BACHELOR’S THESIS
Bachelo ’s deg ee in Biomedical Enginee ing
DICOM VIEWER: INTERACTIVE VIEWER OF DICOM MEDICAL
IMAGES
Repo and Annexes
Au ho : Júlia Romagosa Pé ez
Di ec o : Ch is ian Ma a Miquel
Co-Di ec o : Raúl Bení ez Iglesias
Call: June 2021
DICOM Viewe : In e ac i e iewe o DICOM medical images
i
Resum
El cànce de p òs a a és el segon ipus de cànce més comú i la cinquena causa p incipal de mo pe
cànce en homes du an l'úl im any. To i així, en les úl imes dècades la axa de mo ali a ha disminuï
no ablemen degu a, en e al es ac o s, un diagnòs ic p ecoç i a un a millo a en les eines pe po a -
lo a e me.
T adicionalmen en l'es udi d'aques cànce , els adiòlegs anali zen di e en s modali a s d'ima ge de
o ma indi idual, en e elles 3D T2- Weigh ed Imaging, Di usion-Weigh ed Imaging i Pe usion-
Weigh ed Imaging. D'aques a mane a, obem casos en què l'obse ació del umo es eu
comp omesa pe di e en s aspec es i po no se de ec a co ec amen . Pe això, el nos e objec iu és
desen olupa una aplicació accessible pe a qualse ol cen e mèdic o hospi al, sense necessi a
d'ins al·lació d'un so wa e complex, que se eixi d'ajuda pe al diagnòs ic de cànce de p òs a a i d'una
solució a aques es limi acions.
A pa i de la base de dades p opo cionada pe l'Hospi al de Dijon d'ima ges DICOM (Digi al Imaging
and Communica ions in Medicine) de p òs a a, en aques es udi es p oposa una aplicació web
implemen ada mi jançan Py hon i disponible a a és de Docke , que accedeix di ec amen a la
in o mació de les ima ges i mos a simul àniamen les es modali a s d'ima ges esmen ades. Així
ma eix, p opo ciona una eina pe e ano acions en les di e en s zones de la p òs a a, incloen el umo
si és el cas, i que pe me expo a i o na a impo a -les un cop acaba l'es udi.
Repo
ii
Resumen
El cánce de p ós a a se a a del segundo cánce más ecuen e y la quin a causa p incipal de mue e
po cánce en homb es en el úl imo año. Sin emba go, du an e las úl imas décadas la asa de mue e
ha disminuido no ablemen e debido a, en e o os ac o es, un diagnós ico p ecoz y a una mejo a en
las he amien as pa a ealiza lo.
T adicionalmen e en el es udio de es e cánce , los adiólogos analizan di e en es modalidades de
imagen de o ma indi idual, en e ellas 3D T2- Weigh ed Imaging, Di usion-Weigh ed Imaging y
Pe usion-Weigh ed Imaging. De es a o ma, encon amos casos en los que la obse ación del umo
se e comp ome ida po di e en es aspec os y puede no se de ec ado co ec amen e. Po ello,
nues o obje i o es desa olla una aplicación accesible pa a cualquie cen o médico o hospi al, sin
necesidad de ins alación de un so wa e complejo, que si a de ayuda pa a el diagnós ico de cánce de
p ós a a y de una solución a es as limi aciones.
A pa i de la base de da os p opo cionada po el Hospi al de Dijon de imágenes DICOM (Digi al Imaging
and Communica ions in Medicine) de p ós a a, en es e es udio se p opone una aplicación web
implemen ada median e Py hon y disponible a a és de Docke , que accede di ec amen e a la
in o mación de las imágenes y mues a simul áneamen e las es modalidades de imágenes
mencionadas. Asimismo, p opo ciona una he amien a pa a hace ano aciones en las di e en es zonas
de la p ós a a, incluyendo el umo si es el caso, y que pe mi e expo a y ol e a impo a las una ez
acabado el es udio.
DICOM Viewe : In e ac i e iewe o DICOM medical images
iii
Abs ac
P os a e cance is he second mos common cance and he i h leading cause o cance dea h among
men in he pas yea . Ne e heless, du ing he las decades he dea h a e has no ably dec eased
mos ly as a esul o an ea ly diagnosis and an imp o emen on he ools used o pe o m i .
T adi ionally in he s udy o his cance , adiologis s analyse di e en imaging modali ies indi idually,
including 3D T2-Weigh ed Imaging, Di usion-Weigh ed Imaging and Pe usion-Weigh ed Imaging. In
his way, we can ind cases in which he obse a ion o he umou can be comp omised by di e en
aspec s and may no be de ec ed co ec ly. Fo his eason, ou goal is o de elop a simple, web-based,
applica ion accessible om p ima y ca e cen es o hospi als o help diagnosing p os a e cance and
o e come hese limi a ions.
F om he da abase supplied by he Dijon Hospi al o p os a e DICOM (Digi al Imaging and
Communica ions in Medicine) images, his s udy p oposes a web applica ion implemen ed using
Py hon, which di ec ly accesses he in o ma ion in he DICOM images and displays simul aneously he
h ee images modali ies ha ha e been men ioned. Likewise, i p o ides a ool o make anno a ions
in he di e en a eas o he p os a e, including he umou , and ha allows hem o be expo ed and
impo ed once he s udy has inished.

Repo
i
Acknowledgmen s
This p ojec concludes wi h he wo k o hese las ou yea s. Fi s ly, I would like o hank my supe iso ,
Ch is ian Ma a, o always being willing o help and o all he suppo he has b ough me since he
beginning. Thank you o always being so encou aging, i has been a pleasu e wo king wi h you, as well
as wi h my co-supe iso Raúl Beni ez.
Las bu no leas , i would no ha e been possible o inish his g ade and wo k wi hou he help o my
pa en s, sis e and iends. I am beyond g a e ul wi h hem o being by my side du ing his ime and
making i a lo easie .
DICOM Viewe : In e ac i e iewe o DICOM medical images
Glossa y
ADC Appa en Di usion Coe icien
CZ Cen al Zone
DCE Pe usion-Weigh ed imaging
DICOM Digi al Imaging and Communica ions in Medicine
DWI Di usion-Weigh ed Imaging
MRI Magne ic Resonance Imaging
PCa P os a e Cance
PZ Pe iphe al Zone
SOP Se ice Objec Pai
T2WI 3D T2-Weigh ed Imaging
UID Unique Iden i ie s
Repo
i
Index
RESUM ______________________________________________________________ I
RESUMEN ____________________________________________________________ II
ABSTRACT ___________________________________________________________ III
ACKNOWLEDGMENTS _________________________________________________ IV
GLOSSARY ____________________________________________________________ V
1. INTRODUCTION ___________________________________________________ 5
1.1. P ojec o igin ................................................................................................................. 5
1.2. Mo i a ion ..................................................................................................................... 5
1.3. P ojec amewo k ........................................................................................................ 6
1.3.1. Gene al objec i e ......................................................................................................... 6
1.3.2. Speci ic objec i es ........................................................................................................ 6
2. THEORETICAL FRAMEWORK_________________________________________ 7
2.1. P os a e ana omy ......................................................................................................... 7
2.2. Magne ic Resonance Imaging echniques ................................................................. 8
2.2.1. 3D T2- Weigh ed Imaging (T2WI) ............................................................................... 9
2.2.2. Di usion-Weigh ed Imaging (DWI) .......................................................................... 10
2.2.3. Pe usion-Weigh ed Imaging (DCE) .......................................................................... 11
2.2.4. DICOM o ma ............................................................................................................ 12
2.2.5. Cu en p oblema ic ................................................................................................... 15
3. STATE OF THE ART ________________________________________________ 16
3.1. Li e a u e classi ica ion .............................................................................................. 16
4. PROJECT FRAMEWORK ____________________________________________ 19
4.1. P ojec implemen a ion ............................................................................................. 19
4.1.1. P og amming language .............................................................................................. 20
4.1.2. A chi ec u e ................................................................................................................ 21
4.1.3. G aphic in e ace ........................................................................................................ 22
4.1.4. Image p ocessing ools .............................................................................................. 23
4.1.5. Anno a ion sys em ..................................................................................................... 27
4.1.6. Reposi o ies and deploymen ................................................................................... 27
4.1.7. Ma e ials ..................................................................................................................... 32
DICOM Viewe : In e ac i e iewe o DICOM medical images
ii
5. DICOM VIEWER APPLICATION ______________________________________ 33
5.1. Da abase ...................................................................................................................... 33
5.2. Image iso and il e managemen .......................................................................... 34
5.3. Anno a ion sys em ...................................................................................................... 35
5.4. Case o s udy ............................................................................................................... 37
6. DISCUSSION AND FURTHER WORK __________________________________ 40
7. PROJECT SCHEDULE ______________________________________________ 42
8. ENVIRONMENTAL IMPACT _________________________________________ 43
CONCLUSIONS _______________________________________________________ 44
BUDGET ____________________________________________________________ 46
BIBLIOGRAPHY _______________________________________________________ 47
ANNEX A ____________________________________________________________ 53
A1. Docke ile ..................................................................................................................... 53
A2. Ins uc ions o un docke .......................................................................................... 54
ANNEX B ____________________________________________________________ 55
ANNEX C ____________________________________________________________ 56
DICOM Viewe : In e ac i e iewe o DICOM medical images
7
2. Theo e ical amewo k
2.1. P os a e ana omy
Al hough being he second mos common cance in men wo ldwide, P os a e Cance (PCa) speci ic
su i al is excellen o mos pa ien s. In ac , he dea h a e o p os a e cance has been in signi ican
decline since he mid-1990s. Ea ly diagnosis and a p ope ollow-up, as well as ad anced ea men s
and u he in es iga ion in his ield, ha e led o his d op on he mo ali y a e.
The p os a e is a walnu -sized gland loca ed be ween he bladde and he penis, which su ounds he
p oximal u e h a as i exi s om he bladde . The uppe pa o he p os a e is called he base and he
lowe , na owed pa he apex. Mo eo e , mos pa o he p os a e is co e ed by a hin laye o
connec i e issue called he capsule. An example o he ana omy o a p os a e gland is depic ed in
Figu e 2.1. I is di ided in ou zones: Pe iphe al Zone (PZ), Cen al Zone (CZ), T ansi ion Zone (TZ) and
An e io Fib omuscula S oma (AFS).
Figu e 2.1. Zonal ana omy o he p os a e gland [56]
The Pe iphe al Zone occupies a ound 70% o he p os a e gland, ex ending om he base o he apex
o he pos e io su ace and su ounding he dis al u e h a. I is no mally ep esen ed wi h a high signal
in ensi y in T2-weigh ed images, as i con ains nume ous duc al and acina elemen s wi h spa sely
in e wo en smoo h muscle [4]. In his zone, ca cinoma, ch onic p os a i is, and pos in lamma o y
a ophy a e ela i ely mo e common han in he o he zones, and a ound a 75% o p os a e cance s
o igina e in his zone [1].

Repo
8
The Cen al Zone is a cone-shaped s uc u e which o ms pa o he base o he p os a e, loca ed
be ween he pe iphe al and ansi ion zones, and i is a e sed by he ejacula o y duc s. I accoun s
o app oxima ely 25% o he glandula issue. Tumou s o igina ing in he Cen al Zone, al hough being
less common, end o be mo e agg essi e.
The T ansi ion Zone o ms only 5% o he glandula issue and consis s o wo equal po ions la e al o
he u e h a in he mid-gland. This po ion o he p os a e can enla ge o e he yea s due o he
de elopmen o benign p os a ic hype plasia (BPH), and less p obably adenoca cinoma. On MRI, i is
ep esen ed as nodula a eas wi h di e en signal in ensi y, depending on he ela i e amoun o
glandula and s omal hype plasia [17]. Glandula hype plasia esul s in highe signal in ensi y on T2-
weigh ed images, due o duc al and acina elemen s and sec e ions. On he o he hand, s omal
hype plasia con ains mo e muscula and ib ous elemen s, esul ing in lowe signal in ensi y.
The An e io Fib omuscula S oma (AFMS) co e s he apex o he p os a e. I is composed o ib ous
and smoo h muscula elemen s. Hence, his po ion esul s in low in ensi y signals in T2W1. The dis al
pa o he AFMS is impo an in olun a y sphinc e unc ions, whe eas he p oximal po ion plays a
cen al ole in in olun a y sphinc e unc ions.
In e ms o image anno a ion in p os a e diagnosis, he mos common zones a e CZ, PZ, TZ and Tumou
(Tum), in case i is p esen ed. Fo his eason, ou in e es is mainly ocused on hese ou egions in
his s udy.
2.2. Magne ic Resonance Imaging echniques
Magne ic esonance imaging (MRI) is a non-in asi e imaging echnology ha p oduces h ee
dimensional de ailed ana omical images, used o in es iga e he ana omy and physiology o he body
in bo h heal h and disease. MRI uses a s ong magne ic ield and adio equency pulses o p oduce
de ailed pic u es o o gans, so issues, bone and o he in e nal body s uc u es.
When alking abou MRI i is impo an o be amilia wi h “weigh ed” images, especially wi h T1-
Weigh ed Imaging (T1WI) and T2-Weigh ed Imaging, which can be in 3D (3D T2-Weigh ed Imaging
(T2WI)). These a e ela ed o elaxa ion pa ame e s. As well as ha , he e a e MRI ea u es which ocus
on he pulse sequence. This pulse sequences a e p og ammed se s o changing magne ic g adien s ha
allow he use o image a issue in a ious ways o ob ain impo an diagnos ic in o ma ion abou he
issue. Among his pulse-based ea u es, he mos impo an ones a e he Echo Time (TE) and he
Repe i ion Time (TR). TE e e s o he ime occu ed be ween he applica ion o a adio equency
exci a ion pulse and he peak o he signal induced in he coil. In con as , TR e e s o he ime om
he applica ion o an exci a ion pulse o he applica ion o he nex pulse. Whe eas TE con ols he
DICOM Viewe : In e ac i e iewe o DICOM medical images
9
amoun o T2 elaxa ion, TR de e mines how much longi udinal magne iza ion eco e s be ween each
pulse. Bo h a e measu ed in milliseconds.
Compa ed o o he imaging modali ies, such as X-Ray, CT-scan and ul asound, his echnique is
conside ed o pe o m he bes esul s in e ms o high esolu ion and spon aneous con as o so
issues, besides allowing mul iplana and mul ipa ame ic scanning. Fo hese easons, MRI has been
es ablished as he bes imaging modali y o he de ec ion, localiza ion and s aging o PCa [1].
Fo mos men suspec ed o ha ing p os a e cance , issue is ob ained h ough ans ec al ul asound
guided (TRUS) biopsy du ing which 12 biopsy co es a e andomly aken. S ill, ad ances in MRI ha e
shown imp o ed de ec ion and cha ac e iza ion o p os a e cance by using a mul ipa ame ic
app oach, which combines ana omical and unc ional da a [4]. This mul ipa ame ic MRI e alua ion
includes h ee gene al componen s: high- esolu ion T2WI and a leas wo unc ional MRI echniques,
di usion weigh ed imaging (DWI) and ei he MR spec oscopic imaging (MRSI) o Pe usion weigh ed
imaging (DCE). Fo he pu poses o his p ojec we will discuss T2WI and bo h DWI and DCE.
2.2.1. 3D T2- Weigh ed Imaging (T2WI)
3D T2-Weigh ed Imaging p o ides he bes depic ion o he p os a e’s ana omy, so hey a e some imes
e e ed o as ana omy images. Howe e , i is no ecommended i s use alone, as using i in
combina ion wi h o he echniques inc eases he image quali y in bo h sensi i i y and speci ici y. T2WI
images a e no mally ob ained in wo o h ee o ien a ions.
P os a e cance ypically p esen s as a ound o ill-de ined low signal-in ensi y ocus in he pe iphe al
zone on T2WI, and many o hem can be de ec ed wi hin he high signal-in ensi y backg ound o he
no mal pe iphe al zone glandula issue [4]. Un o una ely, many condi ions such as p os a e
in aepi helial neoplasia, p os a i is, haemo hage, a ophy, sca s and pos - ea men changes can
mimic p os a e ca cinomas. Fu he mo e, he signal in ensi y o he pe iphe al zone can be
comp omised by ho monal abla ion, educing he isibili y o p os a e cance . T ansi ional Zone
umou s a e e en mo e di icul o de ec due o he ac ha he issue low in ensi y o e laps wi h
PCa, and hey a e o en shown as a homogeneous signal mass wi h indis inc ma gins. An example o
a 3D T2-weigh ed image o a p os a e s udy wi h a heal hy and a umou a ea is depic ed in Figu e 2.2.
Repo
10
Figu e 2.2. Example o a 3D T2-weigh ed image o a p os a e s udy [1].
Fo a p ope de ec ion o a PCa, we mus ake in o accoun di e en zones: he capsule, he seminal
esicles and je pos e io bladde wall o ex a-p os a ic umou in asion. Di e en c i e ia ha e o be
conside ed when e alua ing each zone. C i e ia o ex acapsula ex ension a e abu men ; i egula i y
and hickening o he neu o ascula bundle; bulge, loss o capsule and capsula enhancemen ;
measu able ex acapsula disease; and obli e a ion o he ec o-p os a ic angle. As well as ha , an
abno mally low signal in ensi y wi hin he lumen o a ocal hickening o he seminal esicle wall is
sugges i e o seminal esicle in asion. Besides, expansion, low T2 signal in ensi y and illing in o he
p os a e seminal esicle angle can be a sign o seminal esicle in il a ion [5].
2.2.2. Di usion-Weigh ed Imaging (DWI)
Di usion-weigh ed imaging (DWI) allows quali a i e and quan i a i e assessmen o p os a e cance ’s
agg essi eness and p o ides in o ma ion abou he issue in e ms o cell o ganiza ion, densi y and
mic os uc u e.
DWI elies on Appa en Di usion Coe icien (ADC) maps, which a e a measu e o he magni ude o
di usion o wa e molecules wi hin issue and a e commonly clinically calcula ed using MRI wi h DWI.
The mo ion o wa e molecules is mo e es ic ed in issues wi h a high cellula densi y and in ac cell
memb anes, so e y low ADC alues a e clea ly indica i e o cance . Ano he impo an concep in DWI
imaging is he b- alue (s/mm), which iden i ies he measu emen ’s sensi i i y o di usion and
de e mines he s eng h and du a ion o he di usion g adien s. PCa shows a high signal in ensi y wi h
high b- alues and low signal in ensi y/ alue on ADC maps. An example o a DWI o a p os a e s udy
wi h heal hy and umou a ea is depic ed in Figu e 2.3.
DICOM Viewe : In e ac i e iewe o DICOM medical images
11
Figu e 2.3.Example o a Di usion-weigh ed image o a p os a e s udy [1].
High b- alue (800-1000) DW images and ADC maps a e used o a quali a i e assessmen . S ill, no mal
p os a ic issue in some zones like TZ may p esen high signal in ensi y in DWI and low ADC, mimicking
a umou . This can be o e come by using e y high b- alues in combina ion wi h T2WI.
On he o he side, ADC alues a e used o quan i a i e assessmen s, p o iding in o ma ion abou he
umou ’s agg essi eness, and imp o es speci ici y in p os a e cance de ec ion compa ed wi h T2WI
alone. I should, he e o e, be pa o ou ine assessmen s o pa ien s wi h p os a e cance [1].
2.2.3. Pe usion-Weigh ed Imaging (DCE)
Pe usion imaging is based on Dynamic Con as Enhancemen (DCE) o he signal du ing he pass o a
con as agen . The heo e ical basis o his ascula echnique is umou angiogenesis, as he e is a
ela ionship be ween abno mal pe usion and neoangiogenesis in umou s.
In o de o e alua e he umou ’s iabili y, i is commonly used a gadolinium-based con as agen .
Howe e , a compa ison o p e and pos -gadolinium images is no enough o disce n p os a e cance ,
since a no mal p os a e is highly ascula ized. T1WI DCE-MRI imaging da a can be assessed in h ee
ways: quali a i ely, semi-quan i a i ely o quan i a i ely. DCE-MRI images a e acqui ed in a se ies o
axial T1WI g adien echo sequences co e ing he en i e p os a e du ing a bolus injec ion o he con as
medium.
Acco ding o se e al s udies, DCE-MRI gi es be e esul s o p os a e cance localiza ion han T2WI
[1]. An example o a Pe usion-weigh ed image o a p os a e s udy wi h heal hy and umou a ea is
depic ed in Figu e 2.4. Al hough he a ailable in o ma ion is limi ed, i is sugges ed ha DCE may
Repo
12
imp o e s aging. Mo eo e , DCE-MRI is essen ial o he de ec ion o pos -p os a ec omy and
adio he apy ecu ences.
Figu e 2.4.Example o a Pe usion-weigh ed image o a p os a e s udy [1].
2.2.4. DICOM o ma
The Digi al Imaging and Communica ions in Medicine S anda d (DICOM) speci ies a non-p op ie a y
da a in e change p o ocol, digi al image o ma , and ile s uc u e o biomedical images and image
ela ed in o ma ion [12]. DICOM is he globally accep ed s anda d o communica ion and
managemen o a wide ange o medical images. I is used as a way o de ine he o ma o hese images
so ha hey can be exchanged wi h he da a and quali y necessa y o clinical use.
Wi h In e ne becoming he pla o m o new in o ma ion applica ions, DICOM has enabled he
de elopmen o ad anced medical imaging apps ha ha e “changed he ace o clinical medicine” [14].
S uc u e o a DICOM ile
DICOM g oups in o ma ion in o da a se s, consis ing o se e al da a elemen s. In o ma ion Objec s
De ini ion (IOD) a e a key pa in he DICOM s uc u e o ganiza ion, de ining he se o da a elemen s
ha a e being ansmi ed. These da a elemen s s o e he alues o he coded a ibu es o he eal-
wo ld objec ep esen ed in he da a se . They a e o ganized in g oups called modules, as shown in
Figu e 2.5. Each da a elemen is iden i ied by i s ag, numbe ed wi h a unique iden i ie and o de ed
wi hin he DICOM ile om mino o majo [15].
IODs can be associa ed wi h se ices o o m wha is called a Se ice Objec Pai (SOP). Fo ins ance, a
MR image IOD is associa ed wi h he S o age se ice o o m he MR Image S o age SOP Class.

DICOM Viewe : In e ac i e iewe o DICOM medical images
13
Figu e 2.5. DICOM s uc u e [15]
DICOM heade o ma
A DICOM ile con ains a ile heade po ion, a ile me a in o ma ion po ion, and a single SOP ins ance
[1]. The heade is made up o a p eamble o 128-by e, ollowed by he cha ac e s DICOM ( hese a e
used o iden i y whe he i is a DICOM o ma o no ). The ile me a po ion ollows a agged o ma ,
and con ains in o ma ion abou he ile, he se ies, he s udy and he pa ien i belongs o, as i can
be seen in Figu e 2.6.
Repo
14
Figu e 2.6. Example o a DICOM heade [1]
Each DICOM ile is designed o be s andalone, so all he in o ma ion needed o iden i y he ile should
be embedded in each heade . The ac onym UID e e s o a Unique Iden i ie . This in o ma ion is
o ganized in o 4 le els: Pa ien ID, S udyIns anceUID, Se iesIns anceUID, and Ins anceUID:
• Pa ien : Con ains in o ma ion ela ed o he pa ien , such as pa ien ’s name o da e o bi h
among o he s. The key a ibu e in his le el is “Pa ien ID” which is no mally associa ed wi h
his o he medical his o y numbe .
• S udy: Con ains in o ma ion abou he s udy, o ins ance he s udy da e, he acquisi ion
numbe o he s udy desc ip ion. The key a ibu e is “S udyIns anceUID”, which should be
unique o each s udy.
• Se ies: Consis s in a sequence o images o each s udy, and con ains in o ma ion abou a
se ies, such as modali y o se ies numbe . The key a ibu e is “Se iesIns anceUID”.
DICOM Viewe : In e ac i e iewe o DICOM medical images
15
• Ins ance image: Con ains da a abou he ins ance image. The key a ibu e is “Ins anceUID”
which is unique o each image.
2.2.5. Cu en p oblema ic
In o de o make a p ope diagnosis o a p os a e s udy, se e al aspec s need o be aken in o
conside a ion.
Fi s ly, i is eally impo an o ake in o accoun he in e ac i i y be ween he h ee desc ibed MRI
echniques. When a p os a e s udy is ca ied ou , e e y image modali y should be examined
ho oughly. By isualizing he h ee MRI echniques simul aneously, he same egion o in e es can be
seen in each o hem, so he umou isualiza ion is no comp omised by each modali y’s limi a ions
o by any o he p os a e condi ion ha can mislead o an inco ec diagnosis.
On he o he hand, he digi al inpu p omo es he de elopmen o medical image p ocessing, which
has g own in o a as ield o s udy. Applica ions o image p ocessing and isualiza ion a e gene ally
limi ed o he heal h cen e o hospi al. Web-based applica ions ha e become a solu ion o ease he
use o hese applica ions, p o iding a ool accessible om anywhe e, ega dless o he so wa e o he
de ice, wi h no need o a di icul ins alla ion p ocess.
Mo eo e , DICOM is conside ed he s anda d o exchange medical images and i s ela ed in o ma ion.
The e o e, he used applica ion should allow a di ec isualiza ion o DICOM iles and he in o ma ion
con ained in hem.
Fo his eason, in he nex chap e i is ca ied ou a bibliog aphic esea ch o check i he e is any
a ailable applica ion ha sa is ies he p oblema ic exposed abo e. Nex , i is p esen ed a summa y
able wi h all he ela ed wo k ound o e he las decade.
Repo
16
3. S a e o he a
3.1. Li e a u e classi ica ion
The aim o his chap e is o e alua e he cu en a ailable web-based medical applica ions, s udy i s
amewo k and compa e hem. The li e a u e esea ch was ca ied ou using PubMed and Google
Schola , and i ocused on web-based medical applica ions ela ed o medical imaging. In Table 3.1 i
is p esen ed an o e iew o he web-based medical applica ions de eloped in he las decade.
Mos o he cu en ly a ailable applica ions include ools o image isualiza ion and image p ocessing.
Howe e , o he s such as he de eloped by Kaspa e al. [29] ocused mo e on imp o ing accessibili y
in eal- ime, enabling a emo e and in e ac i e s e eoscopic isualiza ion among mul iple iewing
loca ions. Ano he simila app oach was s udied in 2019 by Laja a e al. [42], hey p oposed a solu ion
o deal wi h he lack o a s anda dized whole slide image (WIS) o ma . The de eloped web-app, Visilab
Viewe , allows elepa hology and p ope collabo a i e wo k du ing diagnosis, being able o in e ac
and wo k wi h any WSI based on he DICOM s anda d.
A web-based sys em o e inal image analysis was p oposed in 2010 by O ega e al. [33], p o iding a
amewo k o oph halmologis s and o he expe s in he field o collabo a i ely wo k using e inal
image-based applica ions. Re inal mic oci cula ion is assessed using a semi-au oma ic me hodology
o he compu a ion o he A e iola - o- enula Ra io (AVR). Fu he mo e, in 2016 ano he use ul web
ool o oph halmologis s was de eloped by Remesei o e al. [43] o assess d y eye disease, e alua ing
ea ilm images. I allows manual anno a ions and also a ool ha in eg a es se e al algo i hms o
au oma ic ea ilm pa e n analysis.
In 2011, Bal asa e al. [34] designed a web- ool o suppo diagnos ic clinical ials in ol ing di e en
expe s and hospi als o esea ch cen e s. The image analysis o his p ojec is based on skele al X- ay
imaging; besides i allows o s o e da a and images o he da abase. Mo eo e , in 2014 Kamme e e
al. [38] p esen ed a simila wo k, MyCases, a po able da abase se e o quick and easy s o age o
such cases including whole image se ies, pa ien da a, and anno a ions.
In 2010, Mahmoudi e al. [31] de eloped a web-based, in e ac i e, ex endable, 2D and 3D medical
image p ocessing and isualiza ion applica ion ha equi es no clien ins alla ion. I also includes a wide
ange o medical image p ep ocessing, egis a ion, and segmen a ion me hods, implemen ed using
open sou ce lib a ies. The so wa e suppo s image o ma s such as i , gi , jpeg and DICOM. Ano he
app oach con aining segmen a ion ools was p oposed by Young e al. [45], which consis ed in a web-
based au oma ic spine segmen a ion me hod using deep lea ning ha can be e y p ac ical and
accu a e o spine segmen a ion as a diagnos ic me hod. Mo eo e , Yang e al. [32] also s udied a web-
DICOM Viewe : In e ac i e iewe o DICOM medical images
23
Figu e 4.1. DICOM Viewe g aphic in e ace
4.1.4. Image p ocessing ools
As we ha e seen one o he main pa s o he app a e he h ee simul aneous panels displaying each
one an MRI echnique. In o de o isualize and imp o e i s isualiza ion, i is included a ool o apply
basic image p ocessing il e s o he images. All he image p ocessing il e s a e applied om he Py hon
lib a y Skimage [22]. Di e en modules ha e been used: he module Fil e s (Gaussian, Median and
Sobel), Mo phology (E osion and Dila ion) and las ly Exposu e (Gamma con as ).

Repo
24
Median il e
Median il e is one smoo hing il e and i emo es speckle noise and impulsi e noise om he image,
especially impulse and sal & peppe noise. As he median il e is applied on o an image, each pixel is
eplaced wi h he median alue o i s neighbo s [28].
One o he main ad an ages o he median il e is ha i also p ese es he edges p esen in he image.
Fu he mo e, i does no in oduce new pixel alues since i only e-use exis ing pixel alues om he
window. An example o a Median il e applied o a p os a e image compa ed o a Gaussian il e can
be seen in Figu e 4.2.
Gaussian il e
A Gaussian Fil e is a low pass il e used o educing noise (high equency componen s) and blu ing
egions o an image [28]. In his sense i is simila o he mean il e , bu i uses a di e en ke nel ha
has he shape o he unc ion ‘Gaussian dis ibu ion’ (Equa ion 4.1) o de ine he weigh s inside he
ke nel, which a e used o compu e he weigh ed a e age o he neighbo ing poin s (pixels) in an image.
𝐺 𝜎 =1
2𝜋𝜎2𝑒−(𝑥2+𝑦2)
2𝜎2
Equa ion 4.1. Gaussian dis ibu ion
The di e ences be ween bo h Median and Gaussian smo he ing il e s is shown in Figu e 4.2.
Figu e 4.2. Compa ison be ween Median and Gaussian il e . F om le o igh i can be seen: he image wi hou il e ,
Gaussian and Median il e .
DICOM Viewe : In e ac i e iewe o DICOM medical images
25
Sobel
Figu e 4.3. On he le , a p os a e di usion image wi hou il e and on he igh he same image wi h a Sobel il e .
The Sobel il e is used o edge de ec ion, which a e apid changes in he image in ensi y unc ion. The
e ec s can be obse ed in Figu e 4.3. A Sobel edge de ec ion ope a o consis s o a pai o con olu ion
ke nels as shown in Equa ion 4.2, whe e he second ke nel is a o a ion o he i s one. These ke nels
a e designed o espond maximally o e ical and ho izon al edges [46].
Equa ion 4.2. Sobel ope a o s
E osion
E osion is a mo phological il e whe e he alue o he ou pu pixel is he minimum alue o all pixels
in he neighbo hood. By using e osion, islands and small objec s can be emo ed so ha only
subs an i e pa s emain. I ac s like a local minimum il e [47]. In Figu e 4.4 he e is an example o an
e osion il e applied o a p os a e image in con as o he dila ion il e .
Dila ion
Opposi e o e osion, dila ion il e adds a laye o pixels o bo h he inne and ou e bounda ies o
egions so ha he alue o he ou pu pixel is he maximum alue o all pixels in he neighbo hood.
The e o e, i ac s like a local maximum il e . Dila ion can be used o make objec s mo e isible and ill
in small holes in objec s [47]. The no able di e ence be ween an E osion and Dila ion il e is shown in
Figu e 4.4.
Repo
26
Figu e 4.4. Compa ison be ween an e osion il e on he middle, a dila ion il e on he igh , and a no mal p os a e ana omy
image, on he le
Gamma con as
Gamma co ec ion is used o co ec he di e ences be ween he way a came a cap u es con en ,
he way a display displays con en and he way ou isual sys em p ocesses ligh . I is he name o a
non-linea ope a ion used o code and decode luminance, maximizing he use o he bi s o
bandwid h ela i e o how humans pe cei e ligh and colo .
Figu e 4.5. A Gamma con as il e , on he igh , applied o a Pe usion image, on he le .
Gamma co ec ion is de ined in he simples cases by he exp ession on Equa ion 4.3 [48], whe e A
ep esen s a cons an , ypically om 0 o 1, and he inpu and ou pu alues a e non-nega i e eal
alues. A gamma alue γ < 1 will u n he image da ke while gamma alues > 1 will make he image
appea ligh e . On Figu e 4.5 i can be seen he e ec s o applying his il e wi h a gamma alue o
1.5.
DICOM Viewe : In e ac i e iewe o DICOM medical images
27
Equa ion 4.3. Gamma co ec ion
4.1.5. Anno a ion sys em
The anno a ion sys em p o ides a ool o expe s o manually anno a e egions and classi y hem
be ween he di e en p os a e zones, including he Cen al zone, Pe iphe al Zone, T ansi ion zone and
he Tumou i sel . The py hon lib a y Plo ly has some p ede ined ool o d aw di e en kinds o labels
in images, such as ec angula o open ee o m. Fo he means o ou applica ion he p ede ined ool
we use is a closed ee o m, which le s us d aw a limi ed egion, de ining a zone.
The applica ion s o es he coo dina es o he anno a ions, which can be expo ed in a CSV (Comma
Sepa a ed Values) o ma [50]. A CSV ile is a plain ex ile ha con ains a lis o da a. I has a ai ly
simple s uc u e, a lis o da a sepa a ed by commas. Fo ins ance, a ile wi h wo di e en kinds o
anno a ions, one o he CZ and ano he o a umou would look like is depic ed in Figu e 4.6. Hence,
o a la ge lis he s uc u e would look exac ly he same: he name o he zone ollowed by a coma
and he coo dina es o he closed ee o m label.
The idea is ha you can expo complex da a om he applica ion o a CSV ile, and hen impo he
same da a in o he applica ion again.
4.1.6. Reposi o ies and deploymen
4.1.6.1. Gi Hub
Gi Hub is a code hos ing pla o m o e sion con ol and collabo a ion. We ha e c ea ed a eposi o y
whe e he e can be ound all he iles o he p ojec , which is explained in Annex B.
The main code can be ound on he app.py ile and he ile README explains how o un he app locally.
Basically, you i s need o ins all all he dependencies o un he p ojec , execu ing he command pip
Type, Coo dina es
Cen al
Zone,"M74.67899761336515,108.73329355608591L76.46897370167,109.44928400954653L75.75298329355608,111.23
926014319808L74.67899761336515,110.88126491646777L74.67899761336515,108.37529832935559Z"
Tumo ,"M72.17303102625297,114.10322195704056L73.96300715990454,115.89319809069211L72.53102625298328,1
15.5352028639618L72.53102625298328,114.81921241050118Z"
Figu e 4.6. Example o a CSV ile s uc u e.
Repo
28
ins all - equi emen s. x . Nex , execu e he command py hon app.py (o py hon3 app.py, depending
on he py hon e sion you ha e ins alled in you sys em) and he app will begin o execu e.
Figu e 4.7. Lis o olde s and iles ha can be ound in he DICOM Viewe eposi o y [51]
4.1.6.2. Docke
Dash apps can be deployed o se e s. In ou case we will use Docke , which is an open pla o m o
de eloping, shipping, and unning applica ions [26]. Docke packages and uns an applica ion in an
isola ed en i onmen called a con aine , which con ains e e y hing needed o un he applica ion. The
isola ion and secu i y allow you o un con aine s simul aneously on a gi en hos .
Figu e 4.8. Docke a chi ec u e [26]
In Figu e 4.8 i is shown a schema o Docke wo ks. Docke uses a clien -se e a chi ec u e: he clien
sends commands o he Docke daemon, which ca ies hem ou . Hence, he daemon builds, uns and
dis ibu es Docke con aine s.

DICOM Viewe : In e ac i e iewe o DICOM medical images
29
These con aine s o m pa o he se ies o objec s which a e c ea ed when using Docke . The
ins uc ions and he empla e needed o c ea e con aine s a e held on Images, and hese Images a e
s o ed in a Docke egis y, such as Docke Hub.
The s eps ollowed o con aine ize ou py hon applica ion a e he ollowing:
• We c ea e a Docke ile which con ains he ins uc ions o build he Py hon image
• We build he image
• Run he image as a con aine
• Deploy he applica ion o a hub
Build he Py hon image
In o de o build he image, we i s need o c ea e a Docke ile. A Docke ile is a ex documen ha
con ains all he commands which need o be called on he command line o assemble an image. An
example o a Docke ile can be ound in Annex A1.
Fi s ly, in he oo o ou wo king di ec o y we c ea e a ile named Docke ile, which mus con ain he
ins uc ions desc ibed in Figu e 4.9.
Figu e 4.9. Commands o he Docke ile
• FROM py hon:3 his command ells Docke which base image should use.
• WORKDIR /us /s c/app ins uc s Docke o use his pa h as he de aul loca ion o all
subsequen commands so we can use ela i e pa hs
• COPY equi emen s. x ./ is used o copy ou equi emen s. x ile in o he image, which
con ains all he dependencies needed o un ou applica ion.
• RUN pip3 ins all – no-cache-di - equi emen s. x ins alls he modules in o he image, so
all he needed dependencies a e ins alled.
Repo
30
• COPY . . command akes all he iles loca ed in he cu en di ec o y and copies hem in o
he image
• The EXPOSE ins uc ion in o ms Docke ha he con aine lis ens on he speci ied
ne wo k po .
• CMD [ "py hon", "/app.py"] ells Docke which command we wan o un when ou image
is execu ed inside a con aine .
Once we ha e buil he Docke ile we can build he applica ion Image. To do so we use he docke build
command, shown in Figu e 4.10, which builds Docke images om a Docke ile and a “con ex ”. This
con ex e e s o a se o iles loca ed in he speci ied pa h. The command akes an op ional -- ag lag,
o speci y he name o he image.
Figu e 4.10. Ins uc ions o build a Docke image
Run he image as a con aine
Con aine s a e unnable ins ances o images. Fi s s ep o un a con aine inside an image is o use he
docke un command (Figu e 4.11). Besides, we need o speci y a po o ou con aine , by using he -
- publish command o -p o sho . Fo ins ance, o map he hos ’s po 8050 o he con aine ’s po
8050 we mus ype -p 8050:8050.
This command should be ollowed by -- de ach o -d i we wan ou con aine o un in de ached mode
o in he backg ound. Mo eo e , o speci y he name o he con aine i is necessa y o include he --
name lag on he docke un command.
Figu e 4.11. Command o un he Docke con aine
To lis he buil images we use docke images, and o see he unning con aine we use he command
docke ps. Mo eo e , o s op he con aine om unning we can ype docke s op ollowed by he
con aine name.
DICOM Viewe : In e ac i e iewe o DICOM medical images
31
Push a Docke con aine image o a Docke Hub
Docke Hub eposi o ies a e used o sha e con aine images, which can be pushed o Docke Hub
h ough he docke push command.
Fi s ly, i is needed o name he local image using he Docke Hub use name and he eposi o y name
c ea ed. This can be done by e- agging an exis ing local image docke ag <exis ing-image> <hub-
use >/< epo-name>, as i can be seen in Figu e 4.12. A e ha , i is used docke push <hub-
use >/< epo-name> o inally push he image o he Hub.
Figu e 4.12. Push Docke image o Docke Hub
The applica ion eposi o y [52] in Docke Hub, shown in Figu e 4.13, can be downloaded by using he
command docke pull julia omagosa/dicom- iewe :dicom- iewe , which downloads i locally in he
hos . In o de o do so, you i s need you i s need o ins all Docke Desk op [58], which p o ides he
speed and secu i y needed o use his applica ions on you desk op.
Figu e 4.13. Gi Hub eposi o y in e ace [52]
Repo
32
4.1.7. Ma e ials
In his chap e i is p esen ed he cu en ly wo king DICOM p os a e da abase supplied by he Hospi al
o Dijon. This is composed o ana omical, di usion, and pe usion images o 47 di e en cases o s udy.
The e a e a o al o 64 images o T2WI, 14 o DWI and a ound 640 o DCE imaging pe case.
Fu he mo e, he a e age age o all he pa ien s is a ound 52 yea s. Thei pe sonal in o ma ion, such
as hei comple e name, is anonymized.
T2W1 as spin-echo images we e acqui ed wi h a sub-millime ic pixel esolu ion on an oblique axial
plane. The speci ic acquisi ion pa ame e s we e a TR, TE and ETL o 3600 ms, 143 ms and 109
co espondingly, and a slice hickness o 1.25 mm.
To acqui e DWI images i was used a pulsed g adien spin-echo echnique wi h wo b- alues, 100 and
800 sec/mm2. ADC mapping was gene a ed om he aw da a on a pixel-by-pixel basis, and he TR and
TE we e o 4200 and 101 ms espec i ely.
DCE-MRI was pe o med using a a -supp essed 3D T1 VIBE sequence, which is speci ied by a TR and a
TE o 3.25 and 1.12 ms, as well as a lip angle o 10 deg ees and a empo al esolu ion o 6 sec/slab
o e app oxima ely 5 minu es. As well as ha , i was adminis e ed a bolus injec ion o Gd-DTPA, a a
dose o 0.2 ml Gd-DTPA/kg o body weigh .
DICOM Viewe : In e ac i e iewe o DICOM medical images
39
As a esul , in Figu e 5.9 i is shown he simul aneous iew ha o e s DICOM Viewe o he umou in
each MRI modali y, wi h he p os a e gland also classi ied in i s egions o in e es . F om his case we
can conclude ha wha can be con used as pa o a heal hy p os a e issue in an ana omical image,
can be inally diagnosed as a umou by compa ing i in Pe usion and Di usion images. He e elies he
impo ance o analysing a suspicious egion ho oughly in di e en MR image modali ies, so we can
p o ide an accu a e diagnosis.
Figu e 5.9. Simul aneous iew o a umou in di e en MRI images

Repo
40
6. Discussion and u he wo k
DICOM Viewe is a web-based applica ion o he analysis o p os a e images using h ee di e en MR
image echniques: T2WI, DWI and DCE. One o he mos impo an issues we wan ed o add ess is he
necessi y o simul aneously isualize di e en MRI images o he same pa ien in o de o diagnose
some pa hology o lesion. Tha is because o he high impo ance i has o conside he
in e connec i i y be ween hese di e en echniques o image acquisi ion. Hence, he p oposed
in e ace is di ided in o h ee panels so ha expe s can s udy he same egion o in e es a he same
ime, simul aneously, h ough he di e en echniques. As well as ha , he p esen ed applica ion
p o ides a ool whe e expe s can anno a e impo an indings in a eally simple way, in o de o make
he diagnosis o a pa ien s udy. Once he s udy has been analysed, he indings a e epo ed in a
s anda dized o ma ; cs . Besides, his in o ma ion can be la e impo ed o he app again. In his way,
i p o ides a use ul ool o expe s o manage examina ions wi h di e en ypes o images and da a.
Fu he mo e, DICOM Viewe di ec ly eads and shows DICOM iles, including impo an pa ien
in o ma ion ha could be use ul o expe s. Since DICOM is conside ed he globally accep ed s anda d
o manage wi h medical images, i can easily be adap ed o o he modali ies and p o ocols, al hough
being specialized in p os a e images.
Ano he main ad an age is ha ou ool is designed o wo k in a web-based en i onmen . I is
de eloped using a e y in ui i e p og amming language, Py hon, and i is implemen ed using Dash o
c ea e an in e ac i e and dynamic applica ion. Docke is used o deploy he app, so mul iple use s can
access DICOM Viewe emo ely and no ins alla ion is needed on any compu e . Mo eo e , i can be
execu ed om any ope a ing sys em, including Windows, Linux and MAC. The only equi emen is o
ins all Docke desk op on you compu e , which is ee and as .
On he o he hand, he p ojec we p esen is s ill he i s elease. The e a e some ways in which he
applica ion can be imp o ed. Fo ins ance, he e a e some pa ame e s ega ding he image il e s ha
could ha e had some inpu o modi y hem, such as he gamma alue in he Gamma con as il e .
Mo eo e , we could apply au oma ic segmen a ion il e s ha can help o imp o e he e iciency o
he diagnosis, as cu en ly i is all manually done. Fu he mo e, he applica ion could ha e a simple
way o go di ec ly o a speci ic image numbe .
In addi ion, we ound some limi a ions when expo ing he da abase due o a limi a ion o ime. We
could ha e s udied a unc ion o expo he en i e da abase wi h i s anno a ions, no c ea ing
sepa a ed iles o each one. Docke has also go some sho comings. In case we wan o add mo e
pa ien s o he ac ual da abase, we would ha e o build he Docke image again, as i canno be
DICOM Viewe : In e ac i e iewe o DICOM medical images
41
modi ied. I he applica ion is being used locally ia he Gi Hub eposi o y, he e would no be a
p oblem as iles can easily be added o he da abase olde .
Finally, i has o be ema ked ha his applica ion has been alida ed by he medical image expe s in
he Hospi al o Dijon. This alida ion is based on he pape p esen ed by Ma a e al. [59], whe e i is
analysed he g ound- u h ob en ion o a p os a e cance analysis using MRI by wo di e en expe s.
Medical indings in di e en egions o he p os a e gland a e e alua ed using collabo a i e wo k,
demons a ing ha e alua ion wi h a p e ious knowledge o he o he expe s opinion educes he
a iabili y and inc eases he quali y o he diagnosis.
Repo
42
7. P ojec schedule
In he nex sec ion i is desc ibed he Gan diag am ollowed o de elop his p ojec , which can be
seen in Table 7.1. I comp ises 16 weeks. The i s mee ings s a ed du ing mid-Feb ua y and las ed
un il he second week o June, when we concluded his epo .
The de elopmen o his p ojec ocused on h ee s ages. Fi s o all, i was ca ied ou an ini ial
in o ma ion esea ch o lea n he basics o his wo k and o ge in ouch wi h he Dash s uc u e. Then,
he main pa o he de elopmen ocused on he p og amming pa , whe e we dedica ed mos o he
ime and which needed he mos a en ion. Tha included he deploymen o he applica ion. Finally,
we ga he ed all his in o ma ion and esul s on his epo , and we made he inal conclusions and
hough abou u u e wo k.
Table 7.1. Gan diag am o he p ojec o ganiza ion
DICOM Viewe : In e ac i e iewe o DICOM medical images
43
8. En i onmen al impac
In his chap e we aim o e alua e he en i onmen al impac o his p ojec , ha could esul om he
de elopmen s age and also om he use o i .
Due o he expec a ional si ua ion de i ed om he COVID pandemic, all he mee ings wi h he p ojec
di ec o and co-di ec o we e made ia google mee . Fo his eason, he e has been no emission o
gases om any kind o anspo .
Basically, his p ojec consis s in an applica ion p og ammed exclusi ely in Py hon language using an
in e p e e , PyCha m [55], and deployed ia Docke . Thus, he only ool needed o pe o m his s udy
was a compu e , speci ically a MacBook P o 13-inch o 2017 and ou Thunde bol -3 po s. This de ice
was no acqui ed in he means o his p ojec ; he e o e, we only conside he elec ic powe used o
cha ge i pa o he en i onmen al impac , no accoun ing he manu ac u ing p ocess o i . In o de o
calcula e he impac gene a ed by he compu e 's consump ion o elec ici y, we mus ake in o
conside a ion he emission o CO2. As s a ed in he Gan diag am in Chap e 7, he whole p ojec was
ca ied ou du ing a o al o 16 weeks. Conside ing an a e age o 40 hou s pe week we assume a
numbe o 650 wo ked hou s. Mo eo e , he compu e speci ica ions s a e a powe consump ion o
61 W [53]. In o al, he elec ici y consump ion de i ed om he compu e ’s use o he p ojec adds
up o 44.84 kWh (Equa ion 8.1).
𝐸𝑙𝑒𝑐𝑡𝑟𝑖𝑐𝑖𝑡𝑦 𝑐𝑜𝑛𝑠𝑢𝑚𝑝𝑡𝑖𝑜𝑛 = 650 ℎ ∗ 61 𝑊 = 39650 Wh = 39,65 kWh
Equa ion 8.1. To al elec ici y consump ion de i ed om he compu e use
Wi h he aim o calcula ing he gas emission p oduced by his consump ion, we assume ha one kWh
emi s 0.15 kg o CO2 in Spain, acco ding o he Na ional Commission o Ma ke s and Compe i ion
(CNMC) on da e Ap il 16 h o 2021 [54]. Consequen ly, a o al o 5,94 CO2 kg ha e been emi ed du ing
his p ojec de elopmen (Equa ion 8.2).
𝐶𝑂2 𝑒𝑚𝑖𝑠𝑠𝑖𝑜𝑛 = 39,65 kWh ∗ 0,15 kg
kWh = 5,94 𝐶𝑂2 𝑘𝑔
Equa ion 8.2. C02 emission de i ed om he elec ici y consump ion
On he o he hand, he used DICOM p os a e da abase was no di ec ly acqui ed o his s udy, so we
conside i ou o he p ojec scope. As a esul , no en i onmen al impac de i es om he images
acquisi ion p ocess. Likewise, as he applica ion is web-based, he only impac i s use will ha e is, as
well as in he o he assump ions, he consump ion o elec ic powe .
Repo
44
Conclusions
P os a e cance is he second mos common cance in men wo ldwide. Thus, along wi h he signi ican
endency o applying new echnologies o he so wa e o medical imaging cen e s, has led o he
de elopmen o his p ojec . Ou main goal was o de elop a medical ool o he heal hca e
communi y o manage p os a e cance , demons a ing he capabili ies and ad an ages o
implemen ing a web-based applica ion.
Du ing he second chap e we p esen ed a heo e ical basis o unde s and he p os a e ana omy,
mos ly ocusing on de ining he p os a e gland zones whe e umou s can be o igina ed. Besides, i is
explained he acquisi ion o p os a e images using MRI echniques, which ollows a DICOM p o ocol,
also explained in his sec ion.
In he hi d chap e se e al web-based medical applica ions, designed o medical image pu poses, a e
e iewed and e alua ed. Many o hem p o ided some ools o anno a e di e en image zones. We
wan ed ou ool o di e om o he s so ha i allowed a simul aneous analysis o he same egion o
in e es using di e en MRI echniques (DWI, DCE and T2W1), which inc eases he diagnosis accu acy.
The e a e di e en p os a e condi ions ha can mimic cance , o o he s ha can comp omise he
isualiza ion o umou s. By ca ying ou a simul aneous analysis o he same egion in each modali y,
he p obabili ies o conduc ing a co ec diagnosis inc ease signi ican ly, as i has been demons a ed
in an example shown in he case o s udy. Besides, he con inuous echnological e olu ion c ea es a
need o hese apps o be cons an ly imp o ed.
In his con ex , du ing he ou h chap e we p esen he amewo k o ou applica ion, discussing
di e en implemen a ion op ions and de ining he decided sys em a chi ec u e. As we aimed o ease
all he p ocess ela ed o p os a e cance diagnosis, we decided o use a web-based sys em
a chi ec u e o acili a e he ins alla ion p ocedu e and o ensu e ha e e y heal hca e cen e could
make use o i . The language used o implemen his pla o m was Py hon, making p o i o he Dash
lib a y and i s componen s o de elop an in e ac i e web-based applica ion. Mo eo e , we needed o
implemen a ool ha could ead and access DICOM in o ma ion di ec ly, as i is he s anda d o
manage medical images and i s in o ma ion. Rega ding he design o he images iso , we had in mind
he impo ance o he p e iously men ioned in e ac i i y be ween he di e en image modali ies
ound in he da abase, as well as p o iding some image p ocessing ools o imp o e his isualiza ion.
Conce ning he anno a ion sys em, i is designed o help expe s o manage anno a ing image indings
in an easy and in ui i e way. Besides, hese anno a ions can be expo ed, in a cs o ma , and impo ed
la e in case he adiologis s need i .

DICOM Viewe : In e ac i e iewe o DICOM medical images
45
In chap e i e, i is shown how he de eloped applica ion, DICOM Viewe , looks like and how o use
i . Each sec ion o i is desc ibed in de ail explaining all he p o ided unc ionali ies; including he able
con aining he DICOM in o ma ion, he image iso displaying h ee di e en MRI echniques and also
p esen ing he image p ocessing ools, and inally he sys em o pe o m anno a ions and he me hod
o expo and impo hese indings.
The applica ion code can be ound a he Gi Hub eposi o y [51], whe e i can be downloaded and
execu ed using a py hon in e p e e . Mo eo e , i is also possible o use i h ough Docke . To access
he Docke Hub, you jus need o download he Docke app, which is ee and as o download, and
open he eposi o y o DICOM Viewe [52]. Hence, he e is no need o ins all any complica ed sys em
amewo k.
To sum up, he de eloped i s e sion o DICOM Viewe accomplishes all he o iginal speci ic and
gene al objec i es. We designed a ool des ined o be used by he medical communi y wi h he suppo
o medical s a and hospi als o he diagnosis o p os a e cance . Howe e , he e a e some ea u es
ha we ha e hough abou du ing i s de elopmen ha could imp o e i s use, bu his i s e sion is
e icien and eady o use.
This p ojec has pe sonally gi en me he oppo uni y o gain a huge amoun o knowledge in many
aspec s. I ha e lea ned he ha d wo k behind p og amming a web app, helping me ealize he
pe se e ance p og amming needs. In o de o sol e he p oblems ha may come ac oss you mus
sea ch a lo o in o ma ion and look a hings closely in a di e en way, e hinking you o iginal ideas
and imp o ing hem. Mo eo e , I ha e imp o ed he p e ious expe ience I had wo king wi h Py hon,
which is some hing I am su e i is going o be use ul in he u u e, as his language is widely used o
many o he inali ies ela ed o he medical wo ld. I also alue ha he de eloped applica ion is eady
o be used by medical expe s, and ha his wo k I ha e ca ied ou can be use ul and con ibu e o
many o he p e ious and u u e wo k ela ed o medical diagnosis. Rela ed o his, in Annex C i can be
ound a D a o he pape ha we ha e p esen ed o he IEEE-EMBS In e na ional Con e ence on
Biomedical and Heal h In o ma ics BHI), which will ake place on he 27 h o July.
Repo
46
Budge
In his Chap e i is p esen ed he de ailed budge o his p ojec , b oken down by di e en ca ego ies
and aking in o accoun he numbe o hou s spen in each di e en ask and he cos pe hou o an
enginee . Due o he p ojec scope, i was only needed a compu e wi h he necessa y so wa e. The
compu e was no acqui ed owing o his wo k, so we do no conside i a pa o he de eloped budge .
Mo eo e , all he so wa e we used did no ha e any cos , as PyCha m is an open sou ce p og am.
Table 0.1 B eak down o he p ojec ’s budge
DICOM Viewe : In e ac i e iewe o DICOM medical images
47
Bibliog aphy
[1] Ma a, C. (2015). Web-based applica ion o medical imaging managemen . [PhD Thesis, Uni e si a
de Gi ona- Uni e si é de Bou gogne]. h ps://www. dx.ca /handle/10803/323093#page=9
[2] Sung, H., Fe lay, J., Siegel, R. L., La e sanne, M., Soe joma a am, I., Jemal, A., & B ay, F. (2021).
Global Cance S a is ics 2020: GLOBOCAN Es ima es o Incidence and Mo ali y Wo ldwide o 36
Cance s in 185 Coun ies. CA: A Cance Jou nal o Clinicians, 71(3), 209–249.
h ps://doi.o g/10.3322/caac.21660
[3] Key S a is ics o P os a e Cance . (2021). Ame ican Cance Socie y.
h ps://www.cance .o g/cance /p os a e-cance /abou /key-s a is ics.h ml
[4] Bha sa , A., & Ve ma, S. (2014). Ana omic Imaging o he P os a e. BioMed Resea ch In e na ional,
2014, 1–9. h ps://doi.o g/10.1155/2014/728539
[5] Ba en sz, J. O., Richenbe g, J., Clemen s, R., Choyke, P., Ve ma, S., Villei s, G., Rou ie e, O., Logage ,
V., & Fü e e , J. J. (2012). ESUR p os a e MR guidelines 2012. Eu opean Radiology, 22(4), 746–757.
h ps://doi.o g/10.1007/s00330-011-2377-y
[6] Qiao, L., Li, Y., Chen, X., Yang, S., Gao, P., Liu, H., Feng, Z., Nian, Y., & Qiu, M. (2015). Medical high-
esolu ion image sha ing and elec onic whi eboa d sys em: A pu e-web-based sys em o
accessing and discussing lossless o iginal images in elemedicine. Compu e Me hods and
P og ams in Biomedicine, 121(2), 77–91. h ps://doi.o g/10.1016/j.cmpb.2015.05.010
[7] Wickens, B., Lewis, J., Mo is, D. P., Husein, M., Ladak, H. M., & Ag awal, S. K. (2015). Face and
con en alidi y o a no el, web-based o oscopy simula o o medical educa ion. Jou nal o
O ola yngology - Head & Neck Su ge y, 44(1), 7. h ps://doi.o g/10.1186/s40463-015-0060-z
[8] Shen, H., Ma, D., Zhao, Y., Sun, H., Sun, S., Ye, R., Huang, L., Lang, B., & Sun, Y. (2014). MIAPS: A
web-based sys em o emo ely accessing and p esen ing medical images. Compu e Me hods and
P og ams in Biomedicine, 113(1), 266–283. h ps://doi.o g/10.1016/j.cmpb.2013.09.008
[9] Maglogiannis, I., And ikos, C., Rassias, G., & Tsanakas, P. (2017). A DICOM Based Collabo a i e
Pla o m o Real-Time Medical Teleconsul a ion on Medical Images. Ad ances in Expe imen al
Medicine and Biology, 79–91. h ps://doi.o g/10.1007/978-3-319-57348-9_7
Repo
48
[10] Yuan, R., Shi, S., Chen, J., & Cheng, G. (2018). Radiomics in RayPlus: a Web-Based Tool o Tex u e
Analysis in Medical Images. Jou nal o Digi al Imaging, 32(2), 269–275.
h ps://doi.o g/10.1007/s10278-018-0128-1
[11] Min, Q., Wang, X., Huang, B., & Xu, L. (2020). Web-Based Technology o Remo e Viewing o
Radiological Images: App Valida ion. Jou nal o Medical In e ne Resea ch, 22(9), e16224.
h ps://doi.o g/10.2196/16224
[12] Bidgood, W. D., Ho ii, S. C., P io , F. W., & Van Syckle, D. E. (1997). Unde s anding and Using
DICOM, he Da a In e change S anda d o Biomedical Imaging. Jou nal o he Ame ican Medical
In o ma ics Associa ion, 4(3), 199–212. h ps://doi.o g/10.1136/jamia.1997.0040199
[13] Gibaud, B. (2008). The DICOM S anda d: A B ie O e iew. Molecula Imaging: Compu e
Recons uc ion and P ac ice, 229–238. h ps://doi.o g/10.1007/978-1-4020-8752-3_13
[14] DICOM s anda d. (n. d.). DICOM. h ps://www.dicoms anda d.o g/
[15] DICOM. (n. d.). Lead ools. h ps://www.lead ools.com/help/lead ools/ 19/dh/ o/di- opics-
dicom.h ml
[16] Villei s, G. M., & De Mee lee , G. O. (2007). Magne ic esonance imaging (MRI) ana omy o he
p os a e and applica ion o MRI in adio he apy planning. Eu opean Jou nal o Radiology, 63(3),
361–368. h ps://doi.o g/10.1016/j.ej ad.2007.06.030
[17] Hamme ich, K. H., Ayala, G. E., & Wheele , T. M. (2008). Ana omy o he p os a e gland and
su gical pa hology o p os a e cance . P os a e Cance , 1–14.
h ps://doi.o g/10.1017/cbo9780511551994.003
[18] Pydicom. (n. d.). h ps://pydicom.gi hub.io/
[19] Dash. (n. d.). PyPI. h ps://pypi.o g/p ojec /dash/
[20] In oducing Dash - Plo ly. (2019). Medium. h ps://medium.com/plo ly/in oducing-dash-
5ec 7191b503
[21] Van Rossum, G., & L. D ake, F. (2002). Py hon Re e ence Manual Release 2.2.1.
h p://ci esee x.is .psu.edu/ iewdoc/summa y?doi=10.1.1.406.6230
DICOM Viewe : In e ac i e iewe o DICOM medical images
55
Annex B
In he Gi Hub eposi o y [51] he e a e all he iles needed o un he applica ion. By clicking on he
g een bu on shown in Table A. 1, he olde con aining all he iles will begin o download in a zip
o ma .
In o de o un he app you need o ha e ins alled a py hon in e p e e and ollow he ins uc ions on
he README ile.
Table A. 1. Sc eensho o he DICOM Viewe eposi o y on Gi Hub
The da abase is inside he asse s olde , classi ied by he pa ien ’s iden i ie and di ided in olde s o
each MRI echnique.

Annexos
56
Annex C