNAT logs;
• xna -da a/home/plugins - Ini ially con ains no hing. Se es o cus omize XNAT wi h plugins.
1. po s:
2. - "80:80"
3. expose:
4. - "80"
P og am 4.5 - The exposed po s on he Nginx con aine

BUILDING THE RESEARCH PLATFORM – XNAT@DI
49
4.4
XNAT CONFIGURATIONS
A e building and ins alling he XNAT mul i-con aine en i onmen some cus omiza ions had o be made
di ec ly on he adminis a ion XNAT web in e ace. These changes include:
• Add he DICOM S o e Se ice P o ide – Regis e a new DICOM S o e Se ice P o ide o
XNAT con igu a ions on he p e ious exposed po (8104).
• Change he Session Idle Check In e al o check e e y 10 minu es - This con ols how
o en he sys em checks o see i he e a e incoming DICOM sessions in he p e-a chi e. This
alue is speci ied in milliseconds.
• Enable Si e-wide Anonymiza ion sc ip – An au o anonymiza ion sc ip was added based
on wha is in use a Washing on Uni e si y School o Medicine and Howa d Hughes Medical
Ins i u e.
BUILDING THE RESEARCH PLATFORM – XNAT@DI
50
4.5
XNAT PLUGINS AND PIPELINES
XNAT pla o m has he abili y o use hi d-pa ies se ices o add cus om unc ionali ies o modules. This
is done ei he h ough pipelines which can be au oma ically se o un when, o example, new da a is
uploaded o h ough plugins which add ex a unc ionali ies (gene ally use -in oked). Building he plugins
and pipelines equi es unning a
g adlew
[64] sc ip inside he Docke con aine in which all he compiled
classes and s a ic con en is buil in o a single ou pu ile loca ed in he
build/libs
olde .
A e building he plugin o pipeline, deploying i can be done in h ee s eps:
1. S op Tomca con aine ;
2. Copy JAR ile and build o plugins olde in XNAT iles sys em;
3. S a Tomca con aine again.
Some o he ins alled plugins and pipelines a e:
• Clinical Da a Types [65] - A collec ion o addi ional ypical clinical da a ypes;
• IQ Assessmen [66] – A new IQ da a ype;
• Radiological Assessmen [66] – A new Radiological da a ype, which includes edi and epo
pages;
• m ic on [67] - DICOM and PARREC la o ed pipelines o au o imaging da a con e sion o NIFTI
o ma ;
• Quali y Assessmen P o ocol [68] - QA analysis on unc ional/s uc u al MRI da a;
• m ic ogl [69] - New gene a ion o DICOM- o-NIFTI o ma con e sion pipeline (using dcm2niix);
• DTI-p ep ocessing [70] - Compu e p ep ocessing co ec ions on MRI DTI scans;
• MRI bias ield co ec ion [71] - Co ec ing in ensi y non-uni o mi y (i.e. bias ields);
• MRI ana omical de ace [72] - Au oma ed acial ai s emo al (de acing) o ana omical scan
da a;
• Image Viewe [73]– Neu oimage iewing module buil wi h XTK [74] and Google Closu e [75]
(Fig. 4.3).
BUILDING THE RESEARCH PLATFORM – XNAT@DI
51
Figu e 4.3 - Image Viewe buil -in he XNAT@DI pla o m
4.6
ORGANIZING AND UPLOADING EXISTING DATA
One o he p e-de ined equi emen s was ha exis en da a mus be uploaded o he pla o m. As esea ch
and da a collec ing had al eady been occu ing o many yea s a ICVS and Algo i mi, he e was he need
o ind a way o co ec ly make his da a a ailable h ough XNAT@DI.
The way esea che s ypically sha e da a be o e he he e-desc ibed pla o m was a ailable in ol ed ha ing
an CSV ile wi h subjec in o ma ion like age, gende and hen o ganize he scans in olde s wi h he
co esponden subjec ID. Thus, om his we concluded ha o upload his da a o XNAT@DI, he wo k low
would ha e o ha e wo s eps:
1. Upload subjec da a o he pla o m om he CSV ile;
2. Upload and labeling all he sessions scans and images.
Howe e , some o his CSV iles a e cus om o each p ojec since hey ha e cus om a iables o en imes
ela ed wi h he esea ch ield o he p ojec and no included on he XNAT gene ic ypes. Hence, o each
p ojec , a cus om a iable se wi h speci ic da a ypes was c ea ed. To include new a iables, an XML ile
has o be modi ied wi h he speci ic p ojec con igu a ion.
BUILDING THE RESEARCH PLATFORM – XNAT@DI
52
A e his, he CSV could be uploaded hough he pla o m buil -in uploade . A e c ea ing he subjec s,
all he session scans need o be uploaded o he speci ic subjec . This was done h ough a Py hon sc ip
ha maps all he CSV con en and uploads he co esponding images wi h he co ec label ha iden i ies
he scan.
1. <FieldMapping da a- ype="xna :subjec Da a" c ea e-
da e="Tue Jan 05 03:35:32 PDT 2019" i le="B ainConnec " ID="1294343251432">
2. < ield>xna :subjec Da a/ID</ ield>
3. < ield>xna :subjec Da a/ a he Age</ ield>
4. < ield>xna :subjec Da a/mo he Age</ ield>
5. </FieldMapping>
P og am 4.6 - New ields added o he XML p ojec subjec s con igu a ion ile
1. wi h open('da a.cs ', ' ') as _ ilehandle :
2. cs _ ile_ eade = cs .Dic Reade (_ ilehandle )
3. o ow in cs _ ile_ eade :
4. subjec = ow['Codigo_SW']
5. p in subjec
6. expe imen = subjec + "_CORRELATION"
7. scan = "1"
8. ilepa h = "/AllDa a/" +
9. subjec +
10. "_ nc_ ol_s ime_mc _be _mni_co eg_denoised_wSc ubbing.nii.gz"
11. ile = p ojec .subjec (subjec ).expe imen (expe imen ).scan(
12. scan). esou ce("CORRELATION"). ile(os.pa h.basename( ilepa h))
13. p in ile
14. i no ( ile.exis s()) and os.pa h.is ile( ilepa h):
15. ile.pu ( ilepa h, o ma =" ", con en =" ", ags=" ")
16. p ojec .subjec (subjec ).expe imen (
17. expe imen ).scan(scan).a s.se (" ype", "CORRELATION")
P og am 4.7 - Py hon sc ip o upload all sessions o XNAT@DI
XNATUM – A PYTHON API FOR XNAT
59
Figu e 5.5 - XNAT REST model
Fo example, wi h he REST API i is possible o look o all emale subjec s ha a e 23 yea s o age wi hin
a p ojec o ha e a speci ic answe o an assessmen . Using a me ge o he XNATPy package, which
implies a lo o knowledge in Py hon and APIs o use, wi h he Py hon eques module, XNATUM is able
o e ie e all he assessmen s om a subjec in a single s a emen whe eas he REST API om XNAT
would equi e mul iple HTTP calls.
XNATUM uses Py hon classes o c ea e an ini ial se e con igu a ion, which pe sis s o be used in mul iple
se e ope a ions. The connec ion holds he login in o ma ion, he se e in o ma ion and a session. I
will also send a ping e e y 14 minu es o keep he connec ion ali e, e en i no ope a ion is being execu ed.
XNATUM also allows sel -closing sessions ha can also be used using he Py hon con ex ope a o , o
emo e he possibili y o un o eseen e o s. As soon as he session doesn’ exis anymo e, o example
because an e o /excep ion was h own away, he session is au oma ically disconnec ed.
1. impo xna um
2.
3. session = xna .connec ('h p://mi box.di.uminho.p /XNAT@DI/', use ='xna um', pass
wo d='sec e ')
P og am 5.1 - Example o c ea ing a connec ion o a XNAT ins ance

XNATUM – A PYTHON API FOR XNAT
60
When a session is es ablished and main ained wi hin he scope, i is ai ly easy o explo e he da a on he
XNAT se e . The da a s uc u e o XNAT is mimicked as Py hon objec s. The connec ion gi es access o
a lis ing o all p ojec s, subjec s, and expe imen s on he se e . All le els on he XNAT se e : p ojec s,
subjec s, expe imen s, scans, esou ces, iles can be b owsed.
Al hough, he e a e si ua ions on which he e is he need o looping o e da a. To do his, XNATLis ing
objec s can be used as Py hon dic iona ies.
I he e is he need o download da a, his can be done using a helpe unc ion o ei he download i o a
a ge di ec o y o o he cu en di ec o y. This will c ea e a da a s uc u e simila o ha o XNAT on he
local disk.
To add new da a in o he XNAT se e , i is possible using he XNAT REST impo se ice wi h an HTTP
POST eques . I allows he upload o a zip ile con aining new expe imen s and XNAT will au oma ically
s o e i in he co ec place. Al hough, i is dange ous o add da a s aigh in o he a chi e due o he lack
1. impo xna um
2.
3. wi h xna .connec ('h p://my.xna .se e ') as session:
4. p in (session.p ojec s)
P og am 5.2 - C ea ing a connec ion o a XNAT ins ance using he Py hon con ex ope a o
1. p ojec = sel .session.p ojec s[lp ojec ]
2.
3. o subjec in p ojec .subjec s. alues():
4. o expe imen in subjec .expe imen s. alues():
5. allexpe imen s.append(expe imen )
6. e u n allexpe imen s
1. y:
2. sel .__p ea c_session = sel .session.se ices.impo _(
3. zip name,
4. o e w i e="append",
5. p ojec =p ojec ,
6. subjec =subjec ,
7. expe imen =subjec + "_" + session,
8. des ina ion=des ina ion,
9. igge _pipelines=False,
10. )
11. excep :
12. p in ("Unexpec ed e o du ing XNAT impo :")
13. p in (sys.exc_in o())
P og am 5.4 - XNATUM session upload unc ion
P og am 5.3 - Ge ing all expe imen s o a XNAT p ojec
XNATUM – A PYTHON API FOR XNAT
61
o e iewing. So, he sugges ed wo k low o he XNAT@DI pla o m is o use he des ina ion a gumen as
‘p ea chi e’ in he unc ion o send sessions o he se e . This will make he sessions being uploaded o
he p ojec , bu also makes he p e-a chi e be e iewed by he p ima y esea che be o e being a ailable
o esea ch pu poses.
5.3
USE-CASE EXAMPLES
XNATUM is a powe ul ye easy o use Py hon package. As he ini ial goal was o c ea e a new package
in which esea che s could ely one, e en hose no amilia wi h Py hon p og amming. Fo his eason,
some wo king examples we e c ea ed o se e as he en y poin o use he package.
Download p ojec sessions
This example uses he
download_p ojec _sessions
unc ion o download all images wi hin a p ojec o he
di ec o y whe e he Py hon sc ip is being execu ed.
Download p ojec sessions o di ec o y
This example uses he
download_p ojec _sessions_ o_di ec o y
unc ion o download all images wi hin
a p ojec o he di pa h a gumen olde . I he olde does no exis XNATUM c ea es he olde .
Download subjec sessions
This example uses he
download_subjec _sessions
unc ion o download all images wi hin a p ojec om
a speci ic subjec o he di ec o y whe e he Py hon sc ip is being execu ed.
1. xna = Xna (a gs.se e , a gs.use name, a gs.passwo d)
2. xna _sessions = xna .download_p ojec _sessions(a gs.p ojec )
P og am 5.5 - Download p ojec sessions h ough XNATUM
1. xna = Xna (a gs.se e , a gs.use name, a gs.passwo d)
2. xna _sessions = xna .download_p ojec _sessions_ o_di ec o y(a gs.p ojec , di pa h
)
P og am 5.6 - Download p ojec sessions o di ec o y h ough XNATUM
XNATUM – A PYTHON API FOR XNAT
62
Download subjec sessions o di ec o y
This example uses he
download_subjec _sessions_ o_di ec o y
unc ion o download all images wi hin
a p ojec om a speci ic subjec o he di pa h a gumen olde . I he olde does no exis XNATUM
c ea es he olde .
Download speci ic subjec session o di ec o y
This example uses he
download_single_subjec _session_ o_di ec o y
unc ion o download all images
wi hin a p ojec o he di pa h a gumen olde . I he olde does no exis XNATUM c ea es he olde .
Ge lis o subjec s
This example uses he
ge _lis _subjec s
unc ion o lis o subjec ha exis on a p ojec .
Ge p ojec sessions
1. xna = Xna (a gs.se e , a gs.use name, a gs.passwo d)
2. xna _sessions = xna .download_subjec _sessions(a gs.p ojec , a gs.subjec )
P og am 5.7 - Download subjec sessions h ough XNATUM
1. xna = Xna (a gs.se e , a gs.use name, a gs.passwo d)
2. xna _sessions = xna .download_subjec _sessions_ o_di ec o y(a gs.p ojec , a gs.su
bjec , di pa h)
P og am 5.8 - Download subjec sessions o di ec o y h ough XNATUM
1. xna = Xna (a gs.se e , a gs.use name, a gs.passwo d)
2. xna _sessions = xna .download_single_subjec _session_ o_di ec o y(a gs.p ojec , a
gs.subjec , a gs.session', di pa h)
P og am 5.9 - Download speci ic subjec session o di ec o y h ough XNATUM
1. xna = Xna (a gs.se e , a gs.use name, a gs.passwo d)
2. p in (xna .ge _lis _subjec s(a gs.p ojec ))
P og am 5.10 - Lis all p ojec subjec s h ough XNATUM
XNATUM – A PYTHON API FOR XNAT
63
This example uses he
ge _p ojec _sessions
unc ion o lis all sessions labels wi hin a p ojec , iden i ying
he co esponden subjec .
Ge subjec in o
This example uses he
ge _subjec _in o
unc ion o ge all subjec me ada a wi hin a p ojec .
All he examples desc ibed he e, and mo e a e a ailable in he XNATUM Gi Hub eposi o y
3
.
Upload session
This example uses he
impo _ esou ce
unc ion o upload a session o a p ojec om a subjec . I s a s
o c ea e he session i em s uc u e, so s i by he o ma and hen upload de esou ces iles o he XNAT
se e .
3
XNATUM examples olde on he Gi Hub eposi o y - h ps://gi hub.com/ gllm/xna um/ ee/mas e /examples
1. xna = Xna (a gs.se e , a gs.use name, a gs.passwo d)
2. xna _sessions = xna .ge _p ojec _sessions(a gs.p ojec )
P og am 5.11 - Ge p ojec sessions labels h ough XNATUM
1. xna = Xna (a gs.se e , a gs.use name, a gs.passwo d)
2. p in (xna .ge _subjec _in o(a gs.p ojec , a gs.subjec ))
P og am 5.12 - Ge all in o ma ion om a subjec wi hin a p ojec h ough XNATUM
1. # C ea ing connec ion and load expe imen objec
2. xna = Xna (a gs.se e , a gs.use name, a gs.passwo d)
3. i em = xna .session.p ojec s[a gs.p ojec ].subjec s[a gs.subjec ].expe imen s[0]
4.
5. # Sending ile
6. iles = so ed( glob.glob( '{}/sessions/{}_*.*'. o ma ( a gs.inpu , a gs.subjec
) ) )
7. xna .impo _ esou ce(i em, 'FILE', iles)
8. p in ('File impo ed.')
P og am 5.13 - Upload a session a subjec h ough XNATUM
XNATUM – A PYTHON API FOR XNAT
64
5.4
USING A VIRTUAL RAM DRIVE
Talking speci ically abou he wo k low a Algo i mi and ICVS, XNAT@DI pla o m has a unique wo k low
and se up. Resea che s wo k and de elop sc ip s in Jupy e no ebooks hos ed on he same se e in
which XNAT is unning, he only di e ence is ha hey a e unning on di e en con aine s.
This allows ue collabo a ion be ween all esea che s and he XNAT@DI. The goal he e is o allow o no
eplica e Medical Imaging da a ac oss di e en Jupy e no ebooks and con aine s bu o hos medical
images in XNAT and use hose images di ec ly in all Jupy e no ebooks.
The i s a emp o do his elied on he ac ha XNAT REST API exposes an API which includes he
en i e DICOM ile, he Medical Imaging aw o ma , wi h a ious DICOM da a elemen s.
A DICOM da a elemen , o a ibu e, is composed o he ollowing pa s [63] :
• a ag ha iden i ies he a ibu e, usually in he o ma (XXXX, XXXX) wi h hexadecimal numbe s,
and may be di ided u he in o DICOM G oup Numbe and DICOM Elemen Numbe ;
• a DICOM Value Rep esen a ion (VR) ha desc ibes he da a ype and o ma o he a ibu e alue.
Fo example, he ag (0018,8151) co esponds o he X-Ray Tube Cu en in µA and he ag (0010,0010)
co esponds o he Pa ien ’s Name. The e is also a ag o he speci ic scan image. The Ni y ile o ma ,
a ile o ma used a e he medical image is p ocessed (can be compa ed o JPEG) and wo ks in he
same way as DICOM; wi h di e en ags and a ibu es as all images a e anonymized and combined by
de aul . Taking his in o accoun , he plan was o ead and use all he images om he DICOM and Ni y
iles by selec ing he co esponden ag ield om he XNAT REST API. As explained be o e, he API c ea es
a JSON endpoin o each session inside a "p ojec and subjec " ile and hese endpoin s con ain all
DICOM ags and a ibu es. Howe e , XNAT unca es he a ibu es om he DICOM dump a ailable on
he API o unde 64 cha ac e s. So, i ’s only possible o ead he i s 64 cha ac e s o he “Pixel Da a”
a ibu e and no he comple e images.
Thus, he nex app oach was o use RAM D i e which is a block o andom-access memo y ha so wa e
can use as i he memo y we e a disk d i e. Py hon has an API o gene a e empo a y iles and di ec o ies
ha uses his OS buil -in unc ionali y. The module c ea es empo a y iles and di ec o ies and p o ides
au oma ic cleanup which can be used as con ex manage s. Howe e , as his is a speci ic app oach o
ou pla o m and XNATUM is a global package ha can be used in any XNAT ins alla ion, his is no

XNATUM – A PYTHON API FOR XNAT
65
included in he co e unc ionali y. The e o e, his is added as a wo king example on he XNATUM
eposi o y making use o he unc ionali y p e iously desc ibed o download images by indica ing he
des ina ion olde . The example uses he “ emp ile.mkd emp” [85] Py hon unc ionali y ha c ea es
empo a y di ec o y in he mos secu e manne possible wi h no ace condi ions in he di ec o y c ea ion.
The di ec o y is eadable, w i able, sea chable only by he c ea ing use ID and is a ailable as long as he
use wan s i o be du ing he p og am execu ion. This Py hon unc ionali y, when execu ed, e u ns he
empo a y di ec o y pa h and he e o e he unique hing o do on XNATUM is o indica e his pa h o he
co esponden unc ion so as o download he sessions by p ojec , subjec o session label. When he
p og am is execu ed, all he sessions a e downloaded di ec ly om XNATUM o he empo a y di ec o y
and can be used a e ha in he esea ch sc ip .
5.5
DISCUSSION
His o ically, Neu oimaging esea che s ha e used
ad hoc
p ocedu es o main aining hei analysis and
da a. O e he las ew yea s, he e ha e been se e al a emp s o build sys ems ha manage
Neu oimaging da a. Mo eo e , he abili y o p og amma ically access Neu oimaging da a is becoming
inc easingly impo an o pe o m ba ch analysis and adminis a ion asks, as hese da abases inc ease
in size.
Apa om his, one o he mos widely used Neu oimaging managemen sys ems is XNAT, which includes
many use ul and powe ul ea u es including, as discussed be o e, a REST API. XNATUM p o ides a b idge
be ween he XNAT ins ance used by he esea che o a chi e hei da a and he analysis ool. Combined
wi h Py hon, i can be used as an al e na i e bo h o he XNAT web on end and o a as se o Py hon
ools a ailable in he neu oscience domain. The package ocuses on ease o use, combining REST ul
se ices and XNATPy wi h clea seman ics and helpe ea u es.
Since one o he main ocus o he neu oscience esea ch a UMinho is using Machine Lea ning (ML) and
Deep Lea ning (DL) in he neu oscience/medicine imaging esea ch ield, XNATUM combines a se ies o
ea u es o help esea che s ocuses in he esea ch, abs ac ing om he da a o use. Fo example, he e
is a ea u e on XNATUM o simply ob ain and download all sessions wi hin a XNAT p ojec and
au oma ically label each session as ain da a o es da a.
The e sion 1.0 o XNATUM o Py hon is a ailable publicly since Ma ch 28 o 2019 and i is cu en ly
mainly used by esea che s ha use he XNAT@DI pla o m. As any so wa e in hei ea ly s eps, he e
a e a couple o hings ha can be imp o ed and added o he package. These imp o emen s and new
XNATUM – A PYTHON API FOR XNAT
66
ea u es a e discussed in he nex chap e . The p ima y goal in c ea ing his package was o esea che s
o ha e a simple way o manage da a in he XNAT@DI pla o m, and ha was accomplished since he e
a e al eady esea che s es ing i and/o using i in hei expe imen s. The images bellow lis some o he
s a s om he Pypi eposi o y.
XNATUM – A PYTHON API FOR XNAT
67
Figu e 5.6 - Daily downloads since he package is a ailable a Pypi
Figu e 5.7 - Daily downloads p opo ion be ween di e en e sions o Py hon
Figu e 5.8 - Daily downloads be ween di e en e sions o Py hon
XNATUM – A PYTHON API FOR XNAT
68
Figu e 5.9 - Daily downloads p opo ion be ween di e en e sions o Py hon
Figu e 5.10 - Daily downloads be ween di e en sys em
Figu e 5.11 - Daily downloads p opo ion be ween di e en sys em
REFERENCES
lxx
[26] W. J. Bug, G. A. Ascoli, J. S. G e he, A. Gup a, C. Fennema-No es ine, A. R. Lai d, S. D. La son,
D. Rubin, G. M. Shephe d, J. A. Tu ne , and M. E. Ma one, “The NIFSTD and BIRNLex
ocabula ies: Building comp ehensi e on ologies o neu oscience,”
Neu oin o ma ics
, ol. 6, no.
3, pp. 175–194, Sep. 2008.
[27] W. Raghupa hi Viju Raghupa hi,
An O e iew o Heal h Analy ics
, ol. 04. 2013.
[28] W. Raghupa hi and V. Raghupa hi, “Big da a analy ics in heal hca e: p omise and po en ial,”
Heal.
In . Sci. Sys .
, ol. 2, no. 1, pp. 1–10, 2014.
[29] S. T. C. Wong, D. Tjand a, H. Wang, and W. Shen, “Wo k low-enabled dis ibu ed componen -
based in o ma ion a chi ec u e o digi al medical imaging en e p ises,”
IEEE T ans. In . Technol.
Biomed.
, ol. 7, no. 3, pp. 171–183, 2003.
[30] E. Socie y, “The new EU Gene al Da a P o ec ion Regula ion: wha he adiologis should know,”
Insigh s Imaging
, ol. 8, no. 3, pp. 295–299, 2017.
[31] R. A. Pold ack, “NIH Public Access The u u e o MRI in Cogni i e Neu oscience,” ol. 62, no. 2,
pp. 1216–1220, 2014.
[32] A. Mose , K. Range, and D. M. Yo k, “NIH Public Access,”
Bone
, ol. 23, no. 1, pp. 1–7, 2008.
[33] Na ional Ins i u es o Heal h (NIH), “NIH Launches he Human Connec ome P ojec o Un a el
he B ain’s Connec ions,” 2009. [Online]. A ailable: h ps://www.nih.go /news-e en s/news-
eleases/nih-launches-human-connec ome-p ojec -un a el-b ains-connec ions. [Accessed: 01-
Jan-2019].
[34] L. Geddes, “Human b ain mapped in unp eceden ed de ail,”
Na u e
, Jul. 2016.
[35] D. S. Ma cus, K. A. A chie, T. R. Olsen, and M. Rama a nam, “The open-sou ce neu oimaging
esea ch en e p ise,”
J. Digi . Imaging
, ol. 20, no. SUPPL. 1, pp. 130–138, 2007.
[36] J. Gu neya, T. Olsenb, J. Fla ina, M. Rama a namc, K. A chied, J. Rans o de, R. He icka, L.
Wallacea, J. Clinea, W. Ho ona, and D. S. Ma cusa, “The Washing on Uni e si y Cen al
Neu oimaging Da a A chi e,”
HHS Public Access
, ol. 91, no. 2, pp. 165–171, 2015.
[37] K. Alpe , A. Kogan, T. Pa ish, D. Ma cus, and L. Wang, “The No hwes e n Uni e si y
Neu oimaging Da a A chi e (NUNDA),”
Neu oimage
, ol. 124, pp. 1131–1136, 2016.
[38] T. Doel, D. I. Shaki , R. P a , M. Ae sen, J. Mogg idge, E. Bellon, A. L. Da id, J. Dep es , T.
Ve cau e en, and S. Ou selin, “GIFT-Cloud: A da a sha ing and collabo a ion pla o m o medical
imaging esea ch.,”
Compu . Me hods P og ams Biomed.
, ol. 139, pp. 181–190, Feb. 2017.
[39] D. T. Fe ze and O. C. Wes , “The HIPAA P i acy Rule and P o ec ed Heal h In o ma ion.
Implica ions in Resea ch In ol ing DICOM Image Da abases,”
Acad. Radiol.
, ol. 15, no. 3, pp.

REFERENCES
lxx i
390–395, Ma . 2008.
[40] Ho os, “Ho os P ojec 2015,” 2015. [Online]. A ailable: h ps://ho osp ojec .o g/. [Accessed:
29-May-2019].
[41] “Osi iX DICOM Viewe | The wo ld amous medical imaging iewe ,”
Osi iX
, 2019. [Online].
A ailable: h ps://www.osi ix- iewe .com/. [Accessed: 31-May-2019].
[42] “Medical imaging so wa e: Pla o m o mode n adiology.”
[43] “Bes Medical Imaging So wa e | 2019 Re iews o he Mos Popula Tools & Sys ems.”
[Online]. A ailable: h ps://www.cap e a.com/medical-imaging-so wa e/. [Accessed: 13-Jul-
2019].
[44] “No a Medical Imaging Pla o m.” 2017.
[45] “XNAT.” [Online]. A ailable: h ps://www.xna .o g/. [Accessed: 24-May-2019].
[46] O acle, “O acle VM Vi ualBox,” pp. 1–362, 2012.
[47] VMwa e, “VMwa e,” 2019. [Online]. A ailable: h ps://www. mwa e.com/. [Accessed: 27-Jul-
2019].
[48] G. Kia , K. J. Go golewski, D. Kleissas, W. G. Roncal, B. Li , B. Wandell, R. A. Pold ack, M. Wiene ,
R. J. Vogels ein, R. Bu ns, and J. T. Vogels ein, “Science in he cloud (SIC): A use case in MRI
connec omics,”
Gigascience
, ol. 6, no. 5, pp. 1–10, May 2017.
[49] Docke Inc., “Wha is a Con aine ?: A s anda dized uni o so wa e,”
Docke .Com
, 2019. [Online].
A ailable: h ps://www.docke .com/ esou ces/wha -con aine . [Accessed: 30-Jun-2019].
[50] “Ki ema ic | Docke Documen a ion.” [Online]. A ailable: h ps://docs.docke .com/ki ema ic/.
[Accessed: 02-Jul-2019].
[51] “shipya d p ojec .” [Online]. A ailable: h p://shipya d-p ojec .com/. [Accessed: 02-Jul-2019].
[52] Po aine , “Po aine Managemen , Docke Use In e ace, Con aine So wa e - Auckland,
Singapo e, San F ancisco | Eme ging Technology Pa ne s,” 2019. [Online]. A ailable:
h ps://www.po aine .io/. [Accessed: 31-May-2019].
[53] M. Böhnlein and A. U. Vom Ende, “XML - Ex ensible Ma kup Language,”
Wi scha sin o ma ik
,
ol. 41, no. 3, pp. 274–276, 1999.
[54] F. Falaha i, E. Wes man, and A. Simmons, “Mul i a ia e da a analysis and machine lea ning in
Alzheime ’s disease wi h a ocus on s uc u al magne ic esonance imaging,”
J. Alzheime ’s Dis.
,
ol. 41, no. 3, pp. 685–708, 2014.
[55] Y. Huo, J. Blabe , S. M. Damon, B. D. Boyd, S. Bao, P. Pa a haneni, C. B. Nogue a, S. Chagan i,
V. Na h, J. M. G ee , I. Lyu, W. R. F ench, A. T. New on, B. P. Roge s, and B. A. Landman,
REFERENCES
lxx ii
“Towa ds Po able La ge-Scale Image P ocessing wi h High-Pe o mance Compu ing,”
J. Digi .
Imaging
, ol. 31, no. 3, pp. 304–314, 2018.
[56] D. Ma cus, “XNAT Clinical Analy ics Pla o m ( XCAP ): An In o ma ics Pla o m o Machine
Lea ning Applica ions in Clinical Resea ch.”
[57] “Pos g eSQL: The wo ld’s mos ad anced open sou ce da abase,” 2013. [Online]. A ailable:
h ps://www.pos g esql.o g/. [Accessed: 01-Jun-2019].
[58] B. Shaik, “Simple Pos g eSQL Blog: Pos g eSQL A chi ec u e,” 2013. [Online]. A ailable:
h p://bajis-pos g es.blogspo .com/2013/10/pos g esql-a chi ec u e.h ml. [Accessed: 30-Jun-
2019].
[59] T. Bui, “Analysis o Docke Secu i y,” 2015.
[60] Docke Inc., “O e iew o Docke Compose,”
docs.docke .com
, pp. 1–6, 2015.
[61] C. E ans, “The O icial {YAML} Web Si e,”
{YAML:} {YAML} Ain’ Ma kup Language
, 2011. [Online].
A ailable: h p://yaml.o g/. [Accessed: 06-Jun-2019].
[62] The Apache So wa e Founda ion, “Apache Tomca ,” 2011. [Online]. A ailable:
h p:// omca .apache.o g/. [Accessed: 06-Jun-2019].
[63] Mildenbe ge Pe e , M. Eichelbe g, and E. Ma in, “In oduc ion o he DICOM s anda d,” p. 8,
2001.
[64] “The G adle W appe ,” 2018. [Online]. A ailable:
h ps://docs.g adle.o g/cu en /use guide/g adle_w appe .h ml. [Accessed: 06-Jun-2019].
[65] “Clinical Da a Types - XNAT Ma ke placeXNAT Ma ke place.” [Online]. A ailable:
h ps://ma ke place.xna .o g/plugin/clinical-da a- ypes/. [Accessed: 06-Jun-2019].
[66] “IQ Assessmen - XNAT Ma ke placeXNAT Ma ke place.” [Online]. A ailable:
h ps://ma ke place.xna .o g/plugin/ adiological-assessmen /. [Accessed: 06-Jun-2019].
[67] “m ic on XNAT Pipeline.” [Online]. A ailable: h ps://gi hub.com/jhugue n/xna -
pipelines/ ee/mas e /m ic on.
[68] “QAP.” [Online]. A ailable: h ps://gi hub.com/jhugue n/xna -
pipelines/blob/mas e /QAP/README.md.
[69] “m ic ogl.” [Online]. A ailable: h ps://gi hub.com/jhugue n/xna -
pipelines/ ee/mas e /m ic ogl.
[70] “DTI-p ep ocessing.” [Online]. A ailable: h ps://gi hub.com/jhugue n/xna -
pipelines/ ee/mas e /d i_p ep ocessing.
[71] “MRI bias ield co ec ion.” [Online]. A ailable: h ps://gi hub.com/jhugue n/xna -
REFERENCES
lxx iii
pipelines/ ee/mas e /bias_co ec ion.
[72] “MRI ana omical de ace .” [Online]. A ailable: h ps://gi hub.com/jhugue n/xna -
pipelines/ ee/mas e /m i_ana _de ace.
[73] “xna de / xna -image- iewe -plugin.” [Online]. A ailable: h ps://bi bucke .o g/xna de /xna -
image- iewe -plugin/s c/mas e /. [Accessed: 07-Jun-2019].
[74] “XTK.” [Online]. A ailable: h ps://gi hub.com/x k/X# eadme.
[75] Google, “Closu e Tools - Google De elope s,” 2013. [Online]. A ailable:
h ps://de elope s.google.com/closu e/. [Accessed: 07-Jun-2019].
[76] NRG Lab, “Who Uses XNAT?” [Online]. A ailable: h ps://www.xna .o g/abou /xna -
implemen a ions.php. [Accessed: 11-Jun-2019].
[77] S. D. La son and M. E. Ma one, “On ologies o neu oscience: Wha a e hey and wha a e hey
good o ?,”
F on . Neu osci.
, ol. 3, no. MAY, pp. 60–67, 2009.
[78] E. Mulle , J. A. Bedna , M. Diesmann, M. Gewal ig, M. Hines, and A. P. Da ison,
Py hon in
Neu oscience
. 2015.
[79] SciPy de elope s, “SciPy.o g — SciPy.o g,” 2019. [Online]. A ailable: h ps://www.scipy.o g/.
[Accessed: 12-Jun-2019].
[80] A. J. Na han and A. Scobell, “How China sees Ame ica,”
Fo eign A ai s
, ol. 91, no. 5. p. All,
2012.
[81] “PEP 249 -- Py hon Da abase API Speci ica ion 2.0 | Py hon.o g.” [Online]. A ailable:
h ps://www.py hon.o g/de /peps/pep-0249/. [Accessed: 12-Jun-2019].
[82] R. T. Fielding, “A chi ec u al S yles and he Design o Ne wo k-based So wa e A chi ec u es,”
Uni e si y o Cali o nia, 2000.
[83] “Pypi XNAT sea ch.” [Online]. A ailable: h ps://pypi.o g/sea ch/?q=xna .
[84] Ma hWo ks, “MATLAB - Ma hWo ks - MATLAB & Simulink,”
Ma lab
, 2019. [Online]. A ailable:
h ps://www.ma hwo ks.com/p oduc s/ma lab.h ml. [Accessed: 13-Jun-2019].
[85] Py hon.o g, “ emp ile: Gene a e empo a y iles and di ec o ies,” 2019. [Online]. A ailable:
h ps://docs.py hon.o g/2/lib a y/ emp ile.h ml. [Accessed: 02-Jul-2019].
APPENDICES
APPENDICES
APPENDICES
lxxx
A - XNAT USER GUIDE
A.1 INTRODUCTION
XNAT@DI is an ins ance o he popula Neu oimaging a chi e sys em XNAT (xna .o g) and a chi es all
imaging da a acqui ed a he ICVS and Hospi al o B aga and used in he Depa men o In o ma ics.
This guide co e s he mos commonly pe o med ope a ions wi h he XNAT@DI web po al. Fo mo e
de ailed in o ma ion on how o use he po al please isi he XNAT wiki, o i you can' ind he answe
o you ques ion, please email a XNAT adminis a o .
Figu e A.1 - UMinho XNAT pla o m
A.2 SITE ACCESS
Regis a ion
All UMinho s a and s uden s can egis e o use XNAT@DI. Fo his, hey need o email P o . Vic o Al es
a [email p o ec ed] asking o an accoun in he pla o m. S a and s uden s may use hei
ins i u ional email. A e his i is necessa y o be g an ed access o a p ojec . Non-UMinho use s should
also email P o . Vic o Al es wi h hei p e e ed use name and con ac de ails, including he con ac

APPENDICES
lxxxi
de ails o an in es iga o o he p ojec (s) hey wish o be g an ed access o in he commen s sec ions.
XNAT@DI adminis a ions will enable he accoun a e con i ming wi h he ele an p ojec in es iga o (s).
Login
To access o he po al, access he add ess he pla o m add ess o copy/pas e his add ess in o a web
b owse add ess ba . Nex inse you login de ails in he login box. I any p oblem occu s, y o dele e
you cookies. I he p oblem pe sis s, please con ac P o . Vic o Al es. Login in o ma ion is displayed in
he op igh -hand co ne o all pages on he si e abo e he gene al sea ch ba . To logou simply click he
“logou ” link. Login sessions will expi e a e 15 mins o inac i i y, wi h he emaining ime displayed nex
o he use name o he logged-in use . I you wish o enew you session click he “ enew” link also in he
op igh -hand side o he sc een, and he 15 min coun e will s a again.
P ojec Membe ship
Access o imaging da a is con olled ia p ojec membe ship. Membe s a e added o a p ojec by
adminis a o s and/o p ojec in es iga o .
To ask he adminis a o s o add a new membe o a p ojec , he in es iga o mus email
[email p o ec ed] wi h he name and email add ess o he pe son o be added and he p ojec name
ha he access is o be gi en o.
I you a e a p ojec owne , you can use he "Access" ab o he p ojec page. Access is g an ed by on he
Access ab o he p ojec page, new membe s a e added by selec ing he a ailable use s om he lis .
Membe s can be emo ed om a p ojec by a p ojec owne o MBI adminis a o s by clicking he
“Remo e” bu on nex o hei names.
A.3 NAVIGATION
XNAT@DI ID Con en ions
All p ojec s acqui ed and make a ailable a XNAT@DI mus con ain a name. Gene ally, his name is he
p ojec s name.
APPENDICES
lxxxii
Home Sc een
The home sc een p o ides se e al con enien me hods o na iga e he imaging da a ha is
accessible o he use . A lis o ecen ly accessed p ojec s by he use will appea in he
bo om le co ne o he sc een and a lis o ecen ly uploaded scanning sessions on he igh . Clicking
on hese en ies will na iga e he use o he co esponding p ojec /session. To na ow down he lis o
p ojec s, subjec s o sessions, sea ch c i e ia can be en e ed in o he sea ch box in he middle o he
page. To e u n o he home sc een om any page, selec he XNAT logo in he op le co ne .
Command Ribbon
The command ibbon is accessible om all pages wi hin he si e. Clicking on he “B owse” opens a
hie a chical menu ha p o ides links o all p ojec s, subjec s and imaging sessions ha a e accessible
o he use .
A.4 DATA ACCESS
I is he esea che 's esponsibili y o e i y he quali y o he imaging da a uploaded o XNAT. Gene ally,
he da a uploaded o he XNAT ins ance goes o he "P e-a chi e". The esea ch mus e iew and a chi e
he da a himsel , gi ing a p ojec o i .
Snapsho s and DICOM Heade s
A e an imaging session is uploaded o XNAT, image snapsho s a e gene a ed o
ecognized image ypes and can be displayed along wi h o he basic me a-da a by oggling
he +/- bu on nex o he scan name on he session page. De ailed me ada a can be accessed om he
DICOM heade s ia he link “View DICOM Heade s”.
Image Viewe
XNATUM p o ides a con enien buil -in iewe o isual inspec ion o imaging da a, which
can be accessed by he “ iew images” link on he “Ac ions” menu on he session page.
APPENDICES
lxxxiii
A.5 MANAGE SESSIONS DATA WITH HOROS
To manage sessions h ough Ho os o o he DICOM medical image iewe you jus o ha e o add XNAT
as a new image loca ion on he co esponden DICOM po .
Figu e A..2 - HOROS XNAT loca ion con igu a ion
APPENDICES
lxxxi
B – XNAT DATA MODEL
A da a model is an abs ac model ha o ganizes ce ain elemen s and s anda dizes how hey can ela e
o each o he . I explici ly de e mines he s uc u e o he da a and o ha eason i is some imes e e ed
o as a da a s uc u e.
XNAT se e is a pla o m buil o imaging esea ch pu poses. I s unc ions can a chi e, impo , p ocess
and secu ely dis ibu e he da a. The co e da a-model used in XNAT p e iews he use o he XNAT se e
in a ious p ojec s. The e a e h ee co e da a- ypes in XNAT da a model: P ojec s, Subjec s and
Expe imen s. These h ee da a- ypes ela e wi h each o he .
P ojec s
A p ojec is used o de ine a collec ion o a ce ain da a s o ed in he se e . The p ojec is used o de ine
a secu e s uc u e o da a. Use s can ha e ce ain pe missions o da a wi hin a ce ain p ojec . They can
be Owne s, Membe s o Collabo a o s. This pe mission de ines who can ead, sa e o dele e da a. Fo
example, in a speci ic esea ch s udy we can gi e pe mission o he p incipal esea che o sa e, ead
and dele e da a bu o an in e nal esea che only o ead he da a.
Subjec s
A subjec is anyone who pa icipa es in a s udy. Imaging s udies has adi ionally ocused on human
s udies, howe e , he subjec could be non-human. I exis s in a con ex o a p ojec and i 's "owned" by
he p ojec whe e i was c ea ed on. Also, subjec s can be sha ed wi h o he p ojec s o cap u e
longi udinal da a om a ious s udies.
Expe imen s
An expe imen is an e en by which imaging da a o non-imaging da a is acqui ed. I canno exis ou side
he con ex o a p ojec and can be sha ed in o o he p ojec s.