O iginal Pape
DenTiUS Plaque, a Web-Based Applica ion o he Quan i ica ion
o Bac e ial Plaque: De elopmen and Usabili y S udy
Nicolás Vila-Blanco1,2, MEng; Vicen e F ei e1, BEng; Ca los Balsa-Cas o2,3, BEng, MSc; Inmaculada Tomás2,3,
DDS, MSc, PhD; Ma ía J Ca ei a1,2, MSc, PhD
1Cen o Singula de In es igación en Tecnoloxías In elixen es, Uni e sidade de San iago de Compos ela, San iago de Compos ela, Spain
2Ins i u o de In es igación Sani a ia de San iago de Compos ela, San iago de Compos ela, Spain
3O al Sciences Resea ch G oup, Depa men o Su ge y and Medical Su gical Speciali ies, School o Medicine and Den is y, Uni e sidade de San iago
de Compos ela, San iago de Compos ela, Spain
Co esponding Au ho :
Ma ía J Ca ei a, MSc, PhD
Cen o Singula de In es igación en Tecnoloxías In elixen es
Uni e sidade de San iago de Compos ela
C/Jena o de la Fuen e, s/n
San iago de Compos ela, E15782
Spain
Phone: 34 881816430
Email: [email p o ec ed]
Abs ac
Backg ound: In he den is y ield, he analysis o den al plaque is i al because i is he main e iological ac o in he 2 mos
p e alen o al diseases: ca ies and pe iodon i is. In mos o he pape s published in he den al li e a u e, he quan i ica ion o
den al plaque is ca ied ou using adi ional, non-au oma ed, and ime-consuming indices. The e o e, he de elopmen o an
au oma ed plaque quan i ica ion ool would be o g ea alue o clinicians and esea che s.
Objec i e: This s udy aimed o de elop a web-based ool called DenTiUS and a ious clinical indices o e alua e den al plaque
le els using image analysis echniques.
Me hods: The ool was execu ed as a web-based applica ion o acili a e i s use by esea che s. Expe use s a e ee o de ine
expe imen s, including images om ei he a single pa ien ( o obse e an indi idual plaque g ow h pa e n) o se e al pa ien s
( o pe o m a g oup cha ac e iza ion) a a pa icula momen o o e ime. A no el app oach o de ec ing isible plaque has been
de eloped as well as a new concep known as non isible plaque. This new e m implies he classi ica ion o he emaining den al
a ea in o 3 sub egions acco ding o he isk o accumula ing plaque in he nea u u e. New me ics ha e also been c ea ed o
desc ibe isible and non isible plaque le els.
Resul s: The sys em gene a es esul s ables o he quan i a i e analysis wi h absolu e a e ages ob ained in each image (indices
abou isible plaque) and ela i e measu emen s (indices abou isible and non isible plaque) ela ing o he e e ence momen .
The clinical indices ha can be calcula ed a e he ollowing: plaque index o an a ea pe in ensi y (API index, a alue be ween 0
and 100), a ea g ow h index (g ow h a e o plaque pe uni o ime in hou s; pe cen age a ea/hou ), and a ea ime index ( he ime
in days needed o achie e a plaque a ea o 100% conce ning he ini ial a ea a he same momen ). Images and g aphics can be
ob ained o a momen om a pa ien in addi ion o a ull epo p esen ing all he p ocessing da a. Den is y expe s e alua ed
he DenTiUS Plaque so wa e h ough a usabili y es , wi h he bes -sco ing ques ions hose ela ed o he wo k low e iciency,
alue o he online help, a ac i eness o he use in e ace, and o e all sa is ac ion.
Conclusions: The DenTiUS Plaque so wa e allows au oma ic, eliable, and epea able quan i ica ion o den al plaque le els,
p o iding in o ma ion abou a ea, in ensi y, and g ow h pa e n. Den is y expe s ecognized ha his so wa e is sui able o
quan i ica ion o den al plaque le els. Consequen ly, i s applica ion in he analysis o plaque e olu ion pa e ns associa ed wi h
di e en o al condi ions, as well as o e alua e he e ec i eness o a ious o al hygiene measu es, can ep esen an imp o emen
in he clinical se ing and he me hodological quali y o esea ch s udies.
(J Med In e ne Res 2020;22(9):e18570) doi: 10.2196/18570
J Med In e ne Res 2020 | ol. 22 | iss. 9 | e18570 | p. 1h ps://www.jmi .o g/2020/9/e18570 (page numbe no o ci a ion pu poses)
Vila-Blanco e alJOURNAL OF MEDICAL INTERNET RESEARCH
XSL
•
FO
Rende X
KEYWORDS
compu e -aided diagnoses; compu e -based biomedical applica ions; den al heal h; den al plaque quan i ica ion; web-based ools;
medical in o ma ics
In oduc ion
Den al plaque is a di e se communi y o mic oo ganisms loca ed
on den al su aces in he o m o a bio ilm embedded in an
ex acellula ma ix o polyme s om bo h he hos and
mic obio a [1,2]. I is known ha plaque is di ec ly ela ed o
he appea ance and p og ession o common o al pa hologies
like ca ies and pe iodon al diseases [3,4]. Moni o ing o how
i de elops is, he e o e, a opic o g ea clinical impo ance
when i comes o es ablishing be e s a egies o he con ol
o o al diseases caused by bac e ial bio ilms [5]. Se e al clinical
indices o quan i ying den al plaque ha e been de eloped o e
ecen decades. These ha e been equen ly used by he esea ch
communi y and in he clinical se ing o e alua e he e icacy
o di e en o al hygiene p oduc s. Some o hese adi ional
plaque indices include hose de eloped by Ram jo d [6], G eene
and Ve million [7], and Quigley and Hein [8] and la e modi ied
by Tu esky e al [9], Löe [10], and O'Lea y [11]. In gene al,
mos o hese con en ional indices employ an o dinal scale as
pa o a simple and semiquan i a i e me hod o e alua e
su aces co e ed by den al plaque. Thei applica ion, howe e ,
has majo limi a ions, gi en he g ea subjec i i y inhe en in
conduc ing isual examina ions. Fu he mo e, he isual me hod
is e y imp ecise when plaque le els a e low o pa icula ly
high, and such clinical in es iga ions a e o en labo ious [12].
The planime ic me hod [13] was an imp o emen and in ol ed
aking a pho og aph o dyed plaque and de e mining i s ex en .
This app oach was much mo e accu a e, as i employed a mo e
objec i e measu e o assess plaque le els (i p oduced a
con inuous, a he han o dinal, ou pu ). Addi ionally, as image
senso s ha e imp o ed yea -on-yea , he quali y o he images
gene a ed is be e han e e . Ne e heless, his is s ill a
ime-consuming p ocess because he ee h and plaque egions
in each image mus be ou lined by hand.
I was only in he 21s cen u y ha expe s began o ely on
echniques ha employ analyses o digi al images o quan i y
den al plaque. The main app oach in ol ed he planime ic
me hod, wi h an imaging ool used o de ec he den al plaque
and oo h a eas indi idually and calcula e he a io be ween
hem. An expe ope a o could ou line bo h egions manually
using a g aphical in e ace [14,15]. Also possible we e
semiau oma ic app oaches, whe eby he image-p ocessing
algo i hm equi ed in e en ion om he den al expe o wo k
[16,17], o images could be segmen ed au oma ically using
image-p ocessing echniques [18-20].
Some esea che s used image-p ocessing so wa e [16,17,21-23]
o gene al-pu pose da a-p ocessing ools [24,25], while o he s
de eloped hei own me hods o p ocess hese images [24,26].
Mo e ecen ly, speci ic den al assessmen so wa e has been
used o quan i y plaque le els [18,27].
The compu e ision echniques used p e iously a y om
udimen a y o ex emely complex. One o he simples me hods
was image h esholding, which made i possible o isola e wo
o mo e di e en a eas acco ding o hei colo o ligh
in ensi ies. This echnique was able o dis inguish be ween
disclosed plaque and nonplaque [24,28]; ee h, plaque, and
gingi a pixels [23]; and isola ed ee h, gums, plaque, and
backg ound a eas [29,30]. Mo e sophis ica ed machine lea ning
algo i hms we e subsequen ly de eloped o enhance he esul s.
Ca e e al [20], o example, c ea ed a da abase o mo e han
600,000 pixels o analyze he ela ionship be ween pixel
in o ma ion (in bo h he RGB and HSI space) and pixel loca ion
(disclosed plaque, oo h, and gingi ae). The in o ma ion in his
la ge able enabled a u he s ep o be aken o c ea e a classi ie
capable o labeling a pixel in a new image in i s mos p obable
loca ion. Clus e ing me hods ha e also been employed. An
example is he app oach adop ed by Kang e al [31], who used
a combina ion o uzzy c-means and cellula neu al ne wo k
algo i hms o classi y image pixels in o plaque, oo h su aces,
and backg ounds. A mean-shi based clus e ing algo i hm was
also used o plaque segmen a ion [19].
The di e en me hods a ailable oday p oduce p omising esul s,
wi h many s udies epo ing he sui abili y o au oma ic o
semiau oma ic echniques o assessing den al plaque and, as a
esul , o al hygiene le els [16-18,24,27,29,32]. Speci ically,
quan i a i e ligh ‐induced luo escence digi al (QLF-D) is an
adap a ion o QLF ha employs a modi ied il e se (D007;
Inspek o Resea ch Sys ems BV, Ams e dam, The Ne he lands),
na ow‐band iole ligh (405 nm), and a high‐speci ica ion
digi al single‐lens e lex came a. This con igu a ion has been
speci ically de eloped o enhance he isualiza ion and
quan i ica ion o den al plaque [33-35].
Following his end o digi al de elopmen [36], we p esen
DenTiUS Plaque, which is a ool we ha e de eloped o enable
he au oma ic assessmen o den al plaque le els. DenTiUS
Plaque was i s en isaged as a s andalone applica ion. Howe e ,
as we wan ed o pe mi web access o expe use s, a mo e
gene al pla o m (DenTiUS Lab) was designed o sepa a e he
p ocessing s ages (DenTiUS Plaque p ope module) and enable
he execu ion o adminis a i e asks like use and pa ien
managemen . The sys em also has a common in e ace whe e
use s can log in, include pa ien s in he da abase, and in e ac
wi h he DenTiUS Plaque module independen ly. The en i e
pla o m is web-based, so i can un on any web b owse . I also
has a iendly ope a o in e ace o non echnical use s.
The es o he pape is s uc u ed as ollows. The Me hods
sec ion i s desc ibes he en i e pla o m, including i s main
ea u es and so wa e a chi ec u e, and also con ains a b ie
desc ip ion o he pa ien da abase. Then, i explains DenTiUS
Plaque, desc ibing he kinds o images i can p ocess, he
expe imen 's design, and he p ocessing algo i hms. In he
Resul s sec ion, he di e en pa ame e s o he quan i a i e
analysis o den al plaque o e ed by he ool a e p esen ed, as
well as se e al g aphics, h ough a eal case. Subsequen ly, he
esul s o a ques ionnai e o indi iduals wo king in he ield o
den is y on he usabili y o so wa e a e p esen ed o e alua e
J Med In e ne Res 2020 | ol. 22 | iss. 9 | e18570 | p. 2h ps://www.jmi .o g/2020/9/e18570 (page numbe no o ci a ion pu poses)
Vila-Blanco e alJOURNAL OF MEDICAL INTERNET RESEARCH
XSL
•
FO
Rende X
he ease o use o he ins umen and i s use ulness. Finally, he
pape closes wi h a discussion and some closing ema ks and
iden i ies possible u u e imp o emen s.
Me hods
DenTiUS S uc u e
The DenTiUS pla o m eme ged na u ally as a way o manage
use s, images, and pa ien s in den is y esea ch. I was ini ially
de eloped as a s andalone applica ion o he quan i ica ion o
den al plaque, wi h modules o managing he pa ien s and
expe imen s o a unique use , bu no he di e en p o iles and
in e ac ions o mul iple use s. As ou esea ch g oup is also
wo king in o he den is y ields, all o which employ a sha ed
pa ien and use da abase, he decision was made o de elop an
en i e web-based sys em, DenTiUS Lab, o den al assessmen
expe imen s. DenTiUS Lab in eg a es DenTiUS Plaque's
s uc u e in o a modula pla o m ha isola es he use in e ace
and he pa ien and use da abase om he main DenTiUS
Plaque module. Figu e 1 po ays a gene al block diag am o
he comple e pla o m, which was designed as a con en
managemen sys em. This allows clinicians and esea che s o
sign up, egis e , manage pa ien s, and in e ac wi h DenTiUS
Plaque by uploading plaque images and designing and
p ocessing expe imen s.
Figu e 1. Block diag am o DenTiUS Plaque inside he DenTiUS global pla o m. The modula design acili a es he easy connec ion o he use and
pa ien da abase o he DenTiUS Plaque ool.
DenTiUS was designed as a global pla o m o suppo clinical
den al esea ch h ough he inclusion o di e en modules ha
enable use s o ob ain a se o measu es, g aphs, and images
wi h which o easily each conclusions abou hei expe imen s.
I was de eloped as a web applica ion o se e al easons: (1)
mos pe sonal compu e s come wi h a p e-ins alled web b owse ,
meaning ha an ins alla ion p ocess ha could be complica ed
o nonspecialis use s is a oided; (2) pla o m upda es
(including in e ace imp o emen s, bug ixes, new ea u es, and
new modules) a e deployed in a p ocess ha is comple ely
anspa en , bu also unin usi e, which also ensu es ha all
use s a e employing he mos up- o-da e applica ion, emo ing
any equi emen o suppo old e sions o manage compa ibili y
issues be ween hem; and (3) as ha ing a web b owse is he
applica ion's only equi emen , i can be used on bo h compu e s
and mobile de ices.
The pla o m was implemen ed as a clien -se e applica ion,
which uns on a cen al se e whe e all he da a a e s o ed
(Figu e 2). All he p ocessing asks a e ca ied ou on he web
se e , so he clien does no need a high-pe o mance de ice.
The applica ion logic inside he se e is o ganized in o
con olle s ( o p ocess use eques s), se ices ( o pe o m
ope a ions), and models ( o manage da abase ope a ions), while
he well-known Model-View-Con olle pa e n is used o
manage en i ies, da a access, dependency injec ions, and many
o he elemen s o he applica ion.
The Ja a p og amming language was employed when
de eloping he pla o m because o i s popula i y, powe ,
e bosi y, and ease o main enance [37]. The
Model-View-Con olle pa e n, meanwhile, was cons uc ed
wi h he Sp ing amewo k [38], and da a s o age was managed
wi h bo h Hibe na e [39], which is an objec - ela ional mapping
lib a y, and a Pos g eSQL da abase [40]. Image p ocessing
algo i hms we e execu ed in OpenCV [41], speci ically i s Ja a
implemen a ion [42]. As den al images a e usually e y la ge,
a limi o 5 Mb was se in ela ion o he algo i hm esul s o
ensu e accu acy, wi h he applica ion esizing hem
au oma ically i equi ed.
In summa y, DenTiUS was de eloped wi h main ainabili y and
ex ensibili y as he main objec i es. In his way, he applica ion
was implemen ed by ollowing a modula s uc u e, whe e he
isual, logic, and da a laye s we e isola ed. The unc ionali y
in each laye was di ided in o classes, wi h many abs ac ions
a ailable o imp o e he ex ensibili y. Consequen ly, he only
J Med In e ne Res 2020 | ol. 22 | iss. 9 | e18570 | p. 3h ps://www.jmi .o g/2020/9/e18570 (page numbe no o ci a ion pu poses)
Vila-Blanco e alJOURNAL OF MEDICAL INTERNET RESEARCH
XSL
•
FO
Rende X
elemen s equi ing implemen a ion (i necessa y) o de elop a
new module ela ed o den al esea ch a e he da a de ini ion
(eg, images, da ase s), desc ip ion o a new kind o expe imen
and i s p ocessing algo i hms, and a cus omized epo o ma
(ie, ano he module wi h he same s uc u e as DenTiUS Plaque,
as seen in Figu e 1).
Figu e 2. S uc u al diag am o he DenTiUS pla o m, which is di ided in o a clien side, web se e , and da abase se e . GUI: g aphical use in e ace.
The DenTiUS Lab inpu sc een (Figu e 3A) asks he use o
en e he applica ion o egis e . I a use is new o he sys em,
egis a ion is a e y s aigh o wa d p ocess and akes only
seconds o answe basic ques ions such as name, a ilia ion, and
le el o expe ise. Once logged in, he sys em p esen s he
s a is ics o use, namely inished p ocessings, expe imen s, and
pa ien s (Figu e 3B).
E e y sec ion o DenTiUS Lab is a ailable om bo h he menu
ba and side menu ( isible when enough space is a ailable).
Access o o he u ili ies is also possible h ough he common
in e ace ba a he op o he sc een (see Figu e 3). These
include a “Home” sec ion, which is a ailable by clicking on he
DenTiUS Lab icon and is whe e he ope a o 's s a is ics o use
a e displayed; “Help” page ha explains he pla o m's ea u es,
de ails o he p ocessing algo i hms, and examples o use;
“Downloads” sec ion, whe e sample da a can be downloaded
o use in he applica ion; “Use ” panel (accessible by clicking
on he use name), whe e use s can change hei p o ile da a
and passwo d; and “Language” sec ion (English o Spanish).
An “Admin” sec ion is also a ailable o use s wi h
adminis a i e p i ileges.
A “Pa ien ” da abase in e ace (Figu e 3C) allows use s o
egis e and manage hei pa ien s, who can be il e ed by pa ien
numbe , clinical eco d numbe , pa ien ini ials, and sex. The
use can also c ea e new pa ien de ails by comple ing a o m
wi h da a such as he pa ien 's clinical eco d numbe , bi h da e,
ace, sex, ci y, and coun y (Figu e 3D). The name and su name
ields a e op ional o p ese e pa ien anonymi y.
J Med In e ne Res 2020 | ol. 22 | iss. 9 | e18570 | p. 4h ps://www.jmi .o g/2020/9/e18570 (page numbe no o ci a ion pu poses)
Vila-Blanco e alJOURNAL OF MEDICAL INTERNET RESEARCH
XSL
•
FO
Rende X
Figu e 3. Common modules o he DenTiUS s uc u e ela ing o use s: (A) login sc een, (B) use s a is ics and pa ien s, (C) pa ien da abase, (D) new
pa ien de ini ion.
DenTiUS Plaque
Pa ien Plaque Images
DenTiUS Plaque was he main objec i e behind ou
de elopmen o he DenTiUS pla o m. The goal was o enable
DenTiUS Plaque o pe o m as an au oma ed decision suppo
sys em o help expe s analyze and quan i y he mac oscopically
obse able den al plaque deposi ed on ee h. In he i s s ep,
he so wa e equi es ep oducible images o he
luo escein-dyed plaque aken unde ul a iole ligh .
Fluo escein is a well-known luo och ome in he ield o
den is y, and a pa en o i s use as a den al plaque ma ke was
iled in he USA in 1967 by He be B illian (U.S. Pa en
3-309-274; 1967). In hese images, he b igh blue egion
co esponds o he oo h a ea una ec ed by plaque, while he
J Med In e ne Res 2020 | ol. 22 | iss. 9 | e18570 | p. 5h ps://www.jmi .o g/2020/9/e18570 (page numbe no o ci a ion pu poses)
Vila-Blanco e alJOURNAL OF MEDICAL INTERNET RESEARCH
XSL
•
FO
Rende X
g een egion ma ches plaque deposi s o e bo h he oo h
su aces and gingi a.
To enable he assessmen o den al plaque le els a di e en
imes, an en i y called a “plaque momen ” was de ined o a
single plaque image (usually he on al iew) o a se o images
( on and la e al iews) cap u ed om a pa ien a a pa icula
ime. The possibili y o including la e al iews is a no el y o
he sys em, as hese p o ide a be e iew o he pos e io ee h,
whe e mo e plaque is usually accumula ed.
Plaque momen s can be a ached o a pa ien 's ile ia he
Pa ien s sec ion (Figu e 3C), which con ains in o ma ion abou
each pa ien conce ning he numbe o image se s included o
he o him and he numbe o plaque images dis ibu ed h ough
all hei momen s. Ano he way o a ach new plaque momen s
o a pa ien 's eco d is by en e ing he Plaque sec ion di ec ly
om he “MyDenTiUS” menu, whe e all he plaque momen s
a e lis ed independen ly and can be modi ied o dele ed. In any
case, he use can modi y, dele e, o include a new momen
whe e he images (a leas one) ha e a da e, ime, b ie
desc ip ion, and op ional came a se -up pa ame e s. A his
poin , i is c ucial o de ine he so-called “ e e ence momen ”
o “momen 0.” This canno be dele ed, as i is used as a
e e ence o compu e he plaque-le el g ow h o e ime and
usually co esponds o ci cums ances whe e he e has been
“pe ec den al cleaning.”
Plaque Expe imen Design
The plaque expe imen design module (see Figu e 1) enables
use s o de elop new expe imen s by selec ing speci ic pa ien
da a and uning he pa ame e s o he image p ocessing
algo i hms. The expe imen s a e p esen ed in a lis iew, wi h
se e al sho cu s on he igh side o he sc een ha igge
ope a ions like duplica ing, p ocessing, dele ing, o c opping
he associa ed images. When he use c ea es a new expe imen ,
a sc een appea s wi h he ollowing abs: New, Selec , Cu
Images, and P ocess (Figu e 4). Figu e 4A shows he “New”
ab, on which he name and desc ip ion ields a e manda o y.
On he “Selec ” ab, he use mus choose he plaque momen s
o be p ocessed by selec ing hem om he pa ien da abase.
The sys em au oma ically includes he e e ence momen ela ed
o any selec ed momen . The “Cu Images” ab allows he use
o op ionally selec pa o he images o dele e a i ac s. Finally,
he “P ocess” ab (Figu e 4B) shows a summa y o he
“Expe imen da a” and, below ha , he “P ocessing da a”
pa ame e s, wi h a manda o y name and desc ip ion. This
pe mi s use s o y di e en con igu a ions o he p ocessing
pa ame e s o he same se o images. In his sc een, he expe
use can employ he de aul pa ame e s o change hese o
di e en alues. The meanings o hese pa ame e s will be
explained la e in his sec ion.
The con igu a ion o he expe imen is decided by he use
acco ding o he speci ic ci cums ances (age g oups, gende ,
single o mul iple pa ien s) o objec i es (eg, e ec i eness o
mou hwash, e ec i eness o b ushing de ices, smoking e ec s).
The sys em pe o ms he expe imen s de ined by he use by
p ocessing he da a o all he pa ien s and all he pa ien momen s
included in he expe imen . When he expe imen in ol es
mul iple momen s o he same pa ien , he images wi h he
selec ed oo h a eas a e subjec ed o a p ocess o no maliza ion
in he numbe o pixels.
DenTiUS Plaque does no pe mi compa isons be ween g oups,
as his was no he objec i e o he sys em, bu use s can expo
all he compu ed pa ame e s o manage and p epa e hei own
s a is ics wi h no limi a ions.
J Med In e ne Res 2020 | ol. 22 | iss. 9 | e18570 | p. 6h ps://www.jmi .o g/2020/9/e18570 (page numbe no o ci a ion pu poses)
Vila-Blanco e alJOURNAL OF MEDICAL INTERNET RESEARCH
XSL
•
FO
Rende X
Figu e 4. Plaque expe imen design o he de ini ion o a new expe imen : (A) New ab and (B) P ocess ab, wi h a summa y o he expe imen and
p ocessing da a showing he p ocessing pa ame e s ha will be used when pushing he "P ocess" bu on a he bo om.
P ocessing Algo i hms
A e an exhaus i e analysis o he colo cha ac e is ics o each
egion o in e es , a p ocess is de ined o segmen he den al
a ea, excluding he gingi a egion: Ini ially, he blue and ed
channels a e added oge he , and he esul is h esholded
acco ding o a pa ame e ixed by he use (by de aul , he
algo i hm chooses 75% o he da kes pixels as he backg ound
and 25% o he ligh es as he o eg ound). The isola ed pixels
a e hen emo ed, and a connec ed componen analysis is
pe o med o iden i y objec s om ha bina y image. The ea e ,
a se o condi ions ela ed o egion size (>2000 pixels, by
de aul ) and solidi y (>0.5, by de aul ) a e applied o emo e
nonden al egions. O he speci ic ules, which we e de ined o
elimina e b igh a i ac s, a e included o make he p ocess ully
J Med In e ne Res 2020 | ol. 22 | iss. 9 | e18570 | p. 7h ps://www.jmi .o g/2020/9/e18570 (page numbe no o ci a ion pu poses)
Vila-Blanco e alJOURNAL OF MEDICAL INTERNET RESEARCH
XSL
•
FO
Rende X
au oma ic. All hese pa ame e s can be changed in he “P ocess”
ab (see he “P ocessing da a” pa ame e s in Figu e 4B).
Once he den al a ea is segmen ed, he sys em mus pe o m an
analysis o i s alues o ex ac and measu e he den al plaque.
As men ioned p e iously, plaque is cha ac e ized by a high
g een in ensi y, whe eas he es o he den al egion has a high
blue concen a ion. The i s s ep is hus he de ec ion o he
( isible) plaque by analyzing he di e ences be ween he g een
and blue channels (Equa ion 1). The es o he den al egion is
conside ed o be a i s app oxima ion o he non isible plaque
a ea (Equa ion 2).
P isible = |G-B| o pixels wi h [G-B > 0] (1)
Pnon isible = |G-B| o pixels wi h [G-B ≤ 0 and G>0
and B>0] (2)
whe e P isible is he isible plaque, Pnon isible is he ini ial
app oxima ion o he non isible plaque, and Gand Ba e he
g een and blue channels o he den al a ea, espec i ely.
G-B den al a ea his og ams a e included o assess how le els
o den al plaque inc ease o e ime when o al hygiene s ops,
wi h he posi i e and nega i e sides co esponding o he isible
(G>B) and i s app oxima ion o he non isible plaque (G<B),
espec i ely. Figu e 5A shows an example o a 96-hou
expe imen , whe e momen s we e eco ded e e y 24 hou s,
while Figu e 6 p esen s he esul s o he di e en s eps o he
algo i hm on a on al image o a pa ien a e 96 hou s o pe ec
cleaning. A e p o essional den al cleaning (momen M0), mos
o he his og am a ea co esponded o he i s app oxima ion
o he non isible plaque a ea (Equa ion 2), as i s alues we e
highly concen a ed in he nega i e pa o he g aph. In he
momen s ha ollowed (M1 o M4), he his og am was la ened
and displaced owa d he posi i e side, hus inc easing he
isible plaque le el. In he inal momen (M4), 4 days a e a
den al cleaning, he la ges pa o he his og am was placed on
he posi i e side o he g aph, wi h he isible plaque a ea
co e ing mos o he den al egion. The main inding behind
his s ep was he disco e y and measu emen o he ansi ion
a ea be ween he isible plaque and he es o he den al egion,
based on he isk o plaque de eloping in he nea u u e. As a
esul , he sys em pe o ms an ex a segmen a ion s ep be o e
displaying he non isible plaque a ea (as seen in Figu e 6F).
Figu e 5. G-B his og am whe e he e ical line ep esen s G-B=0 and G and B a e he g een and blue channels, espec i ely. (A) G-B his og ams o
momen s M0 o M4, showing he p og ession o he his og am alues owa d he G-B posi i e alues ( isible plaque); (B) De ini ion o l and δ o M4,
showing plaque (g een), nonplaque (blue), le el 1 non isible plaque ( ed, de ined by δ), and le el 2 non isible plaque (o ange, de ined by l). The
bounda y be ween nonplaque and le el 2 non isible plaque (l) is easily obse able in his cha , as i co esponds o he in ensi y alue whe e he 2
his og ams c oss.
Acco ding o 2 di e en pa ame e s, δand l, he blue/g een
di e ences a e h esholded o p oduce 3 di e en egion masks
in he nega i e pa o he g aph, namely le el 1 non isible
plaque ( isk o being plaque in he ollowing hou s), le el 2
non isible plaque ( isk o being plaque in he ollowing days),
and no plaque (no isk o being plaque in he medium e m):
Pnon isible(le el1) = |G-B| o pixels wi h [G-B ≤0 and
G-B ≥δ] (3)
Pnon isible(le el2) = |G-B| o pixels wi h [G-B <δ and
G-B ≥l] (4)
NonPlaque = |G-B| o pixels wi h [G-B<l] (5)
whe e he lpa ame e is au oma ically calcula ed om
non isible plaque his og ams o cu en and e e ence images
(Figu e 5B); i is de ined as he i s alue whe e he numbe
o non isible plaque pixels in he cu en image is g ea e han
he numbe o non isible plaque pixels in he e e ence image
(momen 0).
The δpa ame e can be se by he use o is au oma ically
esol ed as he absolu e di e ence be ween he a e age pixel
alues o he blue and g een channels in he den al a ea. The
modi ica ion o his pa ame e by he use in he “P ocess” ab
(see Figu e 4B) indica es an inc ease o dec ease in he le el 1
non isible plaque egion ( ed a ea in Figu e 6F). This means
J Med In e ne Res 2020 | ol. 22 | iss. 9 | e18570 | p. 8h ps://www.jmi .o g/2020/9/e18570 (page numbe no o ci a ion pu poses)
Vila-Blanco e alJOURNAL OF MEDICAL INTERNET RESEARCH
XSL
•
FO
Rende X
ha he pa e n o plaque de elopmen can be simula ed by s a ing wi h a minimum alue and p og essi ely inc easing i .
Figu e 6. Resul s o he di e en s eps o he algo i hm ela ing o he on al image o a pa ien a e 96 hou s o pe ec cleaning: (A) on al ul a iole
image, (B) segmen ed den al a ea, (C) isible plaque mask, (D) non isible plaque mask, (E) nonplaque le el 1 mask, (F) inal labeled image (blue:
nonplaque; g een: isible plaque; ed: non isible plaque le el 1; o ange: non isible plaque le el 2).
E hical App o al
Images om Spanish Caucasian subjec s a e pa o a s udy
p o ocol app o ed by he Galician Clinical Resea ch E hics
Commi ee ( egis a ion numbe 2014/008). The image
collec ion was pe o med ollowing he e hical s anda ds o ou
ins i u ion's esea ch commi ee and he 1964 Decla a ion o
Helsinki and i s la e amendmen s [43].
RESULTS
Plaque P ocessing Resul s
Figu e 7 po ays he comple e p ocess o a single pa ien image
se belonging o a pa icula expe imen : Plaque images we e
uploaded o he da abase, he plaque expe imen was designed,
and he expe imen p ocessing was launched.
When he p ocessing inishes, he use can iew he esul s by
clicking on he app op ia e no i ica ion o he “P ocessings”
submenu. The p ocessing esul s a e di ided in o 4 abbed
subsec ions (Figu e 8): “Da a and p ope ies,” “Single
measu emen s,” “Measu emen s download,” and “Images and
cha s.” The “Da a and p ope ies” ab p esen s a summa y o
s a is ics conce ning he p ocessing as well as he p ocessing
pa ame e s. Figu e 8B shows he “Single measu emen s” ab,
which p esen s he esul s o he quan i a i e analysis o each
image in able o m, wi h absolu e a e ages ob ained in each
image (indices abou isible plaque) and ela i e measu emen s
(indices abou isible and non isible plaque) ela ing o he
e e ence momen . As shown in his igu e and jus below he
abs wi h an in o ma ion ale , he use mus click on each image
o display hei p ocessing esul s in he “Single measu emen s”
and “Images and cha s” abs.
The clinical indices de ined in his esea ch p o ide in o ma ion
on he bac e ial plaque on den al su aces, speci ically he a ea
whe e i is p esen , i s in ensi y, a combined alue be ween he
a ea and in ensi y, and di e en g ow h pa ame e s. The sys em
can p oduce den al plaque clinical indices in a compu e ized
o m, including in ela ion o he plaque index o an a ea pe
in ensi y (API index, a alue be ween 0-100), a ea g ow h index
(g ow h a e o plaque pe uni o ime in hou s: pe cen age
a ea/hou ), and a ea ime index ( he ime, in days, needed o
achie e a plaque a ea o 100% conce ning he ini ial a ea a he
same momen ).
In he example in Figu e 8, a e 96 hou s o plaque
accumula ion, ocusing on he isible plaque alues, he pa ien
p esen ed an API index o 21.18 and a hypo he ical plaque
g ow h a e o 0.75% pe hou . A pa icula ly in e es ing
pa ame e supplying he labeled image is he a ea ime. In his
case, he a ea- ime ela i e alue is 1.67 o isible plaque and
0.61 o o al plaque (see Figu e 8B). In o he wo ds, om a
heo e ical poin o iew and assuming a cons an g ow h pa e n
o ha momen , he pa ien would ake 1.67 days o achie e a
isible plaque le el o 100%. This alue is ob iously lowe ,
a ound 0.61 days, when i comes o ealizing plaque le els o
100% o bo h non isible and isible plaque.
The measu emen ables can be cus omized and downloaded in
CSV and sp eadshee o ma s ia he “Measu emen s download”
ab, whe e use s can also selec speci ic pa ame e s wi hin he
absolu e and ela i e measu emen s.
J Med In e ne Res 2020 | ol. 22 | iss. 9 | e18570 | p. 9h ps://www.jmi .o g/2020/9/e18570 (page numbe no o ci a ion pu poses)
Vila-Blanco e alJOURNAL OF MEDICAL INTERNET RESEARCH
XSL
•
FO
Rende X
dis ibu ion, and ep oduc ion in any medium, p o ided he o iginal wo k, i s published in he Jou nal o Medical In e ne
Resea ch, is p ope ly ci ed. The comple e bibliog aphic in o ma ion, a link o he o iginal publica ion on h p://www.jmi .o g/,
as well as his copy igh and license in o ma ion mus be included.
J Med In e ne Res 2020 | ol. 22 | iss. 9 | e18570 | p. 16h ps://www.jmi .o g/2020/9/e18570 (page numbe no o ci a ion pu poses)
Vila-Blanco e alJOURNAL OF MEDICAL INTERNET RESEARCH
XSL
•
FO
Rende X
