i
ASSESSMENT OF INNOVATIVE
TECHNOLOGIES APPLICATION IN MEDICAL
IMAGING IN PORTUGAL
Joana Filipa Soa es da Sil a
Disse a ion p esen ed as pa ial equi emen o ob ain
he deg ee o Mas e o In o ma ion Managemen
i
NOVA In o ma ion Managemen School
Ins i u o Supe io de Es a ís ica e Ges ão de In o mação
Uni e sidade No a de Lisboa
ASSESSMENT OF INNOVATIVE TECHNOLOGIES APPLICATION IN
MEDICAL IMAGING IN PORTUGAL
by
Joana Filipa Soa es da Sil a
Ad iso : Vi o San os
Ad iso : Ca olina San os
No embe 2021
ii
ACKNOWLEDGMENT
Um especial ag adecimen o ao meu ma ido e amigos pela comp eensão du an e es e pe cu so.
Ao P o esso Dou o Vi o San os e à P o esso a Dou o a Ca olina San os exp esso a minha
g a idão pelo apoio, comp eensão e pa ilha de conhecimen os pa a o desen ol imen o e
ealização des a ese de Mes ado.
“Que o igno ado, e calmo
Po igno ado, e p óp io
Po calmo, enche meus dias
De não que e mais deles.
Aos que a iqueza oca
O ou o i i a a pele.
Aos que a ama ba eja
Embacia-se a ida.
Aos que a elicidade
É o sol, i á a noi e.
Mas ao que nada espe a
Tudo que em é g a o.”
Fe nando Pessoa
iii
ABSTRACT
Fo he pas ew yea s, despi e he Po uguese go e nmen announcing he upg ade o he
Po uguese heal hca e sys em, i is possible o e i y he missing in es men in ha
in as uc u e, equipmen and se ices. The lack o quali y and he una ailabili y o mode n
equipmen s and in as uc u es a e some o he usual complain s o he heal hca e p o ide s
and pa ien s.
This s udy aims o analyze and explain, he gaps be ween inno a i e echnologies a ailable in
he ma ke and he p esen implemen ed echnologies in he Medical Imaging a ea in Po ugal
and, wi h he collec ed da a, build an explana o y model o he mo i a ions o he lack o
in es men .
Fo he collec ion da a pu poses, 102 subjec s we e inqui ed, om mul iple Po uguese
hospi als and heal hca e clinics ela ed o medical imaging, wi h physicians, supplie s, se ice
p o ide s, and Senio Diagnos ic and The apeu ic Technicians oles. A e analyzing he da a, i
is possible o conclude ha Po uguese medical imaging se ices need upda ed equipmen and
imp o ed echnology, mo e human esou ces, and aining. Manage s o heal hca e
o ganiza ions, mainly om he public sec o , a e mo i a ed p ima ily o sa e cos s due o he
inc easing heal hca e o ganiza ions' expendi u es. These ac s a e incu ing in w ong diagnoses,
pa ien dissa is ac ion, and lack o con idence in Po uguese heal hca e p o ide s.
KEYWORDS
Technologies; Compu a ional; Medical Imaging; Inno a ion; Heal hca e; Po ugal; Se ices
i
CONTENTS
1. In oduc ion ................................................................................................................1
1.1. Backg ound...................................................................................................................1
1.2. Mo i a ion/Jus i ica ion ...............................................................................................2
1.3. Objec i e ......................................................................................................................3
1.4. S udy ele ance and impo ance ..................................................................................3
2. Me hodology ...............................................................................................................5
2.1. Sys ema ic li e a u e e iew me hodology ..................................................................7
3. Li e a u e Re iew ........................................................................................................9
3.1. Medical Imaging echniques .........................................................................................9
3.1.1. Radiog aphy..........................................................................................................9
3.1.2. Angiog aphy .......................................................................................................10
3.1.3. Compu ed omog aphy ......................................................................................10
3.1.4. Ul asonog aphy .................................................................................................10
3.1.5. Magne ic Resonance Imaging .............................................................................11
3.1.6. Nuclea medicine ................................................................................................12
3.1.7. Mammog aphy ...................................................................................................12
3.2. A i icial In elligence in Medical Imaging ....................................................................13
3.2.1. A i icial In elligence ...........................................................................................13
3.2.2. Machine Lea ning ...............................................................................................14
3.2.3. Deep Lea ning .....................................................................................................14
3.3. Sys ema ic li e a u e e iew esul s p esen a ion ......................................................15
3.3.1. Sys ema ic li e a u e e iew esul s analysis ......................................................16
4. Cu en echnology and Inno a i e echnology su ey ................................................ 19
4.1. Su ey ques ions ........................................................................................................21
4.2. Da a collec ion and da a analysis p ocedu e ..............................................................22
5. Resul s p esen a ion and discussion ........................................................................... 23
5.1. Resul s p esen a ion ...................................................................................................23
5.1.1. Subjec cha ac e iza ion .....................................................................................23
5.1.2. Medical Imaging se ices imp o emen necessi ies ...........................................26
5.1.3. Assessmen o he in en ion o implemen inno a i e echnologies in he medical
imaging se ices ................................................................................................................28
5.2. Resul s discussion .......................................................................................................31

5.3. An in e p e a ion Model ............................................................................................33
6. Conclusion ................................................................................................................. 35
6.1. Limi a ions ..................................................................................................................35
6.2. Recommenda ions o u u e wo k ............................................................................35
REFERENCES ...................................................................................................................... 36
APPENDIX.......................................................................................................................... 40
i
LIST OF TABLES
Table 1 - Me hodology. ................................................................................................................5
Table 2 - Keywo ds. ......................................................................................................................7
Table 3 - In o ma ion sou ces. .....................................................................................................7
Table 4 - Resul s o a icle iden i ica ion. ...................................................................................15
Table 5 - Inclusion and exclusion c i e ia. ..................................................................................15
Table 6 - P isma eco ds. ............................................................................................................16
Table 7 - Su ey ques ions. ........................................................................................................21
Table 8 - Numbe and pe cen age o answe s pe ci y/ egion. ..................................................23
LIST OF FIGURES
Figu e 1 - P isma low diag am. ....................................................................................................8
Figu e 2 - In e p e a ion model o he e ec s caused by he non-in es men in inno a i e
echnologies in MI. .....................................................................................................................33
LIST OF CHARTS
Cha 1 - Map wi h he pe cen age o answe s pe ci y. ............................................................23
Cha 2 - Numbe o esponses pe sec o . ................................................................................24
Cha 3 - Numbe o esponses pe ole. ....................................................................................24
Cha 4 - Medical imaging echniques mos used by he in e iewed g oup. ............................25
Cha 5 - knowledge o compu ed echnology usage by he inqui ed g oup. ............................25
Cha 6 – Numbe o answe s o assess he need o imp o emen on medical imaging se ice in
Po ugal. .....................................................................................................................................26
Cha 7 – Numbe o esponses and pe cen age pe o al answe s o classi ica ion o he
ec i ica ion needed on medical imaging equipmen in Po ugal. .............................................26
Cha 8 – Numbe o esponses and pe cen age pe o al answe s o necessi ies o equipmen
imp o emen . ............................................................................................................................27
Cha 9 - Pe cen age o esponses o he need o imp o emen in medical imaging se ices.
...................................................................................................................................................27
Cha 10 - Pe cen age o answe s o he in en ions o inco po a e inno a i e echnologies
despi e he associa ed cos s. .....................................................................................................28
Cha 11 – Pe cen age o answe s o he in en ion o inco po a e inno a i e echnologies in i e
yea s. ..........................................................................................................................................28
Cha 12 - Ba ie s o compu a ional echnologies in eg a ion heal hca e o ganiza ions in
Po ugal. .....................................................................................................................................29
Cha 13 - Men ioned ba ie s by he inqui ed g oup abou he mo i a ions o he non-
inco po a ion o compu a ional echnologies in medical imaging se ices in Po ugal. ............30
Cha 14 - Compu a ional echnologies weekly used. ................................................................30
ii
LIST OF ABBREVIATIONS AND ACRONYMS
AI
A i icial In eligence
CADe
Compu e -aided de ec ion
CADx
Compu e -aided diagnosis
CAT
Compu ed axial omog aphy
CNN
Con olu ional neu al ne wo k
CT
Compu ed Tomog aphy
DL
Deep Lea ning
EDPSO
Enhanced Da winian Pa icle Swa m Op imiza ion
MRI
Func ional magne ic esonance imaging
GDP
G oss Domes ic P oduc
IP
Image pla e
MI
Medical imaging
ML
Machine Lea ning
MRI
Magne ic esonance imaging
PACS
Pic u e a chi e and communica ions sys ems
PET
Posi on Emission Tomog aphy
PRISMA
P e e ed Repo ing I ems o Sys ema ic Re iews and Me a-Analyses
PSO
Pa icle Swa m Op imiza ion
SPECT
Single Pho on Emission Compu ed Tomog aphy
US
Ul assonog aphy
1
1. In oduc ion
1.1. Backg ound
Heal hca e and echnology, as we know i oday, ha e been wo king oge he o many decades.
A his momen , he echnological e olu ion in heal hca e has a di ec impac on heal hca e
ins i u ions. In such a way, ha is possible o comp ehend he p ocesses' in eg a ion, he
secu i y imp o emen , he da a ea men and he medical p ac ices in hospi als and clinics, and
he usage o equipmen and se ices o diagnosis and ca e (Coccia, 2020).
In he las 40 yea s, i is possible o e i y ha Po ugal made signs o p og ess in heal hca e
ea men s and esou ces, such as echnologies applied o heal h. In his pe iod, Po ugal
eached a highe le el in i s heal h sys em, p esen ing a dec ease o 94% o he G oss Mo ali y
Ra e, compa able wi h he main in e na ional pa ne s ( ime window om 1970 un il 2008)
(Ei a, 2010).
In such a way ha , in P og ama XXI om Go e no Cons i ucional 2015-2019, he Po uguese
go e nmen made he commi men o "Mode nize and in eg a e in o ma ion echnologies and
he exis ing ne wo k o keep he old and sick people mos o he ime in a amilia en i onmen ,
de eloping elemoni o ing and elemedicine" (Go e no de Po ugal, 2014).
Despi e he e olu ion and de elopmen in heal h echnology, he ime be ween he echnology
being deli e ed o he ma ke and he in oduc ion o he Po uguese Na ional Heal hca e
Sys em o Po ugal (SNS) ha e a signi ican disc epancy.
Has Cab al men ioned in 2002, he pa ien s a e unsa is ied wi h he se ices p o ided by public
heal hca e ins i u ions despi e he inc ease in heal h cos s abo e in la ion. He was able o e i y
ha only 50,3% o he pa ien s we e sa is ied wi h he heal hca e se ices p o ided in an
eme gency se ice (Cab al, 2002). The d i es o he dissa is ac ions we e he inadequa e
se ice, he ou da ed hospi al in as uc u es, he lack o means o diagnosis, he incapaci y o
communica ion p esen ed by he se ice p o ide wi h he pa ien , o he lack o adequa e
in o ma ion (Figuei edo, 2008).
As denounced o jou nal Diá io de No ícias by Alexand e Lou enço, p esiden o Po ugal
Associa ion o Hospi al Adminis a o ’s, alleges ha he lack o echnology in es men since
2009 esul ed in a isk o heal hca e echnology obsolescence (Nunes, 2019).
Also, a s udy ca ied ou by Deloi e in 2011 e eals ha he inancial unsus ainabili y o he
heal hca e sys em in Po ugal is one o he main p oblems men ioned by heal hca e
s akeholde s (Deloi e, 2011). O e he las ew yea s, i is possible o obse e he inc ease o
he p i a e and public gene al expendi u e in heal hca e, ha ing he g ow h o public
expendi u e exceeded he g ow h a e o GDP (G oss Domes ic P oduc ).
In compa ison wi h o he Eu opean coun ies, i is possible o e i y ha public expendi u e in
a o al o heal hca e expenses in Po ugal is low, which e eal ha ha e an addi ional
expendi u e in heal hca e suppo ed by he Po uguese popula ion, no e lec ed in axes
(Deloi e, 2011), inc easing he sea ch o p i a e heal hca e se ices.
8
Figu e 1 - P isma low diag am.

9
3. Li e a u e Re iew
The in luence o medical imaging in heal hca e is inc easing. Wi h medical imaging, heal hca e
echnicians can de ec diseases ea lie and ea pa ien s wi h e ec i eness. When hei
applica on is ocused on p e en ion and he apy, his can con ibu e o educing heal hca e
cos s on a global scale.
Medical Imaging consis s o an impo an medical special y ha uses in e nal body image o
ealize diagnosis, ha is, apply echniques and p ocesses o c ea e human body images o clinical
diagnosis and ea diseases.
3.1. Medical Imaging echniques
Medical imaging uses ou ypes o adia ion o ob ain medical images, X- ays, pho ons,
adio equency, and ul asound. Radiog aphy and compu ed omog aphy use x- ay adia ion,
and nuclea medicine use pho ons, being bo h in asi e echniques used o ob ain body images.
Ul asonog aphy and magne ic esonance imaging a e non-in asi e echniques applying
ul asound and adio equency, espec i ely, in o he body o ob ain he images. Mo e de ails
a e speci ied below on how each echnique wo ks and deli e s he medical images.
3.1.1. Radiog aphy
The medical imaging echnique o adiog aphy, usually called X- ay, consis s o an image
diagnosis exam applying adia ion X in o he pa ien 's body and ob aining a plane image. The
image is p oduced by he adia ion X ha ollows he ela ionship be ween he emi ed ays and
he ecei ed ays a e passing o he body (B an , 2019; Webb, 2003).
To ob ain he image is used an X- ay beam, p oduced by a gene a o , will pass h ough he body
sec ion ha is p e ended o analyze. Pa o he X- ay ene gy is abso bed by he body, depending
on he densi y o he s uc u e, and he emaining ene gy goes o he ilm o de ec o , p oducing
he adiog aphy (B an , 2019; Webb, 2003).
The ob ained image ep oduces he di icul y o he X- ay passing h ough he body. The bone,
which is he dense abso bs he adia ion, is shown in he image in a clea one. Howe e , he
s uc u es wi h lowe densi y a e shown in a da ke one, o example, he ai in he lungs (B an ,
2019; Webb, 2003).
Con en ional adiog aphy o ob ain he X- ay images uses a ilm composed o sil e c ys als ha
a e s imula ed by adia ion. In he p ocess o e ela ion, i is equi ed o use mul iple chemical
solu ions ha will p o ide di e en g ey ones o he ilm. This p ocess equi es high ime o
ob ain he inal ilms, he usage o pollu an chemicals, and highe doses o adia ion, being
discon inued and eplaced by digi al adiog aphy (B an , 2019; Webb, 2003).
Nowadays, he usual p ocedu e is digi al adiog aphy and di ec digi al adiog aphy. In digi al
adiog aphy, he image is ob ained by a compu e using a eadable and eusable phospho
display ha is s imula ed by adia ion and called an image pla e (IP). Ano he p ocedu e o
ob ain he image is using di ec digi al adiog aphy, whe e he image is ob ained di ec ly in he
compu e . In his case, he IP is eplaced by a de ec o di ec ly connec ed o he compu e ha
ecei es he da a and p ocesses hem. These me hodologies allow he pos -p ocessing o he
10
images, being an added alue o echnicians and doc o s. Digi al adiog aphy acili a es he
sha ing and a chi e o he exams, imp o ing access o old images, compa ing images, and access
by in e nal o public ne wo ks (B an , 2019; Webb, 2003).
3.1.2. Angiog aphy
The Angiog aphy echnique deli e s images ha show he blood essels selec i ely in he body.
This ype o imaging is used o examine diseases such as clo ing o a e ies and eins and
i egula i ies in sys emic and pulmona y blood low. In X- ay angiog aphy, an iodine-based
con as agen is injec ed in o he bloods eam be o e imaging. The X- ay image shows inc eased
a enua ion om he blood essels compa ed o he issue su ounding hem (B an , 2019;
Webb, 2003).
3.1.3. Compu ed omog aphy
Compu ed omog aphy is used o suppo physicians in diagnosing mul iple diseases and
assessing ea men s. This me hod uses ionic adia ion, simila o con en ional adiog aphy,
wi h complex algo i hms and is suppo ed by compu e sys ems o p oduce he images (B an ,
2019; Webb, 2003).
This echnology, in he beginning, was denomina ed as compu ed axial omog aphy (CAT)
because he echnique was used o p oduce axial images, he gan y su ounded he body,
p oducing one image pe cycle. Be ween cycles, he bed wi h he pa ien mo ed millime e s o
cen ime e s successi ely. La e , his echnology e ol ed o helical, allowing he con inuing bed
mo emen and con inuing gan y o a ion, being called compu ed omog aphy (CT) (B an ,
2019; Webb, 2003).
The numbe o images o slices p oduced is p opo ional o he numbe o gan y de ec o s.
Each slice ma ches a slice o he body in he s udy, and he hickness o slices is p ede ined by
he echnician (B an , 2019; Webb, 2003).
Compu ed omog aphy enables he analysis o ascula sys em injec ing con as in he pa ien ,
complemen ing he s udy, usually called SPECT de ailed in chap e 3.1.6. The con as p oduc s
a e iodized, inc easing he s uc u es con as wi h di e en blood lows, accen ua ing umo s
o in lamma ions (B an , 2019; Webb, 2003).
3.1.4. Ul asonog aphy
Ul asonog aphy (US) imaging is pe o med by a pulse-echo echnique, whe e he ansduce
con e s elec ical ene gy o a b ie pulse o high- equency sound ene gy ansmi ing hem
in o he body. The US ansduce is as well he ecei e , de ec ing he echoes o he sound
ene gy e lec ed om issues, assuming ha he e u ning echoes as he sou ce om along he
line o he sigh o he ansmi ed pulse. The dep h o he echo is de e mined by measu ing he
ound- ip ime o he ligh o he ansmi ed pulse and he e u ning echo, calcula ing he
dep h o he e lec ing issue in e ace, assuming an a e age speed o he sound in he issue o
1,540 m/s. The image is ob ained om mul iple US pulses, and he shape and appea ance o he
esul ing image depend on he design o he ansduce used, sec o , o linea a ay (B an , 2019;
Webb, 2003).
11
US examina ion is achie ed by applying he US ansduce on o he pa ien 's skin using he
wa e -soluble gel as a linking agen o ensu e good con ac and ansmission o he US beam.
The images a e ob ained in any ana omic plane by adjus ing he o ien a ion and angula ion o
he ansduce and he pa ien posi ion. The axial, sagi al, and co onal planes p o ide he
easies ecogni ion o ana omy bu may no be he bes o demons a e all ana omic s uc u es,
and he quali y o he US depends hea ily on he sonog aphe 's skill and diligence. The bes way
o gua an ee he diagnosis is by p o iding guidelines o he sonog aphe (B an , 2019; Webb,
2003).
Dopple ul asonog aphy is an impo an sys em o eal- ime g ay-scale ana omic imaging. The
dopple e ec consis s o a change in he equency o he e u ning echoes, compa ed wi h he
ansmi ed pulse caused by he e lec ion o he sound wa e om he mo ing objec . In medical
imaging, he mo ing objec s o in e es a e he ed blood cells in lowing blood. I he blood low
is away om he ansduce , he echo equency will be shi ed lowe . On he con a y case, i
will be highe . The amoun o he equency shi is p opo ional o he compa a i e eloci y o
he mo ing ed blood cells (B an , 2019; Webb, 2003).
Duplex dopple combines eal- ime g ay-scale imaging wi h dopple o de e mine he posi ion
o he sample olume in isualized blood essels o speci ic a eas in analyzes (B an , 2019; Webb,
2003).
3.1.5. Magne ic Resonance Imaging
Magne ic esonance (MRI) is a non-in asi e echnique ha uses magne ic ields and adio wa es
o p oduce de ailed h ee-dimensional images. In compa ison wi h CT, which analyzes only a
single issue pa ame e , as he x- ay a enua ion, MR analyzes mul iple issue cha ac e is ics,
including hyd ogen (p o on) densi y, he T1 and T2 elaxa ion imes o issue, and blood low
wi hin he issues (B an , 2019; Webb, 2003).
The analyzes o he so issues a e subs an ially be e wi h MR han o he imaging echniques,
whe e he densi y di e ence o he p o ons a ailable con inue o he MR signal di e en ia e
one issue o ano he . The T1 measu emen consis s o he measu e o he p o on's abili y o
exchange ene gy wi h i s su ounding chemical ma ix. The T2 measu emen is he o how o
as a gi en issue loses i s magne iza ion. Mos o he issues can be iden i ied by he di e ence
in hei T1 and T2 elaxa ion imes (B an , 2019; Webb, 2003).
MR echnique is based on he abili y o a small numbe o pho ons wi hin he body o abso b
and emi adio wa e ene gy when he body is placed wi hin a s ong magne ic ield. The capaci y
o abso b and elease adio wa es is di e en a di e en , de ec able, and cha ac e is ic a es
o di e en issues (B an , 2019; Webb, 2003).
In he case o b ain analysis, is usually used unc ional magne ic esonance imaging ( MRI)
analyze he ac i i y o b ain a eas a e a ied and di e en s imuli. This echnique is used in
localized inju ies o he ne ous sys em and epilepsy sou ce loca ion (B an , 2019; Webb, 2003).
12
3.1.6. Nuclea medicine
Nuclea medicine is a medical imaging echnique ha uses adiopha maceu icals ( adioac i e
ace s) o diagnose by images and o apply molecula he apeu ic. To ack he pa h o he
adioac i e ace s is used compu a ional omog aphy echnology, which allows o scan mo e
clea ly umo and moni o he g ow h, as an example (B an , 2019; Webb, 2003).
Radioac i e ace s a e ca ie molecules wi h a s ong bond o a adioac i e a om. The ca ie
molecules a y acco dingly wi h he pu pose o he scan. Some molecules in e ac wi h p o ein
o suga in he body, o hey can a ach adioac i e a oms o blood cells o ace he pa h o he
blood in cases o in e nal bleeding, using SPECT scan o collec he images (B an , 2019; Webb,
2003).
Usually, he adminis a ion o he adioac i e ace s can be done by in a enous injec ion.
Howe e , he adioac i e ace s can also be adminis a ed by inhala ion, o al inges ion, o
di ec injec ion in o an o gan. I depends on he objec i e and he disease ha is being s udied
(B an , 2019; Webb, 2003).
The mos common imaging modali ies in nuclea medicine a e SPECT (Single Pho on Emission
Compu ed Tomog aphy) and PET (Posi on Emission Tomog aphy). The di e ence be ween
hese wo modali ies is he ype o adia ion used. While SPECT scans measu e gamma ays, PET
measu es he ene gy deli e ed when he posi on in e ac s wi h elec ons in he body,
annihila ing each o he and p oducing a small amoun o ene gy in he o m o wo pho ons
shoo o in opposi e di ec ions. The de ec o s in bo h echniques use he in o ma ion o
measu emen s o c ea e images o in e nal o gans. The SPECT scans a e mos ly used o diagnose
and ack hea diseases when PET scan is o de ec cance and moni o i s p og ession,
ea men de elopmen , and me as ases de ec ion (B an , 2019; Webb, 2003).
3.1.7. Mammog aphy
Mammog aphy is a b eas exam based on ionic adia ion ha allows he b eas pa enchyma
analysis, de ec ing nodules, o clus e ing o mic ocalci ica ions. Digi al mammog aphy is a b eas
exam ha allows simple and as e image pos -p ocessing (B an , 2019; Webb, 2003).
Usually, i is known as a women's exam bu can be done in men. The quan i y o adia ion is low
and is he only alid exam o sc eening o b eas cance in he gene al popula ion (B an , 2019;
Webb, 2003).
13
3.2. A i icial In elligence in Medical Imaging
A i icial in elligence has been he echnology mos s udied and de eloped o e he pas ew
yea s o e ol e and suppo medical imaging se ices (Ting, 2018). Nowadays is possible o
obse e he applica ion o a i icial in elligence by suppo ing he adiologis a hei wo kload
and imp o ing he diagnosis, inc easing medical images esolu ion, and suppo ing pa ien ca e.
Machine lea ning and deep lea ning, as subclasses o a i icial in elligence, ha e been de eloped
o suppo adiologis s in decision-making, lesions de ec ion, and pe o ming adiologis s’ asks
wi hou human in e en ion. These AI subclasses ha e a huge da a need a ba ie , along wi h
pa ien s and images da a p o ec ion, making di icul his echnologies de elopmen .
The nex sub-chap e s a e desc ibed he main de elopmen in hese echnologies applied o
medical imaging.
3.2.1. A i icial In elligence
A i icial in elligence, as desc ibed by O en and Pesapane, is a compu e science ha uses
compu a ional algo i hms o dissec complica ed da a c ea ing sys ems ha pe o m asks, ha
usually equi es human in elligence, di iding in mul iple echniques (O en, 2020; Pesapane,
2018). In o he wo ds, i is a compu e science ha ocuses on sol ing asks ha people can
execu e (F ockman, 2019).
The applica ion o AI o medical imaging is one o he mos p omising a eas in heal hca e
inno a ions, bu i s applica ion goes beyond image p ocessing and in e p e a ion. F om image
acquisi ion and p ocessing he epo , ollow-up planning, da a s o age, da a mining, among
o he s. Due o he wide ange o applica ions, i s expec ed a massi e impac on adiologis s daily
li e due o AI (Pesapane, 2018).
Amongs he AI applica ions o image in e p e a ion, he de ec ion is he mos s udied,
ollowed by he ule ou , diagnoses, and de ini ion o he s age o he disease in he medical
images. The de elopmen and applica ion o AI algo i hms in medical imaging, in gene al, need
o ollow h ee-s ep: aining, alida ion, and es ing. A e he AI algo i hm lea ns he ea u es,
hen i can be used o assis clinical p ac ices, educing diagnos ic e o s, p o iding p edic ions,
and educing he in e p e a ion ime o medical images, in compa ison wi h he human eade
in e p e a ion ime (Obuchowski, 2019).
Usually, AI algo i hms ha e he ollowing ou p ac ical applica ions in medical imaging. The i s
one is he eade mode, when an image is applied o an AI algo i hm, and he anomalies a e also
e iewed by a human eade du ing he image in e p e a ion. The second applica ion is he
second eade mode, he applica ion o he AI algo i hm a e he human eade pe o med his
in e p e a ion, o include addi ional indings, ha ing in conside a ion he human eade
in e p e a ion. The hi d applica ion is he iage mode, when he AI algo i hm classi ies
suspicious de ec ions in he images. The las one is he p esc eening mode, when he AI
algo i hm ocuses on cleaning images and p o ides a epo , gi ing only he images wi h
anomalies o he human eade . The AI algo i hm is mos used as he second eade mode as
compu e -aided de ec ion (CADe) applied o de ec b eas , lung, and colon cance imaging,
helping he human eade o ind lesions in he images (Obuchowski, 2019).

14
Nowadays, AI o en de ec s mino image al e a ions wi h mo e ele an ou come a iables,
including a new diagnosis o ad anced disease ha equi es ea men o ha will a ec long-
e m pa ien li e. The bigges change ha AI b ings o medical imaging is he de ec ion o pa e n
changes ha a e no easily de ec ed by humans, showing imp essi e accu acy and sensibili y.
Howe e , he imp o ed sensibili y b ings p oblems by de ec ing sub le changes wi h
inde e mina e signi icance. In he case o sc eening mammog ams, he a i icial neu al ne wo ks
a e no mo e accu a e han adiologis s in de ec ing cance bu ha e mo e sensibili y o
pa hological indings, in pa icula o sub le lesions (O en, 2020).
One s udy pe o med in 2019 showed ha an AI algo i hm (deep lea ning-based algo i hm –
chap e 3.2.2.) was able o dis inguish ou di e en diseases in ches adiog aphy, malignan
pulmona y neoplasm, ac i e pulmona y ube culosis, pneumonia, and pneumo ho ax (Hwang,
2019).
AI includes machine lea ning (ML) and deep lea ning (DL) ha can help c ea e sys ems o
suppo heal h echnicians o pe o m hei daily asks, wi h as e deli e y and inc easing alue
on he imaging analyses (Alexande , 2020). ML and DL can gi e physicians a complemen o
enable he de elopmen o new ea men s, which will be de ailed in chap e s 3.2.2 and 3.2.3.
3.2.2. Machine Lea ning
Machine lea ning (ML) is a sub-b anch o a i icial in elligence, as an elemen o compu e
science ha aims o build and le e age exis ing algo i hms, usually cen e ed on ma hema ical
and s a is ical op imiza ion, in o de o launch gene alized models ha o e pa e ns and
accu a e p edic ions. Meaning ha ML ocuses on con inuous da a imp o emen , da a mining,
and ask au oma ion (F ockman, 2019).
One applica ion o ML in medical imaging is he analyzis o MRI b ain images o de ec b ain
changes ela ed o ea ly ischemic s oke wi hin a na ow ime window wi h g ea e sensi i i y
han a human eade (O en, 2020).
Ano he capabili y in ML is he abili y o dis inguish Pa kinson's disease om a ypical
pa kinsonism and mul iple sys ems a ophies om p og essi e sup anuclea palsy in
conjunc ion wi h he MRI echnique. To dis inguish he diseases, ML, in conjunc ion wi h MRI,
measu es he ee-wa e -co ec ed ac ional aniso opy (FA) o de elop disease-speci ic
machine lea ning compa isons o di e en ia e o ms o pa kinsonism (Abel, 2019).
As men ioned by Chassagnon G., ML has mo e ele ance in image in e p e a ion. On he o he
hand, deep lea ning has shown an ou pe o mance o a speci ic class o p oblems (Chassagnon,
2020).
3.2.3. Deep Lea ning
Deep lea ning (DL) is de ined as a subse o ML whe e he compu e s lea n use ul
ep esen a ions and ea u es au oma ically, di ec ly om he aw da a (La i , 2017). This abili y
was allowed by he inc easing a ailabili y o la ge da ase s and he high compu ing capaci y o
g aphic p ocessing uni s. In medical imaging, he class o deep lea ning ne wo ks mos used and
s udied is he con olu ional neu al ne wo k (CNN) algo i hm (Chassagnon, 2020; Lunde old,
2019).
15
CNN algo i hm is used o sol e compu e ision asks, as hand-w i en numbe s on bank checks
(Hwang, 2019). In medical imaging, CNN ep esen s a powe ul way o lea n use ul
ep esen a ions o images and o he s uc u ed da a. CNN can be applied o imp o e he
e iciency o adiology p ac ices h ough p o ocol de e mina ion based on sho - ex
classi ica ion and also be used o educe he gadolinium dose in con as -enhanced b ain MRI
wi hou comp omising he image quali y. Ano he ele an applica ion is he image egis a ion
o ad anced de o mabili y, allowing he quan i a i e analysis ac oss he ime o di e en images
(Lunde old, 2019).
3.3. Sys ema ic li e a u e e iew esul s p esen a ion
The sys ema ic li e a u e e iew was execu ed in May 2020 ollowing he me hodology
desc ibed a chap e 2.1.
In he i s phase, he iden i ica ion phase, i was possible o iden i y 55878 documen s (Table
4).
Table 4 - Resul s o a icle iden i ica ion.
In o ma ion sou ces
Iden i ied s udies
Science Di ec
n = 54517
Scopus
n = 939
Web o Science
n = 422
In he second phase, he abs ac sc eening was de ined and applied he exclusion and inclusion
c i e ia as desc ibed below in Table 5:
Table 5 - Inclusion and exclusion c i e ia.
Inclusion C i e ia
Exclusion C i e ia
Scien i ic Documen s
Publica ion da e in e io 2015
In es iga ion epo
Publica ion language
Lack o access o he documen
Duplica e eco ds
Thesis
Themes no ele an o his s udy
Conside ing he inclusion c i e ia o his s udy, he e we e included scien i ic documen s and
hesis ha a e di ec ly ela ed o inno a i e compu a ional echnology applied o medical
imaging, used o imp o e he se ice p o ided o pa ien s, and suppo heal hca e echnicians
on a daily basis.
As an exclusion c i e ia, publica ions be o e 2015 we e excluded due o he ac o as e olu ion
o he compu a ional echnologies in ecen yea s, and compu a ional echnologies wi h mo e
han 5 yea s s a o be conside ed ou da ed. A icles no published in Po uguese o English,
lack access o he epo , and duplica ed eco ds we e as well excluded.
16
A e he emo al o duplica ed eco ds, he selec ion o he eligibili y a icles ( hi d phase) was
pe o med in he ollowing s eps:
i) Fi s , all abs ac s wi hou ele ance and clea ly ou side o he s udy scope we e
emo ed.
ii) Then, he abs ac s o he e ie ed a icles we e analyzed i he e we e ela ed o
inno a i e compu a ional echnologies applied o medical imaging.
iii) A e wa ds, he abs ac s o he con inuing a icles we e e alua ed acco dingly o he
de ined inclusi e and exclusi e c i e ia.
i ) Finally, he comple e ex o he a icles was ully analyzed.
In he p ocess o comple e ex analysis, a icles ha epo o e iews o e iews, wi hou
ci a ions o incomple e a icles ha a e no ela ed o he aim o his s udy, we e as well
excluded.
Table 6 - P isma eco ds.
Exclusion c i e ia
SCIENCE DIRECT
SCOPUS
WEB OF SCIENCE
1s phase
Excluded
a icles
To al
Excluded
a icles
To al
Excluded
a icles
To al
A icles iden i ica ion
N/A
54517
N/A
939
N/A
422
2nd phase
Excluded
a icles
To al
Excluded
a icles
To al
Excluded
a icles
To al
Publica ion da e in e io 2015
24885
29632
543
396
189
233
In es iga ion epo
9000
20632
214
182
86
147
Publica ion language
309
20323
8
174
1
146
Lack o access o he documen
15782
4541
146
28
76
70
Ti le and abs ac
4527
14
24
4
66
4
3 d phase
Excluded a icles
To al
Duplica ed a icles
1
21
4 h phase
Excluded a icles
To al
Full-Sc een a icle
10
11
In he included phase (4 h phase), a e pe o ming he ull-sc een o he 21 selec ed a icles, 10
a icles we e excluded, selec ing 11 o hem as ele an a icles o his s udy.
3.3.1. Sys ema ic li e a u e e iew esul s analysis
The main conclusion a e he analysis o he selec ed a icles is ha MRI, CT, and he US a e he
mos s udied medical imaging a eas o embody inno a i e compu a ional echnologies. AI, ML
and DL a e he mos ele an compu a ional echnologies, specially CNNs algo i hm, being he
DL mos ocused in medical imaging.
17
AI implemen a ion in heal hca e se ices is one o he mos desi ed imp o emen s o e icacy
and e iciency o clinical ca e. AI has he capabili y o ecognize pa e ns in images non-
de ec able by humans and suppo physicians' decision-making, based on he la ge olume o
daily gene a ed da a and suppo ing he inc easing demand o heal hca e se ices. The
agg ega ion o mul iple da a in o in eg a ed diagnos ic sys ems b idge medical images,
genomics, pa hology, elec onic heal h eco ds, and social ne wo ks, being some o he
imp o emen s in heal hca e se ices p o ided by AI (Bi, 2019).
Mul iples in es iga o s, since he 1980s, ha e been de eloping ML echniques o CADx
(compu e -aided diagnosis) wi h he objec i e o dis inguish be ween malignan and benign
b eas lesions. CADx consis s o an AI me hod ha au oma ically cha ac e izes he umo by size,
shape, mo phology, ex u e, and kine ics. CADe has been de eloped, as well, o au oma e b eas
lesions de ec ion by MRI (Bi, 2019).
US is he medical imaging echnique ha has he abili y o p oduce eal- ime ideo, p o iding
mo e in o ma ion han a single image ame. To imp o e his echnique, ML has been applied
o he US o allow he classi ica ion o CADx, issue segmen a ion, image egis a ion, and
con en e ie al, being ha he compu e -aided disease diagnosis and classi ica ion consis s o
de ec ing o classi ying lesions. This app oach has been aken mainly in b eas and li e analysis
and has been imp o ed g ea ly om he ecen ad ances in ML (B a ain, 2018).
Ma aci p oposed an au oma ed amewo k o e al de ec ion and p egnancy iabili y, meaning
e al ca diac ac i i y, om a p ede ined ee-hand US sweep o he ma e nal abdomen. This
amewo k was de eloped o add ess he need o empowe bo h he less expe ienced and he
well- ained use s o ul asound. A cen al p oblem o his p oposal was he need o high
deg ee skills o pe o m a well-execu ed ul asonog aphy exam, and he amewo k p oposed
ob ained 83.4% accu acy (Ma aci, 2017).
As men ioned in sec ion 3.2.3., CNN's algo i hm is he mos s udied DL algo i hm applied o
medical imaging, being p oposed o segmen b eas umo s (Caballo, 2020), de ec ion o b ain
anomalies and diseases (Rai, 2020; Go apu, 2018; Nunes, 2019; Coccia, 2020), and p os a e
cance de ec ion and diagnosis (Bi, 2019). The main eason o his applica ion in medical
imaging segmen a ion is due o he ac o unp eceden ed accu acy in classi ying and de ec ing
objec s, ace de ec ion, and segmen a ion in he p ocess o classi ying images (B a ain, 2018).
MRI, as desc ibed in sec ion 3.1.5., plays an essen ial ole in he analysis o many diseases and
condi ions. In he beginning, CNN algo i hms we e used o medical image analysis, bu hei
pe o mance mo i a ed he esea che s o de elop many new CNN a chi ec u es o suppo MRI
image segmen a ion. Image segmen a ion e e s o he p ocess o di iding images in o mul iple
egions wi h simila p ope ies. Go api p oposed a new CNN applica ion, mo e speci ically
DenseNe a chi ec u e, o b ain segmen a ion in o a numbe o classes based on he ype o
issue. The esul s demons a e ha he applied app oach can accu a ely pe o m he
segmen a ion and, in he nea u u e, suppo physicians in Pa kinson MRI segmen a ion
(Go apu, 2018). O he algo i hms s udied and applied o suppo MRI image segmen a ion a e
EDPSO (Enhanced Da winian Pa icle Swa m Op imiza ion) algo i hm and PSO (Pa icle Swa m
Op imiza ion) algo i hm (Vijay, 2016). The usage o DL in 2D image segmen a ion o b eas
masses de ec ed in CT images wi hou con as go a high pe o mance, wi h high classi ica ion
24
Sec o
Almos hal o he inqui ed g oup is employed a he p i a e sec o (46%), he o he hal wo ks
o he public and p i a e sec o (30%) o he public sec o (22%)—only 2% o he g oup wo ks
o he public-p i a e pa ne ships heal hca e ins i u ions.
P o essional quali ica ion
Senio Diagnos ic and The apeu ic Technician ep esen s he highe pe cen age o he inqui ed
g oup (75%), ollowed by he se ice p o ide s (13%), he supplie s (10%), and he physicians
(2%).
2
22
31
47
2%
22%
30%
46%
0%
5%
10%
15%
20%
25%
30%
35%
40%
45%
50%
0
5
10
15
20
25
30
35
40
45
50
Public-p i a e
pa ne ship
Public Public and P i a e P i a e
% pe o al answe s
Nº o answe s
Sec o
Nº o answe s % pe o al answe s
210 13
77
2% 10%
13%
75%
0%
10%
20%
30%
40%
50%
60%
70%
80%
0
10
20
30
40
50
60
70
80
90
Physician Supplie Se ice p o ide Senio Diagnos ic and
The apeu ic
Technician
% pe o al answe s
Nº o answe s
P o essional quali ica ion
Nº o answe s % pe o al answe s
Cha 3 - Numbe o esponses pe ole.
Cha 2 - Numbe o esponses pe sec o .

25
Medical Imaging echniques
To he ques ion “Wha a e he medical imaging echniques you usually wo k wi h?” i was
possible o conclude ha CT, adiog aphy, mammog aphy, and MRI a e he mos common
medical imaging echniques o he inqui ed g oup, ep esen ing 84% o he answe s.
Cha 4 - Medical imaging echniques mos used by he in e iewed g oup.
Knowledge o compu a ional echnology usage
To he ques ion “In which medical imaging echniques you know o ha e implemen ed
compu a ional echnologies?” he mos e e enced echniques a e CT, adiog aphy, MRI, and
mammog aphy, wi h 66% o he inqui ed answe s.
Cha 5 - knowledge o compu ed echnology usage by he inqui ed g oup.
1%
2%
2%
5%
6%
16%
18%
23%
27%
0% 5% 10% 15% 20% 25% 30%
SPECT
PET
US
MRI
Angiog aphy
MRI
Mammog aphy
Radiog aphy
CT
% pe o al answe s
MI echniques
1%
5%
5%
6%
8%
9%
14%
16%
17%
19%
0% 5% 10% 15% 20% 25%
Wi hou knowledge
SPECT
US
PET
MRI
Angiog aphy
Mammog aphy
MRI
Radiog aphy
CT
% pe o al asnwe s
MI echniques
26
Need o imp o emen
Cha 4 ep esen s he answe s o he need o imp o emen in medical imaging se ices whe e
he subjec wo ks. 90,2% o he inqui ed say he medical imaging se ices need imp o emen s,
and only 9,8% o he answe s assess he excellen condi ions o he se ices.
5.1.2. Medical Imaging se ices imp o emen necessi ies
Classi ica ion o he eno a ion needed
So wa e and ha dwa e imp o emen necessi ies we e he highe epo ed, wi h 76 % o he
answe s, ollowed by 14% o so wa e imp o emen needed and 10% o ha dwa e
imp o emen .
10
91
9,8%
90,2%
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
0
10
20
30
40
50
60
70
80
90
100
No Yes
% pe o al answe s
Nº o answe s
Need o imp o emen
Nº o answe s % pe o al answe s
913
70
10% 14%
76%
0%
10%
20%
30%
40%
50%
60%
70%
80%
0
10
20
30
40
50
60
70
80
Ha dwa e So wa e So wa e and Ha dwa e
% pe o al answe s
Nº o answe s
Needs classi ica ion
Nº o answe s % pe o al answe s
Cha 6 – Numbe o answe s o assess he need o imp o emen on medical imaging se ice in Po ugal.
Cha 7 – Numbe o esponses and pe cen age pe o al answe s o classi ica ion o he ec i ica ion needed on
medical imaging equipmen in Po ugal.
27
Need o imp o emen in medical imaging equipmen a he MI se ice
The highe imp o emen necessi y is he ou da ed equipmen (43%) ollowed by he ou da ed
echnology (27%) in he medical imaging se ices. IT cen aliza ion and so wa e incompa ibili y
ep esen s only 30% o he o al answe s.
Cha 8 – Numbe o esponses and pe cen age pe o al answe s o necessi ies o equipmen imp o emen .
Needs o imp o emen in medical imaging se ices
The mos epo ed imp o emen in medical imaging se ices is he upda ed equipmen ,
ep esen ing 33% o he esponses. The necessi y o aining and upda ed so wa e has a simila
ep esen a i i y wi h 25% and 23%, espec i ely. Mo e human esou ces o he se ices
ep esen s 19% o he answe s.
Cha 9 - Pe cen age o esponses o he need o imp o emen in medical imaging se ices.
21
29
45
73
13%
17%
27%
43%
0%
5%
10%
15%
20%
25%
30%
35%
40%
45%
50%
0
10
20
30
40
50
60
70
80
IT cen aliza ion So wa e
incompa ibili y
Ou da ed echnology Ou da ed equipmen
% pe o al answe s
Nº o answe s
Imp o emen needs a MI se ices
Nº o answe s % pe o al answe s
19%
23%
25%
33%
Human Resou ces
Upda ed so wa e
T aining
Upda ed equipemen
28
5.1.3. Assessmen o he in en ion o implemen inno a i e echnologies in he
medical imaging se ices
The in en o he ins i u ions o inco po a e compu a ional echnologies despi e he
associa ed cos s
Despi e he associa ed cos s o he inno a i e echnologies implemen a ion on he medical
imaging se ices, 62% o he inqui ed conside ed possible o execu e. Howe e , 36% men ion
ha inco po a ion may be a possibili y, and only 2% answe ed ha he e is no possibili y o
implemen .
Cha 10 - Pe cen age o answe s o he in en ions o inco po a e inno a i e echnologies despi e he associa ed
cos s.
The in en o he ins i u ions o inco po a e compu a ional echnologies in i e yea s
Almos hal o he inqui ed g oup (42%) does no ha e ce ain y o he in en ion o implemen
compu a ional echnologies in hei se ices. The 36% answe ed a i ma i e o he will o
implemen , and 22% didn’ belie e in he inco po a ion.
Cha 11 – Pe cen age o answe s o he in en ion o inco po a e inno a i e echnologies in i e yea s.
62%
2%
36%
Yes
No
Maybe
36%
22%
42%
Yes
No
Maybe
29
Ba ie s o compu a ional echnologies
41% o he in e iewed g oup selec ed economic ba ie s as he main obs acle o
compu a ional echnologies inco po a ion, ollowed by o ganiza ional ba ie s wi h 27%, human
ba ie s wi h 14%, echnical ba ie s wi h 9%, and p ojec ba ie s wi h 8%.
Cha 12 - Ba ie s o compu a ional echnologies in eg a ion heal hca e o ganiza ions in Po ugal.
Reasons o non-inco po a ion o compu a ional echnologies
To he ques ion “Can you please speci y ba ie s o in eg a ing compu a ional echnologies?” i
was possible o agg ega e i e main ba ie s.
1º ba ie (61%): heal hca e ins i u ions a e ocused on inancial sa ings — he inco po a ion o
compu a ional echnologies comes wi h high cos s and big changes in he se ices.
2º ba ie (19%): physicians a e no a ailable o he changes hose compu a ional echnologies
will p o ide a he medical imaging se ices.
3º ba ie (10%): C-le el managemen o he heal hca e o ganiza ions ha e a high complex
decision-making p ocess, u ning he decision o he imp o emen in o a challenge.
4º ba ie (8%): Lack o us and knowledge by he heal hca e ins i u ions abou compu a ional
echnologies pe o mance, keeping he ocus on oday’s model o wo k (human analysis) ins ead
o compu a ional echnology analysis.
5º ba ie (2%): Un amilia i y o he op boa d o heal hca e o ganiza ions wi h compu a ional
echnologies and he bene i s do he medical imaging se ices.
8%
9%
14%
27%
41%
0% 5% 10% 15% 20% 25% 30% 35% 40% 45%
P ojec ba ie s
Technical ba ie s
Human ba ie s
O ganiza ional ba ie s
Economic ba ie s
% pe o al answe s
Ba ie s

30
Cha 13 - Men ioned ba ie s by he inqui ed g oup abou he mo i a ions o he non-inco po a ion o
compu a ional echnologies in medical imaging se ices in Po ugal.
Op ional ques ion
To he op ion ques ion “Which a e he compu a ional echnologies ha you wo k wi h weekly?”
mos o he inqui ed don’ ha e he knowledge o don’ use any compu a ional echnology.
Howe e , 22% o he ques ioned g oup e e ed o Compu e -aided de ec ion (CAD), and only
1% e e ed o Con olu ional Neu al Ne wo k (CNN).
Cha 14 - Compu a ional echnologies weekly used.
2%
8%
10%
19%
61%
0% 10% 20% 30% 40% 50% 60% 70%
Adminis a ion knowledge
Lack o us
Complex decision-making
Resis ance o change
Focus on sa ings
% pe o al answe s
Men ioned ba ie s
1%
22%
29%
47% CNN
CAD
Non
Wi hou knowledge
31
5.2. Resul s discussion
The collec ion o answe s we e ob ained mainly in he mos popula ed ci ies o Po ugal, wi h
55% o he esponses om Lisbon and Po o. This esul is due o he numbe o heal hca e
o ganiza ions in hese loca ions. As INE published in he la es Po uguese Na ional S a is ic
s udy in 2013, he numbe o hospi als in Lisbon was 59 and 45 in Po o o 238 hospi als in he
coun y, ep esen ing 44% o he o al hospi als in Po ugal (INE, 2013).
The inqui ed g oup is mainly o med by senio diagnos ic and he apeu ic echnicians,
ep esen ing 75% o he answe s. F om he supply chain and se ice p o ide s i was possible o
collec 23% o he answe s and only 2% o he inqui ed g oup ep esen s he physicians.
Nowadays, compu a ional echnologies in medical imaging se ices a e mainly applied o CT,
adiog aphy, mammog aphy, MRI, and angiog aphy, co esponding o he medical imaging
echniques ha he inqui ed g oup p ima ily uses. This in o ma ion is alida ed by he
sys ema ic li e a u e e iew o being he mos s udied medical imaging echniques o
inco po a e inno a i e compu a ional echnologies, wi h AI inco po a ed in hei se ices.
As men ioned by he p esiden o he Po uguese Associa ion o Hospi al Adminis a o ’s, in
Po ugal he e has been a lack o echnology in es men since 2009. This in o ma ion was
co obo a ed by he inqui ed g oup, wi h mo e han 90% o a i ma i e answe s o he need o
imp o emen on hei medical imaging se ices. The emaining almos 10% a e ep esen ed only
by he p i a e sec o .
To answe he ques ion “Wha a e he main p oblems o imp o emen s de ec ed by he
in e enien a medical imaging se ices?” i is possible o obse e ha he so wa e and
ha dwa e (cha 7) was he mos men ioned p oblem. The ou da ed echnology and equipmen
(cha 8), ollowed by ou da ed so wa e and incompa ibili y and he lack o aining (cha 9) a
he MI se ices, a e he main need o imp o emen men ioned by he ques ioned g oup. This
in o ma ion is con i med in he li e a u e e iew by he Po uguese jou nal (Nunes, 2019).
The mo i a ions o he lack o in es men in new echnology a MI se ices in Po ugal a e
mos ly due o economic ba ie s, esis ance o he change, wi h complex decision-making (cha
10, cha 12). As men ioned by he enqui ed g oup, he adminis a ion o heal hca e
o ganiza ions is mainly ocused on inancial sa ings due o he high in es men and p o ound
changes he inno a i e echnologies a ailable in he ma ke ep esen o MI se ices. These
mo i a ions eply o he ques ion, “Wha a e he implemen a ion ba ie s ecognized o
inco po a e compu a ional echnologies a medical imaging se ices?”. Ano he eason o he
lack o in es men mo i a ion is he need o sa e due o he unsus ainabili y o he heal hca e
sys em (Deloi e, 2011). The inc ease o p i a e and public gene al expendi u e on heal hca e is
inc easing he need o inancial sa ing, p omo ing he lack o in es men in inno a i e
echnology, new equipmen , and so wa e, p omo ing he execu ion o he MI se ice wi h he
ou da ed equipmen and so wa e o e he yea s.
The e o e, he con inuing usage o ou da ed equipmen and so wa e leads o an inc easing
sea ch o p i a e esou ces o upda ed equipmen o collec MI wi h he goal o b idge he
p oblems on he public equipmen , aising he public expenses and he delay in diagnosis.
32
Almos hal o he inqui ed g oup belie e he e may be a possibili y o implemen inno a i e
echnologies in he MI se ices in he nea u u e (app oxima ely 5 yea s). When i was
eques ed no o conside he cos s, 62% o he inqui ed g oup belie ed he heal hca e
ins i u ions had he in en ion o inco po a e hese inno a i e echnologies in hei MI se ice.
This con i ms he economic ba ie as he main mo i a ion o he non-implemen a ion o
inno a i e echnologies in MI se ices.
The main ques ion o his s udy is “Wha is he gap be ween inno a i e heal hca e echnologies
a ailable in he ma ke and he implemen ed echnologies in Medical Imaging a eas in
Po ugal?”. The su ey esul s demons a ed ha MI se ices in Po ugal ha e a huge need o
upg aded MI equipmen , o inco po a e inno a i e echnologies, inc ease he in es men in
human esou ces aining, and acqui e mo e human esou ces o ul ill he need a he MI
se ices.
The majo i y o he inqui ed g oup does no use compu a ional echnology a hei MI se ices
o a e no awa e o hei applica ion (76% o he g oup), as demons a ed in cha 14. Howe e ,
his ac does no allow me o a i m ha no compu a ional echnology is used in hei MI
se ices bu demons a es he de iciency in human esou ces aining.
The ML algo i hm – Compu e -aided de ec ion (CADe) – is he compu a ional echnology
men ioned by he inqui ed g oup. This algo i hm had he i s implemen a ion in mammog aphy
in 1998 and nowadays suppo s X- ay and CT image eade s p ima ily o cance de ec ion. Since
he 1980s i has been esea ched CAD x (compu ed-aided diagnosis) o de elop he capabili y o
dis inguish he malignan and benign b eas lesions. This algo i hm is a second eade mode
applied o he images a e he human eade pe o ms hei in e p e a ion, suppo ing he
human eade o ind lesions in he images.
Con olu ional neu al ne wo k (CNN) algo i hm is no ed by one ques ioned pe son. This
algo i hm is a class o DL algo i hm applied o sol e compu ed ision asks, like image
classi ica ion and he con inuous quan i a i e analysis h ough ou ime o di e en images,
inc eases he da abase esou ces analysis, deli e ing a mo e accu a e diagnosis.
The inqui ed g oup doesn’ know he EDPSO algo i hm and PSO algo i hm o image
segmen a ion, and GAN algo i hm o MRI image imp o emen . These algo i hms a e
unde s udied o suppo and imp o e medical image analysis, decision-making, de ec ion, and
classi ica ion o lesions.
The Po uguese p i a e and public heal hca e expendi u e a e exceeding he GDP g ow h a e
inc easing he mo i a ion o inancial sa ings by Po uguese heal hca e ins i u ions. Adding he
esis ance o change, upg ading he MI echnology and equipmen co ela es wi h he missing
in o ma ion abou he inno a i e compu a ional echnologies applied o MI a ailable a he
ma ke . These ac s inc ease he popula ion dissa is ac ion wi h MI heal hca e se ices and he
MI echnicians and physicians o pe o m hei job.
33
5.3. An in e p e a ion Model
A e collec ing he key aspec s o he non-in es men on inno a i e echnologies by he
heal hca e o ganiza ions (causes) and he e ec s o he pa ien s, MI se ices human esou ces
and o he MI in as uc u e i was buil an in e p e a ion model in Figu e 2.
Figu e 2 - In e p e a ion model o he e ec s caused by he non-in es men in inno a i e echnologies in MI.
The cons uc ion o he in e p e a ion model in igu e 3 was based on he da a collec ed a he
quan i a i e su ey, in he li e a u e e iew pe o med du ing his s udy, and in e p e a ion o
he esul s.
A he p esen s udy i was possible o obse e he ollowing causes o he lack o in es men
in inno a i e compu a ional echnologies:
• High inno a i e echnologies cos s and he ocus on inancial sa ings we e he mos
men ioned causes o he lack o in es men by he inqui ed g oup, as shown in cha s
13 and 14.
o 41% o he inqui ed g oup men ioned he economic ba ie as he main ba ie ,
and 61% o he g oup desc ibed he ocus on sa ings by he heal hca e
ins i u ions;
• The esis ance o change by he physicians and lack o us in inno a i e echnologies
by he heal hca e o ganiza ion adminis a ion a e o he causes o he lack o
in es men . The lack o knowledge on how he inno a i e echnology pe o ms and how
i suppo s he adiologis s in hei asks, imp o ing he accu acy o he diagnosis esul s
and as p esen a ion o he esul s.
o 27% o he g oup men ioned he o ganiza ional ba ie , and 14% men ioned he
human ba ie has he ba ie o inno a i e echnologies implemen a ion in MI
se ices in Po ugal, as shown in cha 12;
40
APPENDIX
Tecnologia p esen e nos se iços de Imagiologia Médica em Po ugal
O p esen e ques ioná io em como obje i o a ecolha de dados pa a o es udo "A aliação da
aplicação de ecnologia ino ado a em imagiologia médica em Po ugal" - Disse ação de
Mes ado em Tecnologias e Sis emas de In o mação da Uni e sidade No a de Lisboa –
In o ma ion Managemen School pela aluna Joana Sil a.
Es e es udo em como p incipal obje i o a alia o gap exis en e en e as ecnologias ino ado as
disponí eis no me cado e as ecnologias a ualmen e p esen es nos se iços de Imagiologia
Médica (IM) em Po ugal.
A a és do p esen e ques ioná io p e ende-se ecolhe dados de odas as egiões de Po ugal,
do se o público e p i ado e de di e sas á eas de o mação que in e enham e in e ajam com
se iços de Imagiologia Médica em Po ugal.
O ques ioná io es á di idido em ês secções. A p imei a secção p e ende localiza e ca ac e iza
o sujei o que esponde ao ques ioná io, as écnicas e ecnologias com que abalha e se de e a
melho ias no se iço de Imagiologia Médica. Caso a espos a seja a i ma i a, na segunda secção
é pedido que iden i ique as melho ias necessá ias no se iço de imagiologia médica onde
abalha. Po im, na e cei a secção são ap esen ados dados de ecnologias ino ado as
aplicadas em Imagiologia Médica, em es udo ou disponí eis no me cado, ob idos a a és da
ealização da e isão sis emá ica de li e a u a ealizada no inal do ano 2020. P e ende-se com
a e cei a secção comp eende o conhecimen o das ecnologias e analisa mo i ações de
implemen ação de ecnologias ino ado as nos se iços de Imagiologia Médica em Po ugal.
O ques ioná io em um empo de espos a de ce ca de 10 minu os.
Ag adeço desde já a sua disponibilidade e apoio na ealização desde es udo.
A bibliog a ia u ilizada é disponibilizada no inal do ques ioná io.
Ca ac e ização do sujei o em es udo
Ques ões
A que sec o pe ence a(s) clínica(s) ou hospi al(ais) com que colabo a?
• Público
• P i ado
• Público e P i ado
• Pa ce ia público-p i ado
A que á ea geog á ica pe ence o se iço de saúde com que colabo a?
• A ei o
• Ang a do He oísmo
• Beja
• B aga

41
• B agança
• Cas elo B anco
• Coimb a
• É o a
• Fa o
• Funchal
• Gua da
• Ho a
• Lamego
• Lei ia
• Lisboa
• Pon a Delgada
• Po aleg e
• Po o
• San a ém
• Se úbal
• Viana do Cas elo
• Viseu
• Região No e
• Região Cen o
• Região Sul
• A quipélago dos Aço es
• A quipélago da Madei a
• Todo o país
Qual a sua unção jun o das en idades de Imagiologia Médica?
• Técnico Supe io de Diagnós ico e Te apêu ica
• Médico
• Fo necedo
• Auxilia de se iço
• P es ado de se iços
Quais as écnicas de Imagiologia Médica com que abalha?
• Radiog a ia
• Tomog a ia compu o izada (CT)
• Angiog a ia
• Tomog a ia de emissão de posi ões (PET)
• Tomog a ia compu o izada de emissão de o ão (SPECT)
• Ecog a ia
• Ressonância Magné ica
• Ressonância Magné ica uncional
• Mamog a ia
42
Qual(ais) a(s) écnica(s) de Imagiologia Médica que em conhecimen o da aplicação de
ecnologia(s) compu acional(ais)?
• Radiog a ia
• Tomog a ia compu o izada (CT)
• Angiog a ia
• Tomog a ia de emissão de posi ões (PET)
• Tomog a ia compu o izada de emissão de o ão (SPECT)
• Ecog a ia
• Ressonância Magné ica
• Ressonância Magné ica uncional
• Mamog a ia
• Sem conhecimen o
No(s) se iço(s) de Imagiologia Médica com que colabo a sen e a necessidade de melho ias?
• Sim
• Não
Melho ias necessá ias nos se iços de Imagiologia Médica em Po ugal
DEFINIÇÕES:
Tecnologias compu acionais: equipamen o e sis emas ou aplicações in o má icas com
capacidade de p ocessa ele ada quan idade de in o mação. No caso de imagiologia médica, os
sis emas u ilizados êm no malmen e como obje i o melho a a qualidade das imagens
ecolhidas, diminui a dimensão dos ichei os da imagem e acili a a pa ilha das imagens en e
as pa es in e essadas ( écnicos, u en es e médicos).
Ha dwa e: componen e ísica do equipamen o.
So wa e: conjun o de p og amas ou aplicações, ins uções ou eg as que pe mi em o
equipamen o unciona .
Ques ões
Nos equipamen os de Imagiologia Medica u ilizados no se iço hospi ala ou clínica com que
colabo a, que ipo de necessidades de melho ia ou de ei os encon a?
• So wa e
• Ha dwa e
• So wa e e Ha dwa e
Quais os p oblemas ou necessidades de melho ia que de e a nos equipamen os de Imagiologia
Médica com que abalha?
• Equipamen os en elhecidos
• Incompa ibilidade en e so wa e
• In o mação cen alizada
• Tecnologia desa ualizada
• Adiciona ou a opção
43
No(s) se iço(s) de Imagiologia Médica com que colabo a, que equisi os mencionados abaixo
conside a em al a?
• Fo mação
• Recu sos Humanos
• So wa e ecen e
• Equipamen o ecen e de Imagiologia Médica
• Adiciona ou a opção
Tecnologias compu acionais em es udo e disponí eis no me cado
Após ealiza uma e isão sis emá ica da li e a u a das ecnologias compu acionais aplicadas ou
em es udo em Imagiologia Médica nos úl imos cinco anos oi possí el ecolhe os dados
ap esen ados abaixo.
As á eas de Imagiologia Médica mais es udadas e com maio aplicação de ecnologias
compu acionais são a Tomog a ia Compu o izada, Ul assonog a ia e Ressonância Magné ica.
Po ou o lado, as ecnologias compu acionais mais es udadas pa a aplicação em Imagiologia
Médica são In eligência A i icial, Machine Lea ning e Deep Lea ning.
In eligência A i icial consis e numa á ea da ciência da compu ação ocada na esolução de
a e as que o se humano execu a. Po ou as pala as, em como obje i o eplica o
compo amen o do cé eb o humano na ealização de a e as.
Uma subca ego ia da In eligência A i icial é Machine Lea ning desc i a como um elemen o da
ciência da compu ação que em a capacidade de ap endizagem con ínua sem in e enção
humana, ala ancando algo i mos exis en es, desen ol endo modelos analí icos com a
capacidade de econhecimen o de pad ões e de ealiza p e isões baseado em dados.
Po sua ez, Deep Lea ning é uma subca ego ia de Machine Lea ning que u iliza uma classe
especí ica de algo i mos - edes neu onais - com o obje i o de eplica o uncionamen o do
cé eb o humano e analisa dados con inuamen e, com uma es u u a lógica semelhan e a um
se humano, podendo ap ende , apoia ou oma decisões.
Abaixo ap esen am-se duas abelas. Na Tabela 1 é ap esen ado o ní el de aplicação e
disponibilidade das ecnologias compu acionais no me cado. Na Tabela 2 são desc i as as á ias
aplicações das ês ecnologias compu acionais desc i as acima.
Peço que leia com a enção, analise os dados e in o mações disponibilizadas e esponda às
qua o ques ões abaixo de aco do com a sua ealidade, sendo uma opcional.
Tabela 1 – Disponibilidade de ecnologias compu acionais em Imagiologia Médica.
Tecnologias
Compu acionais
Técnicas de Imagiologia Médica
Radiog a ia
Tomog a ia
Compu acional
Ul assonog a ia
Ressonância
Magné ica
In eligência A i icial
Exis en e21,32
Exis en e8
Em in es igação7
Exis en e5
Machine Lea ning
Em in es igação30
Em in es igação5
Em in es igação7
Em
in es igação1,5,30
Deep Lea ning
Em in es igação24
Em in es igação8
Em in es igação7
Em in es igação20
44
Tabela 2 - Aplicação de ecnologias compu acionais em Imagiologia Médica.
Tecnologias Compu acionais
Aplicações
In eligência A i icial
Sis emas de supo e ao abalho diá io dos p o issionais de saúde, melho ando a apidez
de en ega e melho ando a análise das imagens5
Machine Lea ning
Dis inção de lesões mamá ias malignas de lesões benignas5
Classi icação ou diagnós ico compu acional em ul assonog a ia7
Segmen ação de ecidos em essonância magné ica5
Regis o de imagens e econhecimen o de pad ões2
Deep Lea ning
De e a umo mamá io8
De e a anomalias e doenças ce eb ais12,18,28,33
De e a canc o da p ós a a5
En e ou as aplicações de de eção de anomalias7
Ques ões
Den o do hospi al ou clínica com que colabo a, desconside ando os cus os associados,
conside a que é possí el inco po a algumas das ecnologias desc i as acima?
• Sim
• Não
• Tal ez
Den o do hospi al ou clínica com que colabo a, conside a que é possí el inco po a algumas
das ecnologias desc i as acima num cu o p azo (ap oximadamen e 5 anos)?
• Não
• Sim
• Tal ez
Quais são as p incipais ba ei as pa a a inco po ação de ecnologias compu acionais no hospi al
ou clínica com que colabo a?
• Ba ei as de p oje o
• Ba ei a económica
• Ba ei a écnica
• Ba ei a o ganizacional
• Ba ei a humana
• Ou a opção
De aco do com as ba ei as pa a a inco po ação mencionadas na ques ão acima, pode
especi ica /cla i ica de aco do com a sua ealidade?
45
QUESTÃO OPCIONAL: Qual o ipo de ecnologia compu acional com que abalha numa base
semanal?
• Con olu ional Neu al Ne wo k (CNN)
• Pa icle Swa m Op imiza ion (PSO)
• Gene a i e Ad e sa ial Model (GAN)
• Compu e -aided de ec ion (CAD)
• Sem conhecimen o
• Nenhuma
• Ou a opção