elec onics
A icle
Building S anda dized and Secu e Mobile Heal h
Se ices Based on Social Media
Jesús D. T igo 1,* , Ósca J. Rubio 2, Miguel Ma ínez-Esp onceda 1,Ál a o Alesanco 3,
JoséGa cía3and Luis Se ano-A iezu 1
1Depa men o Elec ical, Elec onic and Communica ions Enginee ing, Public Uni e si y o Na a a,
Ins i u e o Sma Ci ies (ISC), Na a a Ins i u e o Heal h Resea ch (IdiSNA), 31006 Pamplona, Spain;
[email p o ec ed] (M.M.-E.); [email p o ec ed] (L.S.-A.)
2E e is Asse s, 50009 Za agoza, Spain; [email p o ec ed]
3Depa men o Elec onics Enginee ing and Communica ions, Uni e si y o Za agoza, A agón Ins i u e o
Enginee ing Resea ch (I3A), 50018 Za agoza, Spain; alesanco@uniza .es (Á.A.); joga mo@uniza .es (J.G.)
*Co espondence: [email p o ec ed]
Recei ed: 20 Oc obe 2020; Accep ed: 18 Decembe 2020; Published: 21 Decembe 2020


Abs ac :
Mobile de ices and social media ha e been used o c ea e empowe ing heal hca e se ices.
Howe e , p i acy and secu i y conce ns emain. Fu he mo e, he in eg a ion o in e ope abili y
biomedical s anda ds is a s a egic ea u e. Thus, he objec i e o his pape is o build enhanced
heal hca e se ices by me ging all hese componen s. Me hodologically, he cu en mobile heal h
elemoni o ing a chi ec u es and hei limi a ions a e desc ibed, leading o he iden i ica ion o new
po en iali ies o a no el a chi ec u e. As a esul , a s anda dized, secu e/p i a e, social-media-based
mobile heal h a chi ec u e has been p oposed and discussed. Addi ionally, a echnical p oo -o -concep
( wo And oid applica ions) has been de eloped by selec ing a social media (Twi e ), a secu i y
en elope (open P e y Good P i acy (openPGP)), a s anda d (Heal h Le el 7 (HL7)) and an
in o ma ion-embedding algo i hm (modi ying he anspa ency channel, wi h wo e sions). The
es s pe o med included a small-scale and a bounda y scena io. Fo he o me , wo sizes o images
we e es ed; o he la e , he wo e sions o he embedding algo i hm we e es ed. The esul s
show ha he sys em is as enough (less han 1 s) o mos mHeal h elemoni o ing se ices. The
a chi ec u e p o ides use s wi h iendly (images sha ed ia social media), s aigh o wa d ( as and
inexpensi e), secu e/p i a e and in e ope able mHeal h se ices.
Keywo ds: mHeal h; p i acy; secu i y; social media; s anda diza ion
1. In oduc ion
In he las decade, he usage o mobile and sma phones has g own exponen ially. Acco ding o a
mid-2019 epo [
1
], he as majo i y o Ame icans (96%) own a cellphone o some kind, while he sha e
o Ame icans ha own sma phones is 81%. This ac has spu ed he scena io o mobile heal h (also
e e ed o as mHeal h), which, as s a ed by he Wo ld Heal h O ganiza ion, is he “medical and public
heal h p ac ice suppo ed by mobile de ices” [
2
]. The co e age o mHeal h includes se e al aspec s,
such as he acquisi ion, manipula ion, classi ica ion and ansmission o heal h- ela ed in o ma ion [
3
].
Thus, he use s become he cen e o he ac ion, being p oduce s and owne s o hei own biomedical
da a and signals, which can be ubiqui ously ga he ed wi h hei own pe sonal heal h de ices and
ansmi ed by means o hei pe sonal mobile de ices. Fu he mo e, he use s can analyze hei
biomedical da a locally and sha e hem wi h bo h o mal and in o mal ca egi e s. Ul ima ely, such
da a can be ei he s o ed in pe sonal da a aul s o u u e consul a ions o sen o o he sys ems o
se ices, depending on he mHeal h applica ion.
Elec onics 2020,9, 2208; doi:10.3390/elec onics9122208 www.mdpi.com/jou nal/elec onics
Elec onics 2020,9, 2208 2 o 26
All said, al hough he easibili y and a ailabili y o adi ional elemoni o ing mHeal h se ices
ha e been ho oughly desc ibed in he li e a u e [
4
,
5
], he e a e s ill challenges o be sol ed, mainly
ela ed o secu i y and p i acy conce ns. Mo eo e , he ixed s uc u es o such cen alized a chi ec u es
adi ionally p omo ed by manu ac u e s could lead o lack o engagemen , mo i a ion o connec ions
among hei use s, who may decide o c ea e hei own solu ions [
6
]. Indeed, many pa ien s and
communi ies oday a e demanding o become co-p oduce s and, in he end, holde s o hei own
da a. I hey do no ge o ha e con ol, hey may ake o e and c ea e he necessa y ools o gain
sel -empowe men hemsel es. An epi ome he eo is he Nigh scou mo emen [
7
], an open-sou ce,
Do-I -You sel p ojec allowing eal- ime access o a glucose moni o , as well as da a pe sis ence in he
cloud. This p ojec can be seen as an illus a ion o he di e ence be ween “pa ien -cen e ed ca e” and
“ eal pa ien con olled ca e”, whe e i is he pa ien s, he ela i es and he communi ies hey li e in
who inally “make sense” o hings, de ine he goals and s ess cohe ence [8].
Al hough he Nigh scou p ojec has p o ided aluable lessons o he new pa adigm o
empowe ed ci izens, one logical s ep o wa d is he use o social-media-based heal h sys ems, since
hey could os e engagemen , empowe men and communi y building [
9
]. Besides he p edominan ly
ludic cha ac e o social media, new uses in di e en domains a e being in es iga ed and de eloped
nowadays. They a e d i en by he a ac ing ea u es o social media as well as hei ema kable
mass o use s (e.g., as o Sep embe , 2020, Facebook claimed o ha e 3 billion use s [
10
], which
ep esen nea ly one- hi d o he wo ld’s popula ion). Recip ocally, a ac ing new use s o social
media, e.g., hose coming om mHeal h scena ios, would help social media o ea i m hei leading
posi ion in oday’s in e ne pano ama. Indeed, social media p o ide a wide a ie y o ools ha enable
use s o build communi ies a ound hem whe e hey can c ea e, sha e and exchange in o ma ion
in di e en o ma s [
11
]. E e since he social media appea ed, he idea o combining social media
and heal hca e has gained momen um. S udies ha e ound ha heal hca e o ganiza ions, clinicians,
pa ien s and egula o y bodies could bene i om he use o social media [
12
,
13
]. Howe e , in spi e o
he p omising bene i s o social-media-based heal hca e, he e a e some challenges s ill o be sol ed.
Mos o hose issues a e linked o p i acy and secu i y conce ns, bu he e a e also open ques ions
abou usabili y, manipula ion o iden i y, go e nance o con iden iali y, along wi h he a o emen ioned
demo i a ion [9,14,15].
As i has been ma ked in he li e a u e, he e is limi ed e idence ela ed o he e icacy and
e ec i eness o social media in heal hca e [
9
]. Mos p ojec s combining heal hca e and social media so
a use da a mining o analyze sha ed heal h- ela ed da a and ex ac aluable in o ma ion [
16
–
18
]. To
da e, howe e , he e is li le e o in he li e a u e owa ds social-media-based mHeal h sys ems whe e
biomedical da a a e sen h ough social media while aking in o accoun secu i y and s anda diza ion.
An example ha pa ially accomplish such a goal was p esen ed in [
19
]. This sys em le e ages
Twi e o send he main da a o a back-end eposi o y using a web based app oach, while o e ing he
possibili y o sha ing heal h- ela ed messages cons uc ed acco ding o pa icula s a us desc ip o s
using a medical nomencla u e. None heless, as ega ds o p i acy and secu i y, hey only make use o
he buil -in secu i y policies implemen ed by Twi e (open au hen ica ion and p i a e lis s). Rela ed
o his, he au ho s p esen ed a echnical p oo -o -concep sys em o ollowing up ca dio ascula
pa ien s using Twi e and Heal h Le el 7 (HL7) [
20
], which can be conside ed an imp o emen he eo .
Ne e heless, no u he secu i y and p i acy measu es we e implemen ed.
Thus, adequa e p o ec ion policies shall be implemen ed in social-media-based, mHeal h
applica ions o achie e secu i y and p i acy le els in line wi h he demands o use s and he egula ions
applicable. Examples o such egula ions a e he Heal h Insu ance Po abili y and Accoun abili y Ac
(HIPAA) in he Uni ed S a es o he Gene al Da a P o ec ion Regula ion (GDPR) in Eu ope, e ec i e since
2018. The use o HIPAA as a means o achie ing s anda dized da a secu i y and p i acy in mHeal h
scena ios has al eady been p oposed in he li e a u e [
21
]. A common objec i e o hese egula ions is
o gua an ee In o ma ion Assu ance and Secu i y (IAS). T adi ionally, con iden iali y, in eg i y and
a ailabili y—also e e ed o as he CIA- iad—we e he elemen s modelling IAS. Nowadays, he
Elec onics 2020,9, 2208 3 o 26
CIA- iad has e ol ed o a mo e comp ehensi e IAS-oc a e (con iden iali y, in eg i y, a ailabili y,
accoun abili y, audi abili y, au hen ici y, non- epudia ion and p i acy) [
22
]. Such egula ions also
en o ce p e en ion and eac ion o da a b eaches as well as esponsibili y and sanc ions o hose ha
do no ho oughly add ess he a o emen ioned measu es. Social media se ices mus comply wi h
egula o y equi emen s o he coun ies hey a e wo king in. Howe e , uploading unp o ec ed da a o
social media could sow suspicion o mis us among use s. The e o e, he design and implemen a ion
o an addi ional obus secu i y laye , being complemen a y o he measu es al eady implemen ed by
social media, would help o gua an ee independence om hei p i acy policies and aise he us o
he po en ial use s.
Addi ionally, i is ue ha any ad-hoc secu e, p i a e, social-media-based, mHeal h solu ion is
able o exchange biomedical in o ma ion wi hou he need o a common in o ma ion model. Howe e ,
his may no su ice i pe asi e, dis ibu ed, in eg a ed biomedical ecosys ems a e o be achie ed [
23
].
In o de o accomplish a leas seman ic in e ope abili y—acco ding o he model p oposed by
Tu ni sa e al. [
24
]— he con en o he in o ma ion exchange eques s mus be unambiguously
de ined. Hence, biomedical in e ope abili y s anda ds a e highly ecommended. Wi hin he heal hca e
domain, ini ia i es like HL7 o Digi al Imaging and Communica ion in Medicine (DICOM) a e obus ,
widesp ead examples o his s anda diza ion e o .
The elemen s p esen ed in he pa ag aphs abo e— o wi , mHeal h, social media, secu i y/p i acy
and s anda diza ion—a e usually ea ed as sepa a e, o , a leas , loosely in eg a ed ields. The exis ing
li e a u e o e s p elimina y examples desc ibing sys ems ha pa ially co e some o hese ields
combined, being he mos comple e so a he wo k conduc ed by T ian a yllidis e al. [
19
] and a
p e ious wo k by he au ho s [
20
]. Thus, he unde lying hypo hesis o his pape is ha all hese
componen s could be seamlessly me ged o build enhanced heal hca e se ices ( his is co e ed in
Sec ion 2). As a esul , he main objec i e o his wo k is o p opose a gene ic a chi ec u e o building
s anda dized, secu e, p i a e, social-media-based mHeal h se ices (Sec ion 3.1). Fo he sake o
simplici y and con enience, we in oduce he e he ac onym mH3S (a e mHeal h, s anda dized, secu e
and social). Secondly, a p oo o concep o he p oposed mH3S a chi ec u e will be also p esen ed
(Sec ion 3.2). I is composed o (a) Twi e as he social media, (b) e sion 2 o HL7 as a means
o in e ope abili y, (c) openPGP as secu i y en elope and (d) a pa icula embedding algo i hm.
Thi d, bo h he gene ic a chi ec u e p oposed and he echnical p oo -o -concep implemen a ion a e
discussed in Sec ion 4. Conclusions a e d awn in Sec ion 5. Table 1p o ides a lis o he ac onyms
used h oughou he pape .
Table 1. Ac onyms and hei meaning.
Ac onym Meaning
AES Ad anced Enc yp ion S anda d
API Applica ion P og amming In e ace
CA Ce i ica ion Au ho i y
CIA Con iden iali y, In eg i y and A ailabili y
CDSS Clinical Decision Suppo Sys ems
CMS C yp og aphic Message Syn ax
DICOM Digi al Imaging and Communica ion in Medicine
ECG Elec oCa dioG am
EHR Elec onic Heal h Reco d
GDPR Gene al Da a P o ec ion Regula ion
GZIP GNU ZIP
HIPAA Heal h Insu ance Po abili y and Accoun abili y Ac
HIS Heal h In o ma ion Sys em
HL7 Heal h Le el 7
HS Hos Sys em
HTTPS Hype ex T ans e P o ocol Secu e
IAS In o ma ion Assu ance and Secu i y
IEEE Ins i u e o Elec ical and Elec onics Enginee s
Elec onics 2020,9, 2208 4 o 26
Table 1. Con .
Ac onym Meaning
IETF In e ne Enginee ing Task Fo ce
IHE In eg a ing he Heal hca e En e p ise
ISO In e na ional O ganiza ion o S anda diza ion
JPEG Join Pho og aphic Expe s G oup
LOINC Logical Obse a ion Iden i ie s Names and Codes
mH3S S anda dized, secu e/p i a e, social-media-based mobile heal h a chi ec u e
OAu h Open Au ho iza ion
openPGP open P e y Good P i acy
ORU Obse a ion Resul Unsolici ed
PHD Pe sonal Heal h De ice
PHR Pe sonal Heal h Reco d
PKI Public Key In as uc u e
PNG Po able Ne wo k G aphics
QR Quick Response
RGBA/ARGB Red G een Blue +Alpha
RSA Ri es , Shami and Adleman
S/MIME Secu e/Mul ipu pose In e ne Mail Ex ensions
SCP-ECG S anda d Communica ions P o ocol o compu e assis ed Elec oCa dioG aphy
SNOMED-CT Sys ema ized Nomencla u e o Medicine—Clinical Te ms
TLS T anspo Laye Secu i y
TPM T us ed Pla o m Modules
UCUM Uni ied Code o Uni s o Measu e
UMLS Uni ied Medical Language Sys em
2. Ma e ials and Me hods
The me hodology p oposed in his pape consis s o wo s eps. Fi s , he cu en mHeal h
a chi ec u es— ocused on elemoni o ing a chi ec u es—a e illus a ed, co e ing he adi ional usage
and hei e olu ion. Second, he limi a ions o he a chi ec u es composing such e olu ion a e desc ibed
h ough a comp ehensi e e iew o he exis ing li e a u e, which ul ima ely will lead o he iden i ica ion
o new po en iali ies o a no el app oach.
2.1. Analysis o T adi ional Telemoni o ing mHeal h A chi ec u es
A gene ic mHeal h a chi ec u e acili a es he implemen a ion o mHeal h applica ions, usually
g ouped in h ee majo —and in e ela ed— ields: heal h and i ness, independen li ing and disease
managemen . Al hough o he mHeal h applica ions a e possible—e.g., medical e e ence, nu i ion o
wellness applica ions–, he scope o his pape is ocused in applica ions ha a e able o epo medical
s a us upda es o o mal o in o mal ca egi e s. Such applica ions demand a eliable and e icien
acquisi ion o pe sonal biomedical in o ma ion, i s adequa e s o age and a pe asi e, ubiqui ous and
con olled access o he use s ha need o consul his in o ma ion.
To cope wi h he equi emen s o he heal h scena ios desc ibed abo e, di e en mHeal h
a chi ec u es ha e been p oposed and de eloped in he li e a u e [
4
,
5
]. Re e ence [
4
], published in
2015, e iews mHeal h se ices and shows a ypical a chi ec u e he eo . Re e ence [
5
] pe o ms a
su ey on he a chi ec u es o elemoni o ing esea ch p ojec s in 2014 and, om ha knowledge,
de i es a common a chi ec u e o elemoni o ing sys ems, which can be summa ized in h ee dis inc
ie s: senso s, ga eway and emo e se e .
Based on such e iews, a gene ic mHeal h a chi ec u e is illus a ed in Figu e 1, which is u he
de ailed as ollows. The mos basic end- o-end mHeal h a chi ec u e is comp ised by wo elemen s.
Fi s , a Pe sonal Heal h De ice (PHD) o , mo e gene ically, a sensing uni , which collec s and sends he
use ’s biomedical in o ma ion. Second, a Hos Sys em (HS), which s o es he collec ed in o ma ion, o
example a Heal h In o ma ion Sys em (HIS) wi h an Elec onic Heal h Reco d (EHR) o a Pe sonal
Heal h Reco d (PHR), he la e also e e ed o as pe sonal da a aul s. In addi ion, he e a e usually
Elec onics 2020,9, 2208 5 o 26
se e al PHDs a ound he pa ien /use , and hey seldom ha e he connec i i y o each he HS— o da e,
ew PHDs a e In e ne - eady, al hough he pa adigm may be shi ing due o he In e ne o Medical
Things [
25
]. Thus, mos mHeal h a chi ec u es oday include a hi d elemen , namely he concen a o
de ice, a mobile de ice, e.g., cell phone o able , which ga he s he biomedical da a om he di e en
PHDs and o wa ds hem o he HS. Fu he mo e, depending on he in ended mHeal h applica ion,
a ious o he elemen s can be inco po a ed in o he end- o-end a chi ec u e. Fo example, se ice
p o ide s and medical sys ems would be placed be o e and a e he da a a i e a he HS. Examples o
medical sys ems a e ala m sys ems o Clinical Decision Suppo Sys ems (CDSS). They would pe o m
di e en ope a ions included in he scope o he HS, such as he managemen , moni o ing, p ocessing
o ollow-up he use ’s biomedical in o ma ion. Mo eo e , o he elemen s can connec wi h he HS o
ei he sha e medical in o ma ion, such as hi d-pa y hos sys ems, o access ha in o ma ion, such as
a consul a ion sys ems, he eby in e acing he ca egi e s and he use s wi h he HS (see Figu e 1).
As ega ds o he pe sons in ol ed, gene ally, up o ou ypes can be dis inguished in a adi ional
mHeal h a chi ec u e:
•The pa ien s o use s: They eco d he biomedical measu emen s emo ely.
•
The o mal ca egi e s: Nu ses and physicians who e iew he in o ma ion and ollow up
pa ien s/use s.
•
Resea che s: E en ually, a esea che would analyze he da a ga he ed om a pa ien /use o a
g oup o hem o in es iga e in o a speci ic scena io o pa hology.
•
The echnicians: They a e in cha ge o ensu ing ha he hospi al de ices and he back-end se ices
wo k p ope ly.
In he mos basic a chi ec u e, he pa ien /use is he only ac o in ol ed, who moni o s hemsel es
using a mobile applica ion (synch onized o no wi h a se e ac ing as PHR), becoming an in o mal
ca egi e o hemsel es.
Elec onics 2020, 9, x FOR PEER REVIEW 5 o 26
biomedical da a om he di e en PHDs and o wa ds hem o he HS. Fu he mo e, depending on
he in ended mHeal h applica ion, a ious o he elemen s can be inco po a ed in o he end- o-end
a chi ec u e. Fo example, se ice p o ide s and medical sys ems would be placed be o e and a e
he da a a i e a he HS. Examples o medical sys ems a e ala m sys ems o Clinical Decision Suppo
Sys ems (CDSS). They would pe o m di e en ope a ions included in he scope o he HS, such as
he managemen , moni o ing, p ocessing o ollow-up he use ’s biomedical in o ma ion. Mo eo e ,
o he elemen s can connec wi h he HS o ei he sha e medical in o ma ion, such as hi d-pa y hos
sys ems, o access ha in o ma ion, such as a consul a ion sys ems, he eby in e acing he ca egi e s
and he use s wi h he HS (see Figu e 1).
As ega ds o he pe sons in ol ed, gene ally, up o ou ypes can be dis inguished in a
adi ional mHeal h a chi ec u e:
• The pa ien s o use s: They eco d he biomedical measu emen s emo ely.
• The o mal ca egi e s: Nu ses and physicians who e iew he in o ma ion and ollow up
pa ien s/use s.
• Resea che s: E en ually, a esea che would analyze he da a ga he ed om a pa ien /use o a
g oup o hem o in es iga e in o a speci ic scena io o pa hology.
• The echnicians: They a e in cha ge o ensu ing ha he hospi al de ices and he back-end
se ices wo k p ope ly.
In he mos basic a chi ec u e, he pa ien /use is he only ac o in ol ed, who moni o s
hemsel es using a mobile applica ion (synch onized o no wi h a se e ac ing as PHR), becoming
an in o mal ca egi e o hemsel es.
Figu e 1. Ac o s and communica ion lows in a adi ional mobile heal h a chi ec u e.
2.2. Limi a ions o he T adi ional App oach, Cu en E olu ion and Po en ial Fea u es
As echnology e ol ed and mHeal h a chi ec u es we e becoming mo e pe asi e, addi ional
p oblems we e de ec ed and highe a chi ec u al equi emen s we e conside ed necessa y.
One o he i s issues iden i ied was he lack o in e ope abili y [26,27], which has been a
common opic o deba e o da e. This issue has commonly been add essed by c ea ing medical
e minologies and medical s anda ds. Wi hin he o me g oup, one o he mos p ominen is he
Logical Obse a ion Iden i ie s Names and Codes (LOINC). The la e g oup comp ises a wide
a ie y o examples. A p ominen e o would be he In e na ional O ganiza ion o S anda diza ion
(ISO)/Ins i u e o Elec ical and Elec onics Enginee s (IEEE) 11073, in ended o he in e ope abili y
o medical de ices. Ano he example is DICOM, in ended o medical images. Addi ionally, he
S anda d Communica ions P o ocol o compu e assis ed Elec oCa dioG aphy (SCP-ECG) o he
Figu e 1. Ac o s and communica ion lows in a adi ional mobile heal h a chi ec u e.
2.2. Limi a ions o he T adi ional App oach, Cu en E olu ion and Po en ial Fea u es
As echnology e ol ed and mHeal h a chi ec u es we e becoming mo e pe asi e, addi ional
p oblems we e de ec ed and highe a chi ec u al equi emen s we e conside ed necessa y.
One o he i s issues iden i ied was he lack o in e ope abili y [
26
,
27
], which has been a common
opic o deba e o da e. This issue has commonly been add essed by c ea ing medical e minologies
and medical s anda ds. Wi hin he o me g oup, one o he mos p ominen is he Logical Obse a ion

Elec onics 2020,9, 2208 6 o 26
Iden i ie s Names and Codes (LOINC). The la e g oup comp ises a wide a ie y o examples. A
p ominen e o would be he In e na ional O ganiza ion o S anda diza ion (ISO)/Ins i u e o Elec ical
and Elec onics Enginee s (IEEE) 11073, in ended o he in e ope abili y o medical de ices. Ano he
example is DICOM, in ended o medical images. Addi ionally, he S anda d Communica ions P o ocol
o compu e assis ed Elec oCa dioG aphy (SCP-ECG) o he ansmission o ECGs. To conclude
wi h he examples, HL7 is a se o s anda ds o acili a ing he exchange o medical in o ma ion.
A chi ec u ally, some o ganiza ions ha e p oposed a a ie y o echnical amewo ks o p omo e
he use o such s anda ds. The mos p ominen is In eg a ing he Heal hca e En e p ise (IHE). I
p o ides a se o p o iles ha desc ibe clinical in o ma ion needs o wo k low scena ios and ely on
exis ing medical s anda ds o accomplish hem. The e has been e o in he li e a u e p esen ing some
a chi ec u al p oposals o end- o-end s anda d-based mHeal h amewo ks [28,29].
Despi e he undeniable bene i s o s anda d-based mHeal h, he lack o p i acy and secu i y
is s ill one o he majo conce ns [
30
,
31
]. As ega ds o consume s a i ude, ha ing con ol o e
mHeal h p i acy and secu i y ea u es as well as us in p o ide s we e ecen ly iden i ied as
key issues [
32
]. Mo eo e , he e a e na ional and in e na ional egula ions—as he a o emen ioned
HIPAA o GDPR—which compel mHeal h a chi ec u e designe s o ake in o accoun aspec s such as
con iden iali y, in eg i y, a ailabili y, accoun abili y, audi abili y, au hen ici y, non- epudia ion and
p i acy. E o can be ound in he li e a u e aimed a p o iding secu e, s anda d-based mHeal h
a chi ec u es. Fo example, Rubio e al. p oposed a lexible s uc u e ha p o ides ea u es ailo ed o
he needs o di e en mHeal h applica ions, based on a mul i-laye ed, IHE-based ex ension o ISO/IEEE
11,073 [33].
In pa allel, while adi ional mHeal h a chi ec u es we e deployed by heal hca e au ho i ies, hey
we e he eby ine i ably—albei no in en ionally—mainly ocused on medical s a and suppo ing
clinical wo k. In such adi ional s uc u es, he co e o he se ice is loca ed a he HIS, and i is
de eloped by he public o p i a e heal hca e o ganiza ion o e ing he mHeal h se ice. This si ua ion
en ails an e o o main ain he so wa e upda ed and he da a a ailable. Cu en ends in da a
managemen , howe e , indica e a pa adigm shi owa ds cloud compu ing, which enables designe s o
consume di e en esou ces on-demand. Such esou ces include in as uc u e (compu a ion, s o age,
ne wo king), componen s ha acili a e he c ea ion o applica ions and se ices (e.g., middlewa e),
and hi d-pa y so wa e and/o da a. Se ices such as Amazon Web Se ices, Google Cloud Pla o m
o Mic oso Azu e ha e gained no iceable g ound, due o hei wide ange o op ions and lexibili y.
In his con ex , Rahimi e al. e iewed he s a e o he a o cloud compu ing in mobile en i onmen s
and illus a ed hei applica ion o a ious domains, including heal h [
34
]. In he same epo , hey
ale ha secu i y and p i acy a e c i ical aspec s wi h s ill open esea ch issues. None heless, e o
owa ds mobile cloud compu ing in heal h en i onmen s aking in o accoun —albei o di e en
deg ees—s anda diza ion and secu i y/p i acy conce ns can be ound in he li e a u e [
35
–
37
]. In 2012,
Hsieh e al. p oposed cloud and pe asi e compu ing based 12-lead elec oca diog aphy se ice o
ealize ubiqui ous 12-lead ECG ele-diagnosis. In such pape , hey selec ed he Mic oso Azu e cloud
o p ocess and s o e he e ogeneous ECG o ma s (e.g., SCP-ECG o DICOM-ECG). They included some
secu i y and p i acy ea u es. Fo example, au hen ica ion based on oles and in e ne p o ocol add ess
ange, da a enc yp ion ( ia hype ex ans e p o ocol secu e (HTTPS)), sec e key p o ec ed s o age o
ECG ile enc yp ion and e i ica ion while epo s a e e ie ed [
35
]. In 2013,
Ribei o e al.
desc ibed a
solu ion o ou sou cing medical images o Amazon elas ic compu e cloud based on DICOM and a
numbe o IHE p o iles, bu o emos on c oss-en e p ise documen sha ing o images. As ega ds o
secu i y, hey p oposed an enc yp ion me hod which hides access pa e ns o a acke s, ye allows
sea ches h ough he con en [
36
]. In 2016, Hanen e al. published a heal hca e sys em in mobile
cloud compu ing en i onmen s. They used a cloud simula o o con ey DICOM-complian medical
images conside ing some secu i y and p i acy issues, such as au hen ica ion, access con ol o da a
enc yp ion [37]—al hough he eal implemen a ion has no been published so a .
Elec onics 2020,9, 2208 7 o 26
While he cloud is a p omising echnology o mobile heal h ca e en i onmen s, i is mainly
in ended o back-end pu poses—usually including compu ing load. In addi ion, cloud-compu ing
echnologies a e no di ec ly connec ed o he use ’s pe sonal ne wo k. In con as , heal hca e
amewo ks ha a e use - iendly, social, empowe ing, decen alized, echnically easy o deploy and
sel -manageable could lead o a pa adigm shi . This can be achie ed by using social media, e.g.,
Twi e , Facebook, e c., which can be seen as a pa icula iza ion o clouds. A a echnical le el, his
amewo k would be decen alized and highly lexible, designed o enable and p omo e con en s wi h
global each and high equency. This would be achie ed hanks o inexpensi e means (gene ally, no
mone a y cos is cha ged o he end use ) and p ac ical ools a ailable o anybody o publish, sha e
and iew con en s wi hin sho delay. The e o e, he de elopmen o social-media-based mHeal h
se ices has he po en ial o p omo e he ec ui men and ein o ce he engagemen o use s and hei
communi ies. I could also enable as , lexible, use -o ien ed and use -con olled con igu a ion o
mHeal h a chi ec u es and as and inexpensi e s uc u al deploymen . To do so, mHeal h apps ha
use social media o manage and sha e pe sonal biomedical da a in an au oma ic way can be buil by
means o he public Applica ion P og amming In e aces (APIs) exposed by social media companies.
An associa ed issue ha mus be add essed when building social-media-based se ices is ha such
APIs do no allow a high a io o da a sen pe pos (e.g., Twi e limi a ion o cha ac e s). Howe e ,
mo e impo an ly, in e ope abili y, secu i y and p i acy conce ns should no be o e looked.
Conside ing all he abo e, o da e, he bes app oxima ion o an comp ehensi e sys em was
conduc ed by T ian a yllidis e al. [
19
], who used a social media (Twi e ) o moni o pa ien da a.
In o de o uni ocally desc ibe he symp oms o ale s wee ed by he pa ien , his p oposal makes
use o he Sys ema ized Nomencla u e o Medicine—Clinical Te ms (SNOMED-CT) [
38
] and he
Uni ied Medical Language Sys em (UMLS) [
39
] me a hesau us API. Ne e heless, he biomedical
message is no o ma ed acco ding o any biomedical s anda d—lea ing aside he hesau us. Thus, i
could no in e ope a e seamlessly wi h a HIS. Wi h espec o i s secu i y policy, i is s ic ly based on
Twi e - ela ed ea u es. In pa icula , hey make use o lis s o use s o con ol he p i acy o people
subsc ibing o a se ice. They also use Open Au ho iza ion (OAu h) [
40
] o au ho ize he au oma ed
sending and ecei ing o wee s om a use accoun . Finally, hei sys em elies on de aul HTTPS o
secu e communica ions. Ne e heless, his scheme does no implemen end- o-end secu i y, and hus
he in o ma ion can be accessed in clea in he Twi e se e s. An enhancemen o such amewo k
was p oposed by he au ho s in [
20
], whe e a p oo -o -concep sys em o ollowing up ca dio ascula
pa ien s using Twi e and HL7 was implemen ed. Howe e , i was s ill es ic ed o a speci ic medical
s anda d and he secu i y and p i acy measu es implemen ed we e jus hose buil -in by Twi e .
Mo eo e , he sys em elied in a adi ional clien -se e a chi ec u e, which educed he possibili ies
o use s c ea ing hei own sys ems and hus empowe hemsel es.
E o s in he li e a u e ha e p oposed some speci ic amewo ks pa ially ul illing he
equi emen s o mH3S se ices. To da e, howe e , he e is no p oposal in eg a ing all hese conce ns in
a single sys em. The e o e, in his pape we p opose a gene ic a chi ec u e o easy- o-deploy mHeal h
se ices based on social media, which con ey in o ma ion in compliance wi h medical s anda ds,
while enhancing secu i y and p i acy o end use s. Thus, he a chi ec u e will enjoy he ad an ages
ha a social media ne wo k p o ides, such as a use /pa ien social ne wo k, buil -in eliabili y and
scalabili y, o up- o-da e GDPR-complian se e s, while con eying s anda dized biomedical da a wi h
enhanced secu i y and p i acy measu es applied.
3. Resul s
As a esul o he analysis pe o med in Sec ion 2, a gene ic mH3S a chi ec u e is p oposed wi hin
his sec ion. The de ails o he p oposed pla o m and a p oo -o -concep he eo a e ho oughly
desc ibed in he ollowing subsec ions.
Elec onics 2020,9, 2208 8 o 26
3.1. P oposal o a Gene ic mH3S A chi ec u e
The newly c ea ed a chi ec u e is illus a ed in Figu e 2. I depic s a sys em o s o ing, communica ing,
and consul ing medical da a as well as gene a ing and dis ibu ing ala ms, while b idging he adi ional
communica ion gap be ween use s, o mal and in o mal ca egi e s and esea che s.
Elec onics 2020, 9, x FOR PEER REVIEW 8 o 26
Figu e 2. P oposal o he gene ic a chi ec u e.
3.1.1. Ac o s o he P oposed A chi ec u e
The gene al a chi ec u e is comp ised o eigh ac o s (see Figu e 2):
• Use /Pa ien : They will ga he he biomedical in o ma ion and use a mobile phone o able o
pos hem. They could also ecei e eedback om o mal o in o mal ca egi e s.
• PHDs: The medical de ices used o moni o he use ’s s a us.
• In o mal ca egi e : A iend, a neighbo o a ela i e ha helps he use /pa ien o moni o and
con ol hei biomedical da a wi hin he heal hy ange o alues. In o mal ca egi e s a e
ypically in cha ge o a educed numbe o use /pa ien s.
• Fo mal ca egi e : A heal hca e p o ide associa ed o a p o essional o mal sys em, e.g.,
physicians, nu ses o social wo ke s. They a e able o ake ca e o a ela i ely la ge numbe o
pa ien s.
• EHR/HIS: This ep esen s he adi ional EHR/HIS. A p oxy would be equi ed o ecei e he
biomedical da a (by means o he social media API), dec yp , decode and send such
in o ma ion—wi h an app op ia e o ma – o he ac ual EHR/HIS, which would s o e he
in o ma ion. This may need u he policies in o de o ensu e he co ec iden i ica ion o he
sende .
• Cloud p o ide s: Such sys ems may be used o u he compu ing (e.g., signal p ocessing),
edis ibu ion o massi e s o age.
• Resea che : They would pe o m medical esea ch o e a po en ially high olume o
anonymized biomedical in o ma ion ho ough app op ia e da a mining. The open lock in Figu e
2 means ha he con en has been dec yp ed—and anonymized— o esea ch pu poses.
• Social media: The co e o he p oposed sys em, which is used as a backbone o communica ion
and, o a ce ain ex en , s o age pu poses.
I is wo h no ing howe e , ha in speci ic pa icula iza ions o he a chi ec u e, some o hem
may no be p esen . Fo example, in a simple sys em, a use /pa ien and an in o mal ca egi e may
use he social media o exchange biomedical in o ma ion. In a o mal scena io, a pa ien would
communica e wi h a o mal ca egi e h ough he applica ion.
The in ended a ge audience o he sys em depends on he scena io. In a scena io wi h a o mal
ca egi e , hey should e alua e he si ua ion, he pa ien , and hei li e acy (bo h gene al and
echnological). A e he assessmen , he o mal ca egi e would decide whe he o no p esc ibe he
Figu e 2. P oposal o he gene ic a chi ec u e.
3.1.1. Ac o s o he P oposed A chi ec u e
The gene al a chi ec u e is comp ised o eigh ac o s (see Figu e 2):
•
Use /Pa ien : They will ga he he biomedical in o ma ion and use a mobile phone o able o
pos hem. They could also ecei e eedback om o mal o in o mal ca egi e s.
•PHDs: The medical de ices used o moni o he use ’s s a us.
•
In o mal ca egi e : A iend, a neighbo o a ela i e ha helps he use /pa ien o moni o and
con ol hei biomedical da a wi hin he heal hy ange o alues. In o mal ca egi e s a e ypically
in cha ge o a educed numbe o use /pa ien s.
•
Fo mal ca egi e : A heal hca e p o ide associa ed o a p o essional o mal sys em, e.g., physicians,
nu ses o social wo ke s. They a e able o ake ca e o a ela i ely la ge numbe o pa ien s.
•
EHR/HIS: This ep esen s he adi ional EHR/HIS. A p oxy would be equi ed o ecei e
he biomedical da a (by means o he social media API), dec yp , decode and send such
in o ma ion—wi h an app op ia e o ma – o he ac ual EHR/HIS, which would s o e he
in o ma ion. This may need u he policies in o de o ensu e he co ec iden i ica ion o
he sende .
•
Cloud p o ide s: Such sys ems may be used o u he compu ing (e.g., signal p ocessing),
edis ibu ion o massi e s o age.
•
Resea che : They would pe o m medical esea ch o e a po en ially high olume o anonymized
biomedical in o ma ion ho ough app op ia e da a mining. The open lock in Figu e 2means ha
he con en has been dec yp ed—and anonymized— o esea ch pu poses.
•
Social media: The co e o he p oposed sys em, which is used as a backbone o communica ion
and, o a ce ain ex en , s o age pu poses.
Elec onics 2020,9, 2208 9 o 26
I is wo h no ing howe e , ha in speci ic pa icula iza ions o he a chi ec u e, some o hem may
no be p esen . Fo example, in a simple sys em, a use /pa ien and an in o mal ca egi e may use he
social media o exchange biomedical in o ma ion. In a o mal scena io, a pa ien would communica e
wi h a o mal ca egi e h ough he applica ion.
The in ended a ge audience o he sys em depends on he scena io. In a scena io wi h a
o mal ca egi e , hey should e alua e he si ua ion, he pa ien , and hei li e acy (bo h gene al and
echnological). A e he assessmen , he o mal ca egi e would decide whe he o no p esc ibe
he use o he applica ion. In an in o mal scena io, bo h he use and he in o mal ca egi e should
e alua e he ad an ages and d awbacks o he ool a hei disposal. I i is a hospi al de eloping he
applica ion, p ope in o ma ion abou he implica ions could be published so ha use s could make an
in o med decision.
3.1.2. Con igu a ion, Usage and Communica ion Flow
The se -up equi es minimal con igu a ion. Use s will be asked o ins all a speci ic mobile
applica ion (which should be de eloped o o e he mH3S se ice). Such applica ion shall ely
in e nally on a social media API. Use s will he e o e need an accoun o his social media ne wo k.
Since one o he main ea u es o he amewo k is i s secu i y and p i acy, i s o all, he use s—pa ien s,
o mal and in o mal ca egi e s, a hospi al—willing o exchange in o ma ion need o gene a e a pai o
public and p i a e keys. The public key has o be exchanged be o ehand. Fo secu i y easons, he
public key mus be sen h ough a communica ions channel di e en han he one used o biomedical
da a exchange. This could be done by a ious means, e.g., by scanning a Quick Response (QR) code.
The de ails o he secu i y and p i acy scheme a e de ailed la e in Sec ion 3.1.4. Once he public keys
ha e been exchanged, he communica ion can begin. Howe e , depending on he social media chosen
and on how hei accoun s a e con igu ed, he use s may be equi ed o be iend each o he be o ehand.
This can also be done inside o ou side he main applica ion. Wi hin he applica ion howe e , he
ecei e has o in oduce he name o he sende in he social media ne wo k. This will wo k as a
subsc ip ion o he pos s o he sende .
A e he con igu a ion p ocess ends, he biomedical in o ma ion would be ga he ed and
subsequen ly s anda dized, enc yp ed and embedded in an objec media, p omo ing he iendly use
o social media ne wo k (see Sec ion 3.1.4 as well). Immedia ely a e wa d, he use /pa ien would pos
he media objec wi h embedded in o ma ion o he social media (see Figu e 2). All use s subsc ibed
o he sende pos s— o example, an in o mal ca egi e —will ecei e no i ica ion o he upda e and
would be able o dec yp he biomedical in o ma ion, gi en he key exchange ca ied ou be o ehand.
The e o e, he social media would be used as a backbone.
Subsequen ly, he in o ma ion ecei ed could be s o ed o analyzed. The ecei e may answe
o he ini ial sende jus o acknowledgemen o o ecommend o p esc ibe some hing, and he
communica ion could be na u ally ca ied ou in an analogous way. In any case, he in o ma ion
could be also ansmi ed o a HIS/PHR and i could be anonymized and sen o a cloud p o ide o
esea ch pu poses.
3.1.3. Eligible Social Media
The a chi ec u e p oposed elies in exis ing social media, such as Twi e , Facebook, e c. In o de
o a social media o be eligible o se ing as backbone ne wo k o he p oposed mHeal h a chi ec u e,
i should mee a numbe o compulso y equi emen s:
•
The social media mus o e a public API, so ha he medical in o ma ion can be exchanged
p og amma ically. Mos social media oday o e his op ion.
•
The social media mus allow he sha ing o in o ma ion among use s. Na u ally, his is some hing
ha mos social media enable, since i is pa o he ounda ion o social media hemsel es.
Elec onics 2020,9, 2208 16 o 26
al hough he modi ica ion could be decided unila e ally, hey a e always announced in ad ance.
Social media o e empo al windows o adap he applica ions elaying on hei API o he new
si ua ion. I he API is discon inued o he c i e ia o being an eligible social media a e no
longe me , he mHeal h se ice as i is would be o ced o de ini i e closu e. Such eliance on
co po a ions could be g appled wi h by using—o implemen ing—an open, decen alized social
media, such as he Twi e -like mic oblogging sys ems GNU social [
49
] o Mas odon [
50
], o he
Facebook-like, non-p o i , use -owned, dis ibu ed, social media Diaspo a [
51
]. Howe e , use s
and ca egi e s would be o ced o mig a e o—o a leas c ea e a new accoun o — his se ice,
while he numbe o ac i e use s in he main social media—Twi e , Facebook, e c.—is a he big,
which clea ly eases adop ion. Use s may join exis ing se e s o open social media, which could
be a simple op ion, o al e na i ely, se e s could be ins alled and con olled by use hemsel es
o by o mal o in o mal ca egi e s. Howe e , ins alling and main aining a se e is no a simple
p ocess o mos use s. Wi h he p oposal depic ed in his pape , howe e , he only equi ed
se -up is he ins alla ion o he applica ions in mobile phones, which is easy and s aigh o wa d,
e en o use s wi h educed echnological li e acy.
Ano he poin o deba e is he p oposal o using social media as a backbone. As opposed o
adi ional clouds, social media o e limi ed capabili ies o s o age and sea ching, and hey do
no p o ide compu ing powe wha soe e . None heless, social media a e use o ien ed, and hey
ce ainly enable da a communica ion, which may su ice o mos elemoni o ing use cases (see
Sec ion 4.2. o a speci ic analysis on speed and da a capaci y).
The decen alized na u e o he a chi ec u e p esen ed he e is also deba able. The au ho s p opose
an a chi ec u e based on social media opposed o cloud se ices, while social media a e hemsel es
cloud based. Howe e , social media a e used in he sys em as a backbone. In adi ional a chi ec u e
p oposals, he biomedical da a we e con eyed h ough a ypical clien /se e a chi ec u e o a
compu ing cloud o sen o a single HIS loca ed a he hospi al, whe e hey would be pe sis ed.
By using he p oposed a chi ec u e, he biomedical da a a e sca e ed h oughou he mobile
de ices o he use s/pa ien s and o mal/in o mal ca egi e s. Thus, he a chi ec u e is, by na u e,
decen alized. Biomedical da a do go h ough and a e s o ed in he se e s o he social media
selec ed, bu enc yp ion p e en s da a om being accessed by hem.
The ac ual wo h o in oducing social media is some hing ha equi es u he esea ch and alls
beyond he scope o he pape . The pe cei ed wo hiness should be alida ed in eal scena ios.
In any case, he success o he pla o m may depend on he inal use s and hei echnological
li e acy, as well as on he speci ic a chi ec u e selec ed, among o he ac o s.
•
S anda ds and In e ope abili y: In p inciple, any s anda d is eligible o he p oposed a chi ec u e.
I is no ewo hy howe e , ha some s anda ds may be mo e sui able o a speci ic use case o may
be mo e p ac ical depending on whe he such s anda d is used in some o he place wi hin he
o e all a chi ec u e. In any case, he inclusion o ele an biomedical s anda ds in he a chi ec u e
allows he s aigh o wa d in eg a ion o he ga he ed biomedical heal h in o ma ion in heal hca e
sys ems, whene e equi ed.
•
Secu i y and p i acy: The claimed secu i y and p i acy is p o ided by he end- o-end en elope
enc yp ion. This p o ides con iden iali y, ia a combina ion o symme ic and public-key
enc yp ion. Addi ionally, he digi al signa u es p o ide au hen ica ion and in eg i y alida ion.
The in eg i y alida o could be c ucial in an embedding applica ion like his, since i can be used
o e i y ha he con en o he message has no be ampe ed wi h.
The p oposed a chi ec u e only adds one manda o y laye o enc yp ion, ha en o ced by openPGP.
The es o he secu i y measu es a e buil -in ea u es p o ided by he social ne wo k o he
in e ne p o ocols. Gene ally, he use s canno op -ou o such measu es, bu hey a e applied in a
anspa en way.
Once con igu ed, he use o he applica ion is s aigh o wa d. Howe e , a ade-o be ween
secu i y/p i acy and usabili y is always p esen when designing echnological a chi ec u es. In

Elec onics 2020,9, 2208 17 o 26
gene al, including addi ional secu i y measu es educes usabili y. In he p oposed app oach, he
secu i y and p i acy p ocedu e added implies he managemen o a pai o public/p i a e keys
(gene a ion, exchange and s o age). Use s mus c ea e a pai o public/p i a e keys. This has
o be done only once and i is as easy as pushing a bu on. The exchange o he public key is
only equi ed when con ac ing o a new use . The p ocedu e could also be simple, o example,
ansmi ing he key h ough a messaging app o using a QR code. Finally, use s would only ha e
o manage a way o access he keys s o ed in he key ing (e.g., passwo d, pin, d awn pa e n o
e en biome ic au hen ica ion, such as inge p in o acial ecogni ion). Thus, he addi ional
complexi y o he pla o m could be conside ed low. Al hough he added complexi y is na u ally
a subjec i e ma e , he au ho s conside ha he imp o ed secu i y and p i acy ou weighs he
dec eased usabili y.
The secu i y and p i acy measu es p esen ed he ein a e s uc u ed unde an end- o-end pa adigm,
om he use /pa ien de ice o he o mal/in o mal ca egi e de ice. This means ha he secu i y
p o ocols used a e s acked as di e en laye s ha pa ially o e lap each o he , p o iding secu i y
along he whole communica ion and s o age p ocess. Fo example, he eques s o he social media
API may be ca ied ou using HTTPS/TLS, bu his co e s only he segmen om he de ices o
he social media se e . The addi ional secu i y en elope p o ides u he enc yp ion so ha he
in o ma ion is no in accessible e en wi hin he social media se e s.
Fo inc eased secu i y, he algo i hms selec ed shall be—whene e possible—di e en om and
complemen a y o he algo i hms al eady de ined by he physical ne wo k, he social media and
he biomedical s anda d—i hey implemen any. This way, i one c yp og aphic algo i hm is
comp omised, he o he may s ill be alid. In any case, i any o he chosen algo i hms o a
speci ic implemen a ion we e comp omised, i would ha e o be eplaced wi h ano he exis ing
algo i hm conside ed secu e and sui able o he applica ion a such ime.
Use s/pa ien s may con igu e mul iple ecei e s o biomedical in o ma ion wi hin he applica ion.
The a chi ec u e p oposed implies he need o c ea ing one enc yp ion pe ollowe . This may
cause scalabili y issues, al hough his is no a p oblem in small-scale p ojec s wi h ew o mal
o in o mal ca egi e s. This end- o-end enc yp ion is equi ed in o de o p e en po en ial
ea esd opping—including he social media se ice p o ide i sel —and is widely implemen ed
in digi al se ices oday—such as Teleg am, Signal o Wha sApp.
Secu i y and p i acy could s ill be comp omised due o, o example, he use /pa ien exposing
his/he p i a e key. In ha case, he biomedical in o ma ion should be emo ed om he social
media ne wo k. In gene al, online social ne wo ks a e exposed o a a ie y o p i acy and secu i y
h ea s, such as phishing, ake p o ile c ea ion, o iden i y clone a acks [
52
]. The p oposal p esen ed
he ein is ulne able o hese a acks inasmuch as i elies on bo h he in e ne and on online social
ne wo ks. In e ne use s in gene al and online social ne wo k use s in pa icula mus be awa e
o hem and ollow gene al ecommenda ions. Fo example, hey should cus omize adequa e
p i acy se ings o he social ne wo k o build us wi h hose applica ions and pe sons ecei ing
o managing he use s’ pe sonal in o ma ion [
52
]. In addi ion, online social ne wo ks oday
o e ecommenda ion o minimize hese po en ial h ea s. Fo example, Twi e has published
a webpage o o e guidance and help wi h gene al sa e y and secu i y conce ns [
53
], such as
po en ially comp omised/hacked accoun s, epo impe sona ion accoun s, and ake accoun s,
among o he s.
In any case, o he a chi ec u e o unc ion, he concep and scope o us needs o be add essed. In
a scena io wi h a o mal ca egi e , bo h he ca egi e and he inal use mus us he applica ion.
Howe e , his could be achie ed i he applica ion is de eloped and dis ibu ed ia o icial
channels by he hospi al i sel . In addi ion, use s mus na u ally us he doc o sugges ing he
use o he applica ion. E ec i e heal h ca e is based on subs an ial us be ween pa ien s and
p o essionals, including he clinical ools hey choose o use [
54
]. In any case, pa ien s should be
p ope ly in o med and should sign an in o med consen be o ehand. Fo he in o mal ca egi e
Elec onics 2020,9, 2208 18 o 26
scena io, he use and he ca egi e mus us each o he . This could be easily achie ed, since hey
know each o he pe sonally and would e en be ela i es. They mus also us he applica ion.
Such applica ion could be he same one as in he o mal ca egi e scena io, ha is, de eloped and
published by a hospi al hey us . Finally, ega dless o he scena io, he applica ion code could
be open sou ce in o de o enhance us h ough anspa ency.
Ha ing his concep o us among use s de eloped, he e is no eason o mis us he o he
end. The e o e, he exchange o he public keys is always pe o med in a way ha can always be
conside ed us wo hy. The exchange o public keys could be easily pe o med in pe son. Fo he
o mal ca egi e scena io, he exchange is pe o med a he momen o en olling in he p og am,
o example, by showing a p in ed QR code o he o he use o scan. Ne e heless, i he public
keys a e sen h ough he in e ne , he e a e mechanisms o con i m ha he keys ha e no been
ampe ed wi h. Fo example, use s could gene a e a oken o he public keys (e.g., a hash) and
exchange hem h ough an al e na i e channel (e.g., hey would be sho enough o be ead o e
he phone).
All said, despi e o he measu es p oposed, no secu i y scheme is 100% secu e o p i a e. The e is
always oom o secu i y o p i acy b eaches. Howe e , he p oposed a chi ec u e p o ides use s
wi h a amewo k ha is secu e and p i a e enough o c ea e a easible mobile heal h scena io. In
gene al, a p ojec like his does equi e cons an main enance and supe ision. Speci ically, hose
pa s ela ed o c yp og aphy (lib a ies, algo i hms, keys, e c.) migh be comp omised and need
special a en ion.
•
P esen a ion: We p oposed embedding he biomedical da a blob in o a mul imedia con en . The
a ionale behind his decision is wo old. Fi s , because o he social bene i : i is clea ly iendlie
o sha e a beau i ul image han an a ay o appa en ly nonsense by es. Addi ionally, he e is
also an economy ac o , measu ed in numbe o pos s pe biomedical da a. Social media usually
es ic he maximum amoun o cha ac e s (e.g., 140— ecen ly 280—Unicode codepoin s in a
wee ) o he maximum size o an image o ideo (a ew megaby es usually) sha ed in a single
pos . Thus, he mul imedia app oach was selec ed since i is possible o embed a mo e da a in a
single social media pos i a mul imedia con en is used, compa ed o plain ex . In exchange, he
app oach selec ed inc eases he o e head and he p ocessing load, which implies highe delays.
Ne e heless, his is no an issue, since he a chi ec u e is no in ended o wo k in eal ime, and
he enc yp ion algo i hms a e as enough, gi en he amoun o biomedical da a ha i in a single
mul imedia con en .
•
Embedding: Designe s ha e o ake in o accoun ha social media may ampe wi h he uploaded
mul imedia con en in a way ha may no be public, and, i he embedding algo i hm selec ed is
no obus enough, he in o ma ion could be co up ed pa ially o comple ely. In he p oposed
a chi ec u e, a obus enough embedding algo i hm could p e en in p ac ice any po en ial da a
co up ion. Howe e , i he da a a e co up ed, he ecipien will de ec ha he in o ma ion has
no been ecei ed p ope ly. The en elope p o ides a mechanism o check his. Thus, he ecei e
could ask o a e ansmission. Then, he embedding could be ca ied ou using a mo e obus
algo i hm o p e en he same p oblem om occu ing again. The sende ’s applica ion could
be p o ided wi h a pool o embedding algo i hms, o de ed in e ms o obus ness, so ha he
simples / as es one would selec ed, as long as he da a could be e ie ed a he o he end. In any
case, he e is a ade-o be ween speed and obus ness.
4.2. Discussion abou he P oo o Concep
•
Social Media: P i a e messages ha e been used o add ess a pa icula ecei e . Howe e , use s
may exchange o he images aside om he heal hca e se ice. I ha is he case, checking he
alpha channel and looking o he openPGP da a s uc u e is an e ec i e mechanism o de ec i
he e is ac ual biomedical in o ma ion embedded in o he pic u e.
Twi e o e s a kind o g oups ( e e ed o as “lis s”), which can be public o p i a e. Howe e ,
Elec onics 2020,9, 2208 19 o 26
use s a e no allowed o pos di ec ly o a lis , so ha e e y membe o he lis ecei e he wee .
Al e na i ely, in he sys em p oposed, use s can send a p i a e message o all ecei e s (see he
pa ag aph below).
We p opose sha ing he pic u es h ough p i a e messages, al hough his could be ca ied ou ia
he public imeline. While he o me would enable he possibili y o mul iple ecei e s in one
single pos /image, he la e enhances p i acy.
As ega ds o he Twi e API maximum a es, i can be no ed ha hey a e gene ous enough o
s o e-and- o wa d, small-scale p ojec s, bu hey may no be enough in scena ios equi ing highe
da a a e ansmission o when dealing wi h a as amoun o use s/pa ien s and/o ca egi e s.
Twi e o e s a ee sea ch API, bu i is es ic ed o he pas se en days and i is ocused on
ele ance—and no in comple eness. This limi s he use o Twi e as a eliable pe sis ence handle .
Howe e , hey o e some en e p ise, high-le el sea ch APIs ha could be implemen ed wi h a
mone a y cos .
Finally, since use s only publish images ha a e aligned wi h hei as e and he opic o hei
usual social pos s, he sys em is minimally in asi e wi h ega d o in o ma ion noise. In any case,
use s may also c ea e a di e en accoun o his pu pose.
•
S anda ds and In e ope abili y: Ve sion 2 o HL7 is su icien o he use case po ayed in he
p oo o concep —a use connec ing wi h an in o mal ca egi e . The amoun o da a ansmi ed in
an HL7 e sion 2 message does no gi e ise o any issues – ega ding he embedding, o example.
In he p oo o concep , he e is no in eg a ion wi h pe sonal heal h eco ds o highe HIS, so one
can en ision ha he use o a s anda d may no be necessa y. Howe e , he implemen a ion o
HL7 lea es open he possibili y o u he in eg a ion.
•
Secu i y and P i acy: We chose o implemen openPGP o e CMS. Technical di e ences among
hem a e no decisi e. Indeed, bo h can be used as a base o Secu e/Mul ipu pose In e ne Mail
Ex ensions (S/MIME), which is an In e ne Enginee ing Task Fo ce (IETF) s anda d o public key
enc yp ion and digi al signing o MIME da a. CMS is widely used by companies by building
a PKI h ough X.509 ce i ica es, which had o be app o ed and signed by a CA be o ehand,
which usually equi es some kind o paymen . openPGP, on he o he hand, p o ides an open,
decen alized, ee sys em o ce i ica e managemen , which a e he main easons why we chose
o implemen his en elope.
The secu i y lib a y chosen, Spongy Cas le, based on Bouncy Cas e, cu en ly suppo s all
e sions o And oid. Howe e , And oid has announced he dep eca ion o Bouncy Cas le.
The e o e, he implemen ed applica ions will equi e changing o (mos likely) he de aul And oid
implemen a ion in he o hcoming u u e.
Rega ding OAu h, we ha e implemen ed e sion 1.0a. Twi e suppo s bo h e sions 1.0a and
2.0. Howe e , only e sion 1.0a can be used o pos ing on behal o ano he accoun .
•
P esen a ion: The image sha ed in he p oo o concep is p e-s o ed in he applica ion. Howe e ,
in o de o enhance he use sa is ac ion and pe sonal binding, he applica ion could be imp o ed
by le ing use s upload hei own images.
We ha e selec ed an image wi h a RGBA/ARGB colo model in o de o modi y he alpha channel
while embedding he in o ma ion. One could en ision ha agged o ma s could be selec ed
o embed he in o ma ion in ags, he eby elimina ing dis o ion. Howe e , his could no be
implemen ed in his speci ic p oo o concep , since, a he momen o w i ing, Twi e s ips and
disca ds he me ada a om uploaded images.
•
Embedding: The algo i hm implemen ed wo ks lawlessly unde he condi ions documen ed. As
he anspa ency in o ma ion is being modi ied, he p ocess na u ally dis o s he image. Less
dis o ing and mo e obus algo i hms could be implemen ed. Howe e , as discussed abo e,
he e is a ade-o be ween speed and obus ness/dis o ion. In ou implemen a ion, e sion 2
o he embedding algo i hm esul s in less noisy images, bu i akes longe o comp ess hem
(see Table 2).
Elec onics 2020,9, 2208 20 o 26
The ac ha he image becomes dis o ed does no ha e any e ec on he “clinical” unc ioning o
he pla o m, since he pic u e i sel has no clinical meaning. Howe e , a dis o ed image may
diminish he “social” componen o he pla o m. Unde he condi ions es ed, wi h only a ew
by es o in o ma ion being ansmi ed, li le dis o ion can be pe cei ed by he human eye (see
and zoom he dog pic u e in Figu e 4b).
Addi ionally, since he social media p ocessing algo i hm may no be public and, mo eo e , he
owne s o he social media can change he in e nal algo i hms o e ime wi hou no i ica ion,
he e is a need o e iew and upda e he p oposed embedding algo i hms pe iodically. I is wo h
no ing as well ha , since social media selec hei p ocessing algo i hms, an embedding algo i hm
ha wo ks o a social media may no wo k o ano he .
•
Da a capaci y: In o de o discuss he pe o mance o he p oposed echnical p oo o concep ,
we p esen below an analysis in e ms o da a capaci y and speed (a he ansmi e end). Two
scena ios we e conside ed: he one p oposed o p oo -o -concep (a small-scale elemoni o ing
sys em ansmi ing jus a ew by es) and a bounda y scena io o he p oo o concep (when all
he space a ailable is used up). In o de o assess he speed, wo sizes o images we e es ed o
he small-scale scena io. Fo he bounda y scena io, bo h e sions o he embedding algo i hm
we e es ed.
Fo calcula ing he space a ailable in he la e scena io, some calcula ions ha e o be pe o med,
acco ding o Twi e ’s cu en image suppo policies [
55
]. Fo his p oo o concep , we ha e
decided o use he alpha channel o embedding pu poses. Among he image o ma s suppo ed
by Twi e , WebP and Po able Ne wo ks G aphics (PNG) p o ide such channel. Howe e , Twi e
s a es ha i will anscode all WebP o ma s o 85% quali y Join Pho og aphic Expe s G oup
(JPEG) wi h 4:2:0 ch oma subsampling. This may dis o he embedded in o ma ion. A PNG-32
(8 bi s pe channel ARGB), on he o he hand, will be le as-is i he image has mo e han 256
colo s and i is 900 pixels o smalle in he longes dimension ( ha is, i can i in o 900
×
900). I i
has ewe colo s, i will be anscoded o PNG-8. I i is g ea e han 900
×
900, i will be es ed o
conside i hey will emain PNG o i hey will be con e ed o JPEG, being he la e mo e likely.
Thus, i hese equi emen s a e me , an image wi h 900
×
900 pixels and alpha channel could
be uploaded and i will no be anscoded by Twi e . Thus, he ecei e applica ion will be
able o e ie e he embedded in o ma ion lawlessly. Gi en ha a pixel is composed o 4 by es
(ARGB/RGBA), a maximum o 900
×
900 =810,000 alpha-channel by es a e a ailable. Howe e ,
no all o hose by es would be ac ual biomedical in o ma ion. Fi s , he begin PGP and end
PGP lines, he e sion line, he digi al signa u e and he a o emen ioned edundancy check shall
be deduc ed (95 by es). The emaining da a a e encoded in Radix-64. In such encoding, 4*(n/3)
cha ac e s a e needed o ep esen n by es, and his needs o be ounded up o a mul iple o 4. In
his case, 1 by e is needed o padding. A e decoding, he esul ing da a include he session key
enc yp ed wi h he RSA algo i hm (in ou p oo -o -concep , once enc yp ed, i s leng h is 2048
bi s, ha is, 256 by es) and he biomedical da a enc yp ed wi h he AES session key. AES wo ks
in blocks, in ou case o 128 bi s (16 by es). The e o e, some padding may be needed. In his
si ua ion, 4 padding by es a e equi ed. A he end, a maximum o 607,168 by es o biomedical
da a would be a ailable.
The amoun o biomedical da a could i in app oxima ely 600 kB depends on he da a hemsel es
and he medical s anda d selec ed, i any. Fo example, he size o he HL7-complian message
gene a ed in he p oo o concep (including weigh , hea a e, oxygen sa u a ion and blood
p essu e) is jus 439 by es in HL7- o ma ed plain ex , 1035 by es a e enc yp ing and encoding.
In a 900
×
900 image, i will dis o only he i s ow and pa ly he second ow (ou o 900 ows).
A non-comp essed SCP-ECG ile (12 ECG leads, 1000 samples/second, 2 by es/sample, 10 s) is jus
240 kB o ECG da a, plus a ew by es o SCP-ECG me ada a. A numbe o comp ession echniques
could be applied o enable la ge egis a ion imes. Fo he case o digi al images, a DICOM ile
would also ha e some me ada a, bu a pic u e o 600 kB would p obably ha e enough quali y
Elec onics 2020,9, 2208 21 o 26
o a elemoni o ing scena io (e.g., elede ma ology). I is wo h no ing ha DICOM i sel also
suppo s image comp ession using JPEG2000 [
56
], and i could be used in e nally. Al hough he e
is lack o consensus abou he ole able deg ee o comp ession, JPEG2000 ha e been epo ed
o allow comp ession a ios anging 30:1–50:1 wi hou a ec ing he clinical quali y. E en ha
could be enhanced by selec ing an op imized, ad-hoc pa ame iza ion o he speci ic medical use
case [57]. Thus, wi h a 40:1 comp ession a io, a 600 kB comp essed image would mean a 24 MB
aw image (8 megapixels RGB), which should be enough o a elede ma ology consul a ion.
Fu he mo e, his is conside ing jus one wee . Na u ally, mo e han one wee could be used,
enabling he ansmission o la ge amoun s o da a (e en he whole EHR) in a ew wee s. The
size o an EHR depends on a a ie y o ac o s. As a e e ence, in a 2011 epo o he Be h Is ael
Deaconess Medical Cen e , a eaching hospi al o Ha a d Medical School, i was published ha
hey gene a e 1 e aby e o clinical ex da a (s uc u ed and uns uc u ed) pe yea and 19 e aby e
o image da a pe yea [
58
]. Wi h 250,000 ac i e pa ien s a ha ime, ha means 80 megaby es
pe pa ien pe yea . Tha ansla es o app oxima ely 138 wee s pe pa ien pe yea , ha is, one
wee pe pa ien e e y 2–3 days.
•
Speed: The speed o he whole p ocedu e depends on a a ie y o ac o s. Fo example, he
p ocesso , he p og amming language, he p og amming coding e iciency, he amoun o da a
o be embedded, he size o he image, he embedding algo i hms and he size o he keys,
among o he s.
Rega ding he c yp og aphy algo i hms, he es da a published in [
59
] ha e been used as a
e e ence. Such es s we e un wi h an In el Co e 2 a 1.83 GHz (simila o cu en mobile de ices).
The algo i hms selec ed a e hose compa able o ou p oo o concep : AES (128-bi key) and
RSA (2048-bi key). The esul s a e shown in Table 2 o he wo a o emen ioned scena ios. Such
able also shows he ime equi ed o encoding in Radix-64, conside ing he same 1.83 GHz
p ocesso . The speed used o he calcula ions is 2 cycles/by e, which is a ypical speed o base64
encoding/decoding, al hough i could be up o en imes as e using ec o ins uc ions [
60
]. The
da a show ha enc yp ing and encoding is a ma e o milliseconds (sub o al 1, ow numbe 05 in
Table 2), e en o ~600 kB o da a. The mos ime-consuming ask would be he RSA enc yp ion
( ow 02 in Table 2) o bo h scena ios.
As ega ds he ime equi ed o embedding, he e a e no publicly a ailable da a, because his
is no a s anda d algo i hm. The au ho s ha e pe o med a pool o es s unde he ollowing
condi ions. The p ocess measu ed includes opening a bi map image, se ing up he necessa y
a iables, modi y he equi ed alpha-channel by es and gene a e a comp essed ile in PNG o ma
in ARGB/RBGA.
Fi s , he sys em was es ed o he small-scale elemoni o ing scena io. Mo e speci ically, 1035 by es
(co esponding o he size o he enc yp ed HL7-complian message) we e embedded in o he
alpha channel o he dog pic u e in Figu e 4b, es ing wo sizes: 900
×
900 pixels and a escaled
e sion o 225
×
225 pixels (exac ly 16 imes less pixels). Secondly, we pe o med wo es s (one pe
each e sion o he embedding algo i hm) using all space a ailable (
900 ×900
alpha by es). The
a ionale o es ing wo di e en e sions o he embedding algo i hm is ha he PNG comp ession
uses DEFLATE [
61
], a lossless comp essed da a o ma in ol ing a combina ion o Lempel–Zi -77
comp ession algo i hm and Hu man coding. This implies ha as e comp ession could be
achie ed o images wi h small a ia ions o colo . In o he wo ds, he highe he anspa ency,
he as e he comp ession. Con e sely, he image would be noisie and he e o e less iendly.
The es s we e un in a cu en (as o 2019), mid- ange mobile phone wi h an oc a-co e MediaTek
Helio G90T, including 2 ARM Co ex-A76 a 2.05 GHz and 6 ARM Co ex-A55 a 2.0 GHz,
6 GB o RAM and unning And oid 9.0 (PPR1.180610.011). We pe o med 10 execu ions pe
es and calcula ed he a e age ime and he s anda d de ia ion. This p o ides he eade wi h
a no ion o he cen al alue o he esul s (a i hme ic mean) and he amoun o dispe sion
o he se o esul s (s anda d de ia ion). Th oughou he pape , he esul s a e exp essed as

Elec onics 2020,9, 2208 22 o 26
“mean ±s anda d_de ia ion”.
The esul s a e shown in Table 2. In gene al, he s anda d de ia ion is low compa ed o he
a e age. This sugges s ha he da a a e clus e ed a ound he mean. The embedding is as ( ow
12 in Table 2) o bo h small-scale and bounda y scena ios, aking up less han 30 ms in any case.
I is pa icula ly as o he smalle image, whe e i needs only 3.7
±
0.5 ms. When compa ing he
wo embedding algo i hms o he bounda y scena io, we can obse e ha he shi ing p ocess
i sel ( ha is, adding 0x85 o each by e) only adds 0.3 ms (5.6
±
0.5 ms s. 5.3
±
0.5 ms, ow 11
in Table 2). Fo he small-scale scena io, he wo embedding algo i hms we e no es ed, as hey
would beha e simila ly, since only a ew by es a e being modi ied, and hus he image emains
la gely una ec ed (see zoomed images in ow 07 in Table 2).
As ega ds he PNG comp ession, i can be seen ha i depends la gely on he size o he image
(38.1
±
0.6 ms a 225x255 s. he es , se e al imes slowe , a 900
×
900, ow 13 in Table 2). The
o he ac o a ec ing he PNG comp ession ime is he e sion o he embedding algo i hm. The
comp ession o he shi ed e sion is slowe compa ed o he non-shi ed one (703.5
±
4.2 ms s.
548.5
±
3.8 ms, ow 13 in Table 2). On he o he hand, he image is clea e , which could imp o e
use s’ engagemen ( ow 07 in Table 2).
The da a show ha he bo leneck o he whole p ocedu e is he PNG comp ession ( ow 13
in Table 2). App oxima ely 80-95% o he o al ime ( ow 14 in Table 2) is spen on he PNG
comp ession. In hese es s, we use he buil -in comp esso p o ided by And oid, which is
single- h eaded. I is a guably easible o pa allelize he comp ession p ocess, achie ing a
educ ion o ime ha would scale linea ly wi h he numbe o co es o he machine. Besides
he co es o he cen al p ocessing uni , mobile phones nowadays also inco po a e mul i-co e
g aphic p ocessing uni s, which could also be used o pa allelized comp ession. To he bes o
ou knowledge, he e is no And oid lib a y a ailable able o pe o m a pa allel comp ession o
PNG iles (a he momen o w i ing). The e exis , howe e , Ja a-based pa allel comp esso s
o GZIP [
62
], a ile o ma simila o PNG, since i is also based on DEFLATE. Ne e heless,
p og amming such a comp esso alls ou o he scope o his pape .
As a conclusion, o bo h he small-scale elemoni o ing scena io and he bounda y scena io, he
whole p ocess could be pe o med a he as (less han 1 s in any case). I can be ca ied ou
conside ably as e wi h smalle images: 48.2
±
0.9 ms (225
×
225) s. 719.6
±
9.6 ms (900
×
900).
In addi ion, using he non-shi ed algo i hm, he p ocess can be pe o med as e in he bounda y
scena io o he de imen o noisie images. In any case, he esul s show ha he speed would no
be a c i ical issue o mos mHeal h scena ios.
Table 2. Speed and size analysis.
Row Scena io Small-Scale Bounda y
01 Plain ex (biomedical) da a (by es) 439 607,168
02 RSA enc yp ion (ms) 6.370
03 AES se -up and enc yp ion (ms) 0.004 5.309
04 Radix-64 encoding (ms) 0.001 0.664
05 Sub o al 1: enc yp ion and encoding (ms) 6.375 12.343
06 Embedding Algo i hm Ve sion Shi ed Shi ed Shi ed
No shi ed
07 Zoomed (uppe - igh zone)
image anspa ency o e iew
Elec onics 2020, 9, x FOR PEER REVIEW 22 o 26
as hey would beha e simila ly, since only a ew by es a e being modi ied, and hus he image
emains la gely una ec ed (see zoomed images in ow 07 in Table 2).
As ega ds he PNG comp ession, i can be seen ha i depends la gely on he size o he image
(38.1 ± 0.6 ms a 225x255 s. he es , se e al imes slowe , a 900 × 900, ow 13 in Table 2). The
o he ac o a ec ing he PNG comp ession ime is he e sion o he embedding algo i hm. The
comp ession o he shi ed e sion is slowe compa ed o he non-shi ed one (703.5 ± 4.2 ms s.
548.5 ± 3.8 ms, ow 13 in Table 2). On he o he hand, he image is clea e , which could imp o e
use s’ engagemen ( ow 07 in Table 2).
The da a show ha he bo leneck o he whole p ocedu e is he PNG comp ession ( ow 13 in
Table 2). App oxima ely 80-95% o he o al ime ( ow 14 in Table 2) is spen on he PNG
comp ession. In hese es s, we use he buil -in comp esso p o ided by And oid, which is
single- h eaded. I is a guably easible o pa allelize he comp ession p ocess, achie ing a
educ ion o ime ha would scale linea ly wi h he numbe o co es o he machine. Besides he
co es o he cen al p ocessing uni , mobile phones nowadays also inco po a e mul i-co e
g aphic p ocessing uni s, which could also be used o pa allelized comp ession. To he bes o
ou knowledge, he e is no And oid lib a y a ailable able o pe o m a pa allel comp ession o
PNG iles (a he momen o w i ing). The e exis , howe e , Ja a-based pa allel comp esso s o
GZIP [62], a ile o ma simila o PNG, since i is also based on DEFLATE. Ne e heless,
p og amming such a comp esso alls ou o he scope o his pape .
As a conclusion, o bo h he small-scale elemoni o ing scena io and he bounda y scena io, he
whole p ocess could be pe o med a he as (less han 1 s in any case). I can be ca ied ou
conside ably as e wi h smalle images: 48.2 ± 0.9 ms (225 × 225) s. 719.6 ± 9.6 ms (900 × 900).
In addi ion, using he non-shi ed algo i hm, he p ocess can be pe o med as e in he
bounda y scena io o he de imen o noisie images. In any case, he esul s show ha he speed
would no be a c i ical issue o mos mHeal h scena ios.
Table 2. Speed and size analysis.
Row Scena io Small-Scale Bounda y
01 Plain ex (biomedical) da a (by es) 439 607,168
02 RSA enc yp ion (ms) 6.370
03 AES se -up and enc yp ion (ms) 0.004 5.309
04 Radix-64 encoding (ms) 0.001 0.664
05 Sub o al 1: enc yp ion and encoding (ms) 6.375 12.343
06 Embedding Algo i hm Ve sion Shi ed Shi ed Shi ed No shi ed
07 Zoomed (uppe - igh zone)
image anspa ency o e iew
08 Image dimension (wid h × heigh in pixels) 225 × 225 900 × 900
09 By es embedded (by es) 1035 900 × 900 = 810,000
10 Embedding se up (ms) 3.6 ± 0.5 20.9 ± 1.2 24.3 ± 3.2 22.3 ± 1.8
11 Embedding (ms) 0.1 ± 0.3 3.9 ± 0.9 5.6 ± 0.5 5.3 ± 0.5
12 Embedding sub o al (ms) 3.7 ± 0.5 24.8 ± 1.9 29.9 ± 3.3 27.6 ± 1.1
13 PNG comp ession (ms) 38.1 ± 0.6 688.4 ± 9.1 703.5 ± 4.2 548.5 ± 3.8
14 % PNG comp ession (100 * ow 13/ ow 16) 79.1 95.7 94.3 93.2
15 Sub o al 2: embedding and comp ession (ms) 41.8 ± 0.9 713.2 ± 9.6 733.4 ± 5.2 576.1 ± 4.1
16 TOTAL TIME: sub o al 1 + sub o al 2 (ms) 48.2 ± 0.9 719.6 ± 9.6 745.7 ± 5.2 588.4 ± 4.1
As a inal e lec ion on Sec ion 4.2, i is wo h no ing ha he elemen s selec ed o a speci ic
p oo o concep can be changed independen ly. I he designe decide o use Facebook ins ead o
Twi e , he o he pa icula selec ions (HL7, openPGP, o he embedding algo i hms) would s ill be
alid.
5. Conclusions
Elec onics 2020, 9, x FOR PEER REVIEW 22 o 26
as hey would beha e simila ly, since only a ew by es a e being modi ied, and hus he image
emains la gely una ec ed (see zoomed images in ow 07 in Table 2).
As ega ds he PNG comp ession, i can be seen ha i depends la gely on he size o he image
(38.1 ± 0.6 ms a 225x255 s. he es , se e al imes slowe , a 900 × 900, ow 13 in Table 2). The
o he ac o a ec ing he PNG comp ession ime is he e sion o he embedding algo i hm. The
comp ession o he shi ed e sion is slowe compa ed o he non-shi ed one (703.5 ± 4.2 ms s.
548.5 ± 3.8 ms, ow 13 in Table 2). On he o he hand, he image is clea e , which could imp o e
use s’ engagemen ( ow 07 in Table 2).
The da a show ha he bo leneck o he whole p ocedu e is he PNG comp ession ( ow 13 in
Table 2). App oxima ely 80-95% o he o al ime ( ow 14 in Table 2) is spen on he PNG
comp ession. In hese es s, we use he buil -in comp esso p o ided by And oid, which is
single- h eaded. I is a guably easible o pa allelize he comp ession p ocess, achie ing a
educ ion o ime ha would scale linea ly wi h he numbe o co es o he machine. Besides he
co es o he cen al p ocessing uni , mobile phones nowadays also inco po a e mul i-co e
g aphic p ocessing uni s, which could also be used o pa allelized comp ession. To he bes o
ou knowledge, he e is no And oid lib a y a ailable able o pe o m a pa allel comp ession o
PNG iles (a he momen o w i ing). The e exis , howe e , Ja a-based pa allel comp esso s o
GZIP [62], a ile o ma simila o PNG, since i is also based on DEFLATE. Ne e heless,
p og amming such a comp esso alls ou o he scope o his pape .
As a conclusion, o bo h he small-scale elemoni o ing scena io and he bounda y scena io, he
whole p ocess could be pe o med a he as (less han 1 s in any case). I can be ca ied ou
conside ably as e wi h smalle images: 48.2 ± 0.9 ms (225 × 225) s. 719.6 ± 9.6 ms (900 × 900).
In addi ion, using he non-shi ed algo i hm, he p ocess can be pe o med as e in he
bounda y scena io o he de imen o noisie images. In any case, he esul s show ha he speed
would no be a c i ical issue o mos mHeal h scena ios.
Table 2. Speed and size analysis.
Row Scena io Small-Scale Bounda y
01 Plain ex (biomedical) da a (by es) 439 607,168
02 RSA enc yp ion (ms) 6.370
03 AES se -up and enc yp ion (ms) 0.004 5.309
04 Radix-64 encoding (ms) 0.001 0.664
05 Sub o al 1: enc yp ion and encoding (ms) 6.375 12.343
06 Embedding Algo i hm Ve sion Shi ed Shi ed Shi ed No shi ed
07 Zoomed (uppe - igh zone)
image anspa ency o e iew
08 Image dimension (wid h × heigh in pixels) 225 × 225 900 × 900
09 By es embedded (by es) 1035 900 × 900 = 810,000
10 Embedding se up (ms) 3.6 ± 0.5 20.9 ± 1.2 24.3 ± 3.2 22.3 ± 1.8
11 Embedding (ms) 0.1 ± 0.3 3.9 ± 0.9 5.6 ± 0.5 5.3 ± 0.5
12 Embedding sub o al (ms) 3.7 ± 0.5 24.8 ± 1.9 29.9 ± 3.3 27.6 ± 1.1
13 PNG comp ession (ms) 38.1 ± 0.6 688.4 ± 9.1 703.5 ± 4.2 548.5 ± 3.8
14 % PNG comp ession (100 * ow 13/ ow 16) 79.1 95.7 94.3 93.2
15 Sub o al 2: embedding and comp ession (ms) 41.8 ± 0.9 713.2 ± 9.6 733.4 ± 5.2 576.1 ± 4.1
16 TOTAL TIME: sub o al 1 + sub o al 2 (ms) 48.2 ± 0.9 719.6 ± 9.6 745.7 ± 5.2 588.4 ± 4.1
As a inal e lec ion on Sec ion 4.2, i is wo h no ing ha he elemen s selec ed o a speci ic
p oo o concep can be changed independen ly. I he designe decide o use Facebook ins ead o
Twi e , he o he pa icula selec ions (HL7, openPGP, o he embedding algo i hms) would s ill be
alid.
5. Conclusions
Elec onics 2020, 9, x FOR PEER REVIEW 22 o 26
as hey would beha e simila ly, since only a ew by es a e being modi ied, and hus he image
emains la gely una ec ed (see zoomed images in ow 07 in Table 2).
As ega ds he PNG comp ession, i can be seen ha i depends la gely on he size o he image
(38.1 ± 0.6 ms a 225x255 s. he es , se e al imes slowe , a 900 × 900, ow 13 in Table 2). The
o he ac o a ec ing he PNG comp ession ime is he e sion o he embedding algo i hm. The
comp ession o he shi ed e sion is slowe compa ed o he non-shi ed one (703.5 ± 4.2 ms s.
548.5 ± 3.8 ms, ow 13 in Table 2). On he o he hand, he image is clea e , which could imp o e
use s’ engagemen ( ow 07 in Table 2).
The da a show ha he bo leneck o he whole p ocedu e is he PNG comp ession ( ow 13 in
Table 2). App oxima ely 80-95% o he o al ime ( ow 14 in Table 2) is spen on he PNG
comp ession. In hese es s, we use he buil -in comp esso p o ided by And oid, which is
single- h eaded. I is a guably easible o pa allelize he comp ession p ocess, achie ing a
educ ion o ime ha would scale linea ly wi h he numbe o co es o he machine. Besides he
co es o he cen al p ocessing uni , mobile phones nowadays also inco po a e mul i-co e
g aphic p ocessing uni s, which could also be used o pa allelized comp ession. To he bes o
ou knowledge, he e is no And oid lib a y a ailable able o pe o m a pa allel comp ession o
PNG iles (a he momen o w i ing). The e exis , howe e , Ja a-based pa allel comp esso s o
GZIP [62], a ile o ma simila o PNG, since i is also based on DEFLATE. Ne e heless,
p og amming such a comp esso alls ou o he scope o his pape .
As a conclusion, o bo h he small-scale elemoni o ing scena io and he bounda y scena io, he
whole p ocess could be pe o med a he as (less han 1 s in any case). I can be ca ied ou
conside ably as e wi h smalle images: 48.2 ± 0.9 ms (225 × 225) s. 719.6 ± 9.6 ms (900 × 900).
In addi ion, using he non-shi ed algo i hm, he p ocess can be pe o med as e in he
bounda y scena io o he de imen o noisie images. In any case, he esul s show ha he speed
would no be a c i ical issue o mos mHeal h scena ios.
Table 2. Speed and size analysis.
Row Scena io Small-Scale Bounda y
01 Plain ex (biomedical) da a (by es) 439 607,168
02 RSA enc yp ion (ms) 6.370
03 AES se -up and enc yp ion (ms) 0.004 5.309
04 Radix-64 encoding (ms) 0.001 0.664
05 Sub o al 1: enc yp ion and encoding (ms) 6.375 12.343
06 Embedding Algo i hm Ve sion Shi ed Shi ed Shi ed No shi ed
07 Zoomed (uppe - igh zone)
image anspa ency o e iew
08 Image dimension (wid h × heigh in pixels) 225 × 225 900 × 900
09 By es embedded (by es) 1035 900 × 900 = 810,000
10 Embedding se up (ms) 3.6 ± 0.5 20.9 ± 1.2 24.3 ± 3.2 22.3 ± 1.8
11 Embedding (ms) 0.1 ± 0.3 3.9 ± 0.9 5.6 ± 0.5 5.3 ± 0.5
12 Embedding sub o al (ms) 3.7 ± 0.5 24.8 ± 1.9 29.9 ± 3.3 27.6 ± 1.1
13 PNG comp ession (ms) 38.1 ± 0.6 688.4 ± 9.1 703.5 ± 4.2 548.5 ± 3.8
14 % PNG comp ession (100 * ow 13/ ow 16) 79.1 95.7 94.3 93.2
15 Sub o al 2: embedding and comp ession (ms) 41.8 ± 0.9 713.2 ± 9.6 733.4 ± 5.2 576.1 ± 4.1
16 TOTAL TIME: sub o al 1 + sub o al 2 (ms) 48.2 ± 0.9 719.6 ± 9.6 745.7 ± 5.2 588.4 ± 4.1
As a inal e lec ion on Sec ion 4.2, i is wo h no ing ha he elemen s selec ed o a speci ic
p oo o concep can be changed independen ly. I he designe decide o use Facebook ins ead o
Twi e , he o he pa icula selec ions (HL7, openPGP, o he embedding algo i hms) would s ill be
alid.
5. Conclusions
Elec onics 2020, 9, x FOR PEER REVIEW 22 o 26
as hey would beha e simila ly, since only a ew by es a e being modi ied, and hus he image
emains la gely una ec ed (see zoomed images in ow 07 in Table 2).
As ega ds he PNG comp ession, i can be seen ha i depends la gely on he size o he image
(38.1 ± 0.6 ms a 225x255 s. he es , se e al imes slowe , a 900 × 900, ow 13 in Table 2). The
o he ac o a ec ing he PNG comp ession ime is he e sion o he embedding algo i hm. The
comp ession o he shi ed e sion is slowe compa ed o he non-shi ed one (703.5 ± 4.2 ms s.
548.5 ± 3.8 ms, ow 13 in Table 2). On he o he hand, he image is clea e , which could imp o e
use s’ engagemen ( ow 07 in Table 2).
The da a show ha he bo leneck o he whole p ocedu e is he PNG comp ession ( ow 13 in
Table 2). App oxima ely 80-95% o he o al ime ( ow 14 in Table 2) is spen on he PNG
comp ession. In hese es s, we use he buil -in comp esso p o ided by And oid, which is
single- h eaded. I is a guably easible o pa allelize he comp ession p ocess, achie ing a
educ ion o ime ha would scale linea ly wi h he numbe o co es o he machine. Besides he
co es o he cen al p ocessing uni , mobile phones nowadays also inco po a e mul i-co e
g aphic p ocessing uni s, which could also be used o pa allelized comp ession. To he bes o
ou knowledge, he e is no And oid lib a y a ailable able o pe o m a pa allel comp ession o
PNG iles (a he momen o w i ing). The e exis , howe e , Ja a-based pa allel comp esso s o
GZIP [62], a ile o ma simila o PNG, since i is also based on DEFLATE. Ne e heless,
p og amming such a comp esso alls ou o he scope o his pape .
As a conclusion, o bo h he small-scale elemoni o ing scena io and he bounda y scena io, he
whole p ocess could be pe o med a he as (less han 1 s in any case). I can be ca ied ou
conside ably as e wi h smalle images: 48.2 ± 0.9 ms (225 × 225) s. 719.6 ± 9.6 ms (900 × 900).
In addi ion, using he non-shi ed algo i hm, he p ocess can be pe o med as e in he
bounda y scena io o he de imen o noisie images. In any case, he esul s show ha he speed
would no be a c i ical issue o mos mHeal h scena ios.
Table 2. Speed and size analysis.
Row Scena io Small-Scale Bounda y
01 Plain ex (biomedical) da a (by es) 439 607,168
02 RSA enc yp ion (ms) 6.370
03 AES se -up and enc yp ion (ms) 0.004 5.309
04 Radix-64 encoding (ms) 0.001 0.664
05 Sub o al 1: enc yp ion and encoding (ms) 6.375 12.343
06 Embedding Algo i hm Ve sion Shi ed Shi ed Shi ed No shi ed
07 Zoomed (uppe - igh zone)
image anspa ency o e iew
08 Image dimension (wid h × heigh in pixels) 225 × 225 900 × 900
09 By es embedded (by es) 1035 900 × 900 = 810,000
10 Embedding se up (ms) 3.6 ± 0.5 20.9 ± 1.2 24.3 ± 3.2 22.3 ± 1.8
11 Embedding (ms) 0.1 ± 0.3 3.9 ± 0.9 5.6 ± 0.5 5.3 ± 0.5
12 Embedding sub o al (ms) 3.7 ± 0.5 24.8 ± 1.9 29.9 ± 3.3 27.6 ± 1.1
13 PNG comp ession (ms) 38.1 ± 0.6 688.4 ± 9.1 703.5 ± 4.2 548.5 ± 3.8
14 % PNG comp ession (100 * ow 13/ ow 16) 79.1 95.7 94.3 93.2
15 Sub o al 2: embedding and comp ession (ms) 41.8 ± 0.9 713.2 ± 9.6 733.4 ± 5.2 576.1 ± 4.1
16 TOTAL TIME: sub o al 1 + sub o al 2 (ms) 48.2 ± 0.9 719.6 ± 9.6 745.7 ± 5.2 588.4 ± 4.1
As a inal e lec ion on Sec ion 4.2, i is wo h no ing ha he elemen s selec ed o a speci ic
p oo o concep can be changed independen ly. I he designe decide o use Facebook ins ead o
Twi e , he o he pa icula selec ions (HL7, openPGP, o he embedding algo i hms) would s ill be
alid.
5. Conclusions
08 Image dimension (wid h ×heigh in pixels) 225 ×225 900 ×900
09 By es embedded (by es) 1035 900 ×900 =810,000
10 Embedding se up (ms) 3.6 ±0.5 20.9 ±1.2 24.3 ±3.2 22.3 ±1.8
11 Embedding (ms) 0.1 ±0.3 3.9 ±0.9 5.6 ±0.5 5.3 ±0.5
12 Embedding sub o al (ms) 3.7 ±0.5 24.8 ±1.9 29.9 ±3.3 27.6 ±1.1
13 PNG comp ession (ms) 38.1 ±0.6 688.4 ±9.1 703.5 ±4.2 548.5 ±3.8
Elec onics 2020,9, 2208 23 o 26
Table 2. Con .
Row Scena io Small-Scale Bounda y
14 % PNG comp ession (100 * ow 13/ ow 16) 79.1 95.7 94.3 93.2
15
Sub o al 2: embedding and comp ession (ms)
41.8 ±0.9 713.2 ±9.6 733.4 ±5.2 576.1 ±4.1
16 TOTAL TIME: sub o al 1 +sub o al 2 (ms) 48.2 ±0.9 719.6 ±9.6 745.7 ±5.2 588.4 ±4.1
As a inal e lec ion on Sec ion 4.2, i is wo h no ing ha he elemen s selec ed o a speci ic p oo
o concep can be changed independen ly. I he designe decide o use Facebook ins ead o Twi e ,
he o he pa icula selec ions (HL7, openPGP, o he embedding algo i hms) would s ill be alid.
5. Conclusions
In his pape , a gene ic mH3S a chi ec u e has been p oposed. Such a chi ec u e p o ides use s
wi h an enhanced heal hca e se ice. The no el y o his app oach can be summa ized as ollows:
•
No el a chi ec u e: Ins ead o a clien -se e a chi ec u e o a cloud-based a chi ec u e, we
p opose an end- o-end sys em ha le e ages online social ne wo ks as a backbone.
•
Empowe ed use s and pa ien s: Use s do no ely on anyone o c ea e a secu e, p i a e
communica ion channel. They decide wha o sha e and wi h whom.
•
Use - iendly way: A ac i e and ela able media objec s (images, ideos) wi h biomedical da a
embedded a e sha ed h ough a social media ne wo k.
•
S aigh o wa d deploymen : Use s only need o ins all a mobile applica ion and pe o m some
mino con igu a ion.
•
A o dable: The use s a e no asked o spend any money o use he sys em. Only a mobile phone
wi h in e ne connec ion is equi ed.
•
High up ime a ailabili y: The sys em le e ages online social ne wo ks as a backbone. Thus, he
se e s a e almos always up.
•
Imp o ed secu i y and p i acy: This is due o he secu i y en elope, based on a hyb id
c yp osys em, he e o e combining he con enience o public-key app oaches wi h he e iciency
o symme ic-key schemes. Social ne wo ks con ey he in o ma ion, bu no e en hey a e able o
ead he biomedical da a a elling h ough hei se e s.
•
Reduced added complexi y: Use s a e equi ed o manage some key- ela ed aspec s, bu his is
common p ac ice o cu en use s o sma phones and applica ions and i could be ca ied ou
easily and swi ly.
•
Augmen ed in eg abili y: This is p o ided by he in e nal suppo o medical in e ope abili y
s anda ds.
Addi ionally, a echnical p oo -o -concep implemen a ion o such a chi ec u e has been de eloped
by selec ing a speci ic social media (Twi e ), a secu i y en elope (openPGP), an in e ope abili y
s anda d (HL7) as well as a speci ic embedding algo i hm. To accomplish such a sys em, wo And oid
applica ions we e de eloped: one o use s/pa ien s and he o he o o mal/in o mal ca egi e s. This
implemen a ion demons a es he easibili y o he pla o m. The es s show ha he p ocess is as : less
han 1 s, e en o p epa ing ( ha is, enc yp ion, encoding and embedding) ~600 kB o biomedical da a.
Thus, he addi ional complexi y o he p ocedu e does no en ail imp ac ical delays, and he e o e, he
pla o m can be conside ed as enough o mos mHeal h elemoni o ing se ices.
As a inal e lec ion, i can be highligh ed ha , al hough he a chi ec u e p esen ed and discussed
in his manusc ip has been mo i a ed by a biomedical con ex , i could ce ainly be applied o o he
con ex s. Fo example, by eplacing he medical in e ope abili y s anda d wi h ano he s anda d o
da a o ma sui able o he applica ion. The e o e, he gene ic a chi ec u e p oposed he e can ac ually
be seen as an enable o payload anspa ency.
Elec onics 2020,9, 2208 24 o 26
Au ho Con ibu ions:
Concep ualiza ion, J.D.T.,
Ó
.J.R., and
Á
.A.; me hodology, J.D.T. and
Ó
.J.R.; so wa e,
J.D.T. and M.M.-E.; alida ion, J.D.T.; o mal analysis, J.D.T. and
Ó
.J.R.; in es iga ion, J.D.T.; esou ces, J.D.T.,
J.G. and L.S.-A.; da a cu a ion, J.D.T.,
Ó
.J.R. and M.M.-E.; w i ing—o iginal d a p epa a ion, J.D.T. and
Ó
.J.R.;
w i ing— e iew and edi ing, J.D.T.,
Ó
.J.R., M.M.-E.,
Á
.A., J.G. and L.S.-A.; isualiza ion, J.D.T. and
Ó
.J.R.;
supe ision, J.G. and L.S.-A.; p ojec adminis a ion, J.D.T., J.G. and L.S.-A.; unding acquisi ion, J.D.T., J.G. and
L.S.-A. All au ho s ha e ead and ag eed o he published e sion o he manusc ip .
Funding:
This esea ch was unded by Public Uni e si y o Na a a (p ojec e e ence numbe PJUPNA29);
Minis e io de Econom
í
a, Indus ia y Compe i i idad om Gobie no de España and Eu opean Regional
De elopmen Fund ( e e ence numbe TIN2016-76770-R); Gobie no de A ag
ó
n (Re e ence G oup T31_20R);
and FEDER 2014-2020 “Cons uyendo Eu opa desde A agón”.
Con lic s o In e es : The au ho s decla e no con lic o in e es .
Re e ences
1.
Pew Resea ch Cen e Mobile Fac Shee . A ailable online: h p://www.pewin e ne .o g/ ac -shee /mobile/
(accessed on 7 Sep embe 2020).
2.
Wo ld Heal h O ganiza ion (WHO) mHeal h. New Ho izons o Heal h h ough Mobile Technologies.
A ailable online: h p://www.who.in /goe/publica ions/goe_mheal h_web.pd (accessed on 7 Sep embe 2020).
3.
Adibi, S. Mobile Heal h: A Technology Road Map; Sp inge Publishing Company, Inc.: New Yo k, NY, USA,
2015; ISBN 978-3-319-12816-0.
4.
Sil a, B.M.C.; Rod igues, J.J.P.C.; de la To e D
í
ez, I.; L
ó
pez-Co onado, M.; Saleem, K. Mobile-heal h: A
e iew o cu en s a e in 2015. J. Biomed. In o m. 2015,56, 265–272. [C ossRe ] [PubMed]
5.
Hamdi, O.; Chalou , M.A.; Oua a a, D.; K ie , F. eHeal h: Su ey on esea ch p ojec s, compa a i e s udy o
elemoni o ing a chi ec u es and main issues. J. Ne w. Compu . Appl. 2014,46, 100–112. [C ossRe ]
6.
Awo i, J.; Lee, J.M. A Make Mo emen o Heal h: A New Pa adigm o Heal h Inno a ion. JAMA Pedia .
2017,171, 107–108. [C ossRe ] [PubMed]
7.
Lee, J.M.; Hi sch eld, E.; Wedding, J. A Pa ien -Designed Do-I -You sel Mobile Technology Sys em o
Diabe es: P omise and Challenges o a New E a in Medicine. JAMA 2016,315, 1447–1448. [C ossRe ]
8.
Chen, C.; Haddad, D.; Selsky, J.; Ho man, J.E.; K a i z, R.L.; Es in, D.E.; Sim, I. Making Sense o Mobile
Heal h Da a: An Open A chi ec u e o Imp o e Indi idual- and Popula ion-Le el Heal h. J. Med. In e ne .
Res. 2012,14. [C ossRe ]
9.
Househ, M.; Bo ycki, E.; Kushni uk, A. Empowe ing pa ien s h ough social media: The bene i s and
challenges. Heal h In o m. J. 2014,20, 50–58. [C ossRe ]
10. Facebook S a is ics. A ailable online: h ps://abou . b.com/company-in o/(accessed on 7 Sep embe 2020).
11.
Kaplan, A.M.; Haenlein, M. Use s o he wo ld, uni e! The challenges and oppo uni ies o Social Media.
Bus. Ho iz. 2010,53, 59–68. [C ossRe ]
12.
Househ, M. The use o social media in heal hca e: O ganiza ional, clinical, and pa ien pe spec i es. S ud.
Heal h Technol. In . 2013,183, 244–248.
13.
Rozenblum, R.; Ba es, D.W. Pa ien -cen ed heal hca e, social media and he in e ne : The pe ec s o m?
BMJ Qual. Sa . 2013,22, 183–186. [C ossRe ]
14.
G ajales, F.J., III; Sheps, S.; Ho, K.; No ak-Lausche , H.; Eysenbach, G. Social Media: A Re iew and Tu o ial
o Applica ions in Medicine and Heal h Ca e. J. Med. In e ne Res. 2014,16, e13. [C ossRe ]
15.
Ho s-F aile, S.; A ique, S.; Maye , M.A.; Denecke, K.; Me olli, M.; Househ, M. The Unin ended Consequences
o Social Media in Heal hca e: New P oblems and New Solu ions. Yea b. Med. In . 2016, 47–52. [C ossRe ]
16.
G o e , S.; Aujla, G.S. Twi e da a based p edic ion model o in luenza epidemic. In P oceedings o he 2015
2nd In e na ional Con e ence on Compu ing o Sus ainable Global De elopmen (INDIACom), New Delhi,
India, 11–13 Ma ch 2015; pp. 873–879.
17.
Al-Bah ani, R.; Danilo ich, M.K.; Ag awal, A.; Choudha y, A. Towa ds In o mal Ca egi e Iden i ica ion in
Social Media. In P oceedings o he 2016 IEEE In e na ional Con e ence on Heal hca e In o ma ics (ICHI),
Chicago, IL, USA, 4–7 Oc obe 2016; p. 414.
18.
Nai , L.R.; She y, S.D.; She y, S.D. Applying spa k based machine lea ning model on s eaming big da a o
heal h s a us p edic ion. Compu . Elec . Eng. 2017. [C ossRe ]
Elec onics 2020,9, 2208 25 o 26
19.
T ian a yllidis, A.K.; Kou kias, V.G.; Chou a da, I.; Magla e as, N. A Pe asi e Heal h Sys em In eg a ing
Pa ien Moni o ing, S a us Logging, and Social Sha ing. IEEE J. Biomed. Heal h In o m.
2013
,17, 30–37.
[C ossRe ] [PubMed]
20.
T igo, J.D.; Eguzkiza, A.; Ma
í
nez-Esp onceda, M.; Se ano, L. A ca dio ascula pa ien ollow-up sys em
using Twi e and HL7. In P oceedings o he Compu ing in Ca diology (CinC), Za agoza, Spain, 22–25
Oc obe 2013; pp. 33–36.
21.
Lux on, D.D.; Kayl, R.A.; Mishkind, M.C. mHeal h Da a Secu i y: The Need o HIPAA-Complian
S anda diza ion. Telemed. E Heal h 2012,18, 284–288. [C ossRe ]
22. Che dan se a, Y.; Hil on, J. A Re e ence Model o In o ma ion Assu ance & Secu i y. In P oceedings o he
2013 In e na ional Con e ence on A ailabili y, Reliabili y and Secu i y, Regensbu g, Ge many, 2–6 Sep embe
2013; pp. 546–555.
23.
Velsen, L.V.; He mens, H.; d’Hollosy, W.O.N. A ma u i y model o in e ope abili y in eHeal h. In P oceedings
o he 2016 IEEE 18 h In e na ional Con e ence on e-Heal h Ne wo king, Applica ions and Se ices
(Heal hcom), Munich, Ge many, 14–17 Sep embe 2016; pp. 1–6.
24.
Tu ni sa, C.D. Ex ending he Le els o Concep ual In e ope abili y Model. In P oceedings o he IEEE
Summe Compu e Simula ion Con e ence (SCSC), New Je sey, NY, USA, 24–28 July 2005.
25.
Islam, S.M.R.; Kwak, D.; Kabi , M.H.; Hossain, M.; Kwak, K. The In e ne o Things o Heal h Ca e: A
Comp ehensi e Su ey. IEEE Access 2015,3, 678–708. [C ossRe ]
26.
Voska ides, S.; Pa ichis, C.S.; Is epanian, R.S.H.; Ky iacou, E.; Pa ichis, M.S.; Schizas, C.N. Mobile Heal h
Sys ems: A B ie O e iew. In P oceedings o he SPIE 4740, Digi al Wi eless Communica ions IV, O lando,
FL, USA, 1–5 Ap il 2002; Volume 4740, pp. 124–132.
27.
P ice, S.; Summe s, R. Clinical knowledge managemen and m-heal h. In P oceedings o he Second Join
EMBS-BMES Con e ence: 24 h Annual Enginee ing in Medicine and Biology Socie y Con e ence and he
Annual Fall Mee ing o he Biomedical Enginee ing Socie y, Hous on, TX, USA, 23–26 Oc obe 2002; Volume 3,
pp. 1865–1866.
28.
Ma
í
nez, I.; Escayola, J.; Ma
í
nez-Esp onceda, M.; Muñoz, P.; T igo, J.D.; Muñoz, A.; Led, S.; Se ano, L.;
Ga c
í
a, J. Seamless In eg a ion o ISO/IEEE11073 Pe sonal Heal h De ices and ISO/EN13606 Elec onic
Heal h Reco ds in o an End- o-End In e ope able Solu ion. Telemed. E Heal h
2010
,16, 993–1004. [C ossRe ]
29.
Cla ke, M.; de Fol e , J.; Ve ma, V.; Gokalp, H. In e ope able End- o-End Remo e Pa ien Moni o ing Pla o m
based on IEEE 11073 PHD and ZigBee Heal h Ca e P o ile. IEEE T ans. Biomed. Eng. 2017. [C ossRe ]
30.
Ha ey, M.J.; Ha ey, M.G. P i acy and secu i y issues o mobile heal h pla o ms. J. Assn. In . Sci. Technol.
2014,65, 1305–1318. [C ossRe ]
31.
A o a, S.; Y i, J.; Nilsen, W. P i acy and Secu i y in Mobile Heal h (mHeal h) Resea ch. Alcohol Res.
2014
,
36, 143–151.
32.
A ienza, A.A.; Za cadoolas, C.; Vaughon, W.; Hughes, P.; Pa el, V.; Chou, W.-Y.S.; P i s, J. Consume A i udes
and Pe cep ions on mHeal h P i acy and Secu i y: Findings om a Mixed-Me hods S udy. J. Heal h Commun.
2015,20, 673–679. [C ossRe ]
33.
Rubio,
Ó
.J.; T igo, J.D.; Alesanco,
Á
.; Se ano, L.; Ga c
í
a, J. Analysis o ISO/IEEE 11073 buil -in secu i y and
i s po en ial IHE-based ex ensibili y. J. Biomed. In . 2016,60, 270–285. [C ossRe ] [PubMed]
34.
Rahimi, M.R.; Ren, J.; Liu, C.H.; Vasilakos, A.V.; Venka asub amanian, N. Mobile Cloud Compu ing: A Su ey,
S a e o A and Fu u e Di ec ions. Mob. Ne w. Appl. 2014,19, 133–143. [C ossRe ]
35.
Hsieh, J.C.; Hsu, M.W. A cloud compu ing based 12-lead ECG elemedicine se ice. BMC Med. In . Decis.
Mak. 2012,12, 77. [C ossRe ] [PubMed]
36.
Ribei o, L.S.; Viana-Fe ei a, C.; Oli ei a, J.L.; Cos a, C. XDS-I ou sou cing p oxy: Ensu ing con iden iali y
while p ese ing in e ope abili y. IEEE J. Biomed. Heal h In . 2014,18, 1404–1412. [C ossRe ]
37.
Hanen, J.; Kechaou, Z.; Ayed, M.B. An enhanced heal hca e sys em in mobile cloud compu ing en i onmen .
Vie nam. J. Compu . Sci. 2016,3, 267–277. [C ossRe ]
38.
Sys ema ized Nomencla u e o Medicine—Clinical Te ms (SNOMED-CT). A ailable online: h ps://www.
snomed.o g/snomed-c /(accessed on 7 Sep embe 2020).
39.
Uni ied Medical Language Sys em (UMLS). A ailable online: h ps://www.nlm.nih.go / esea ch/umls/
(accessed on 7 Sep embe 2020).
40. Open Au ho iza ion (oAu h). A ailable online: h ps://oau h.ne /(accessed on 7 Sep embe 2020).
41.
Housley, R. Vigil Secu i y CMS. A ailable online: h ps:// ools.ie .o g/h ml/ c5652 (accessed on 7 Sep embe 2020).