IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS, VOL. 29, NO. 7, JULY 2025 5191
Smoking De ec ion and Cessa ion: An Upda ed
Scoping Re iew o Digi al and Mobile
Heal h Technologies
Mi ko Casu , F ancesco Gua ne a , Giusy Ri a Ma ia La Rosa ,
Sebas iano Ba ia o , Senio Membe , IEEE, Pasquale Caponne o , Ricca do Polosa ,
and Rosalia Emma
Abs ac —Digi al and mobile heal h echnologies o e
p omising solu ions o smoking de ec ion and cessa ion.
This scoping e iew examines he cu en s a e o esea ch
and de elopmen in his ield, encompassing sma phone
applica ions, wea able de ices, and senso -based sys ems.
We analyzed 49 s udies published be ween 2019 and 2023
om PubMed and ACM Digi al Lib a y, ocusing on echnol-
ogy ea u es, ou comes, and e alua ion me hods. Wea able
senso s and sma phone apps show po en ial in comba -
ing smoking addic ion and imp o ing qui a es. Mo ion
senso s o hand- o-mou h ges u e de ec ion achie e high
accu acy in con olled se ings bu ace challenges in eal-
wo ld applica ions. Machine lea ning models and wi eless
signal de ec ion echniques yield encou aging esul s bu
Recei ed 17 Sep embe 2024; e ised 23 Decembe 2024 and 27
Janua y 2025; accep ed 3 Ma ch 2025. Da e o publica ion 10 Ma ch
2025; da e o cu en e sion 4 July 2025. This esea ch did no ecei e
any speci ic g an om unding agencies in he public, comme cial, o
no - o -p o i sec o s. The wo k o Ricca do Polosa (as ask leade ),
Mi ko Casu, Sebas iano Ba ia o and Rosalia Emma has been sup-
po ed by Minis y o Uni e si y and Resea ch (MUR) in he amewo k o
he Na ional Reco e y and Resilience Plan (PNRR)E63C22000900006,
SPOKE 1, Wo k Page Heal h unde p ojec SAMOTHRACE (Sicilian
Mic o and Nano Technology Resea ch and Inno a ion Cen e ). The wo k
o F ancesco Gua ne a has been suppo ed by MUR in he amewo k o
PNRR PE0000013, unde p ojec “Fu u e A i icial In elligence Resea ch
–FAIR”.(Co esponding au ho : Mi ko Casu.)
Mi ko Casu is wi h he Depa men o Ma hema ics and Compu e
Science, Uni e si y o Ca ania, 95123 Ca ania, I aly, and also wi h he
Depa men o Educa ional Sciences, Sec ion o Psychology, Uni e si y
o Ca ania, 95123 Ca ania, I aly (e-mail: mi ko[email p o ec ed]).
F ancesco Gua ne a is wi h he Depa men o Ma hema ics and
Compu e Science, Uni e si y o Ca ania, 95123 Ca ania, I aly (e-mail:
ancesco.gua [email p o ec ed]).
Giusy Ri a Ma ia La Rosa is wi h he Depa men o Clinical & Expe -
imen al Medicine, Uni e si y o Ca ania, 95123 Ca ania, I aly (e-mail:
giusy[email p o ec ed]).
Sebas iano Ba ia o is wi h he Depa men o Ma hema ics and Com-
pu e Science, Uni e si y o Ca ania, 95123 Ca ania, I aly, and also wi h
he Cen e o Excellence o he Accele a ion o Ha m Reduc ion, Uni-
e si y o Ca ania, 95123 Ca ania, I aly (e-mail: [email p o ec ed]).
Pasquale Caponne o is wi h he Depa men o Educa ional Sci-
ences, Sec ion o Psychology, Uni e si y o Ca ania, 95123 Ca ania,
I aly, and also wi h he Cen e o Excellence o he Accele a ion o
Ha m Reduc ion, Uni e si y o Ca ania, 95123 Ca ania, I aly (e-mail:
p[email p o ec ed]).
Ricca do Polosa and Rosalia Emma a e wi h he Depa men o Clin-
ical & Expe imen al Medicine, Uni e si y o Ca ania, 95123 Ca ania,
I aly, and also wi h he Cen e o Excellence o he Accele a ion o
Ha m Reduc ion, Uni e si y o Ca ania, 95123 Ca ania, I aly (e-mail:
[email p o ec ed]; [email p o ec ed]).
Digi al Objec Iden i ie 10.1109/JBHI.2025.3549255
equi e u he e inemen . Sma phone apps p o ide pe -
sonalized plans and p og ess acking, hough mos ely
on manual logging and lack igo ous scien i ic e alua ion.
Ou indings sugges ha digi al heal h echnologies could
signi ican ly enhance smoking cessa ion e o s. Howe e ,
mo e obus e alua ion me hods and in eg a ion o senso
da a wi h machine lea ning a e needed o imp o e usabil-
i y and e ec i eness. Con inued esea ch and inno a ion
in his ield a e c ucial o de eloping eliable, p ac ical
solu ions and in eg a ing hese echnologies in o clinical
p og ams.
Index Te ms—Smoking de ec ion, heal h echnologies,
smoking cessa ion, medical mobile apps, echnology
e iew, wea able de ices.
I. INTRODUCTION
TOBACCO smoking emains a leading cause o p e en able
illness and p ema u e dea h wo ldwide, despi e declining
p e alence a es [1],[2]. In he U.K., smoking- ela ed dea hs
accoun ed o 16% o all dea hs in 2016 [1]. The economic
impac o smoking is subs an ial, wi h global annual cos s ex-
ceeding US$500 billion [3]. Smoking beha io is main ained by
nico ine’s ein o cing p ope ies and he dis an na u e o heal h
consequences [2]. E ec i e in e en ions o educe smoking
p e alence include ax inc eases, social ma ke ing, and b ie ad-
ice omheal hp o essionals[2].Wo kplacesmokingcessa ion
p og amsha eshowncos -e ec i eness,wi hbene i -cos a ios
up o 8.75 and signi ican employe cos sa ings [3]. While a i-
ous cessa ion measu es ha e p o en e ec i e and cos -e ec i e,
challenges emain in add essing pe sis en inequali ies in smok-
ing a es among ce ain g oups, such as manual wo ke s and
indi iduals wi h se ious men al illness [1].
O e he pas decade, we ha e wi nessed a apid p oli e a-
ion o po able de ices ha ha e become cen al o ou daily
li es [4],[5]. No ably, sma phone echnology, coupled wi h
e e -expanding bandwid h connec i i y and he g ow h o social
ne wo ks, has undamen ally ans o med he way we conduc
nea ly all ou daily ac i i ies, ushe ing in an e a o pe asi e
digi al echnology [4],[6],[7],[8]. In addi ion o sma phones,
he e has been a signi ican up ick in he adop ion o a ious
wea able de ices and home/o ice ins alla ions [9],[10],[11],
[12], all in e connec ed and con ollable h ough simple sma -
phone applica ions. This in e connec ed de ice ecosys em is
© 2025 The Au ho s. This wo k is licensed unde a C ea i e Commons A ibu ion 4.0 License. Fo mo e in o ma ion, see
h ps://c ea i ecommons.o g/licenses/by/4.0/
5192 IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS, VOL. 29, NO. 7, JULY 2025
gea ed owa ds enhancing he in elligence o ou de ices and
en i onmen s, leading o he eme gence o concep s like sma
homes and sma o ices [13],[14],[15],[16]. C ucially, hese
wea able and emo e de ices a e equipped wi h speci ic senso s
ha can cap u e da a ela ed o indi iduals o hei su ound-
ings, which can hen be sha ed and p ocessed collabo a i ely
among di e en de ices [17],[18],[19]. The goal is o de i e
insigh s and c ea e added alue o he use expe ience, o e ing
oppo uni ies o bo h da a cap u e and use suppo . A no el
and inno a i e applica ion o sma de ice echnology lies in i s
po en ial o assis wi h smoking cessa ion ea men s [20],[21],
[22]. In he ealm o smoking de ec ion and cessa ion echnolo-
gies, he e is a ecognized issue ha hese echnologies a e no
ully op imized o eal-li e scena ios. While exis ing echnolo-
gies ha e demons a ed po en ial, hei pe o mance in eal-li e
scena ios con inues o pose a challenge. The p esen wo k is a
scoping e iew upda ing a p e iously published wo k [17] wi h
he aim o conduc ing a compa a i e examina ion o a ious
sma phoneapplica ions(apps),wea able echnologiesdesigned
o au oma ic smoking de ec ion, and o he ins ances whe e
echnology can play a ole in suppo ing smoking cessa ion
in e en ions.
In his pape , we aim o p o ide an o e iew and analysis o
he cu en s a e-o - he-a echnology ocusing on au oma ed
smoking de ec ion and smoking cessa ion echnologies. An
au oma ic smoking de ec ion echnology is a solu ion designed
o asce ain he numbe o ciga e es smoked by an indi idual
wi hin a speci ied obse a ion pe iod [16],[23]. This encom-
passes app oaches ha necessi a e minimal use in e en ion
(i.e., au oma ic), encompassing all s ages in ol ed in de ec ing
smoking e en s, om collec ing senso y da a o making he inal
in e ence, as opposed o solu ions elian on sel - epo ing by
pa icipan s (e.g., dia y apps). In he ollowing sec ions, he
mos ele an apps and echnologies designed o help people
s op smoking a e shown and compa ed. A summa y o he
e ised solu ions o help use s qui smoking is p esen ed in he
Discussion.
II. METHODS
A. Resea ch Ques ion
This scoping e iew was conduc ed ollowing he P e-
e ed Repo ing I ems o Sys ema ic Re iews and Me a-
Analyses (PRISMA) Ex ension o Scoping Re iews (Suppl.
Ma e .1) [24]. The aim was o syn hesize and explo e he cu en
applica ions o digi al de ices o au oma ically de ec ing he
use o ciga e es. Addi ionally, his e iew will also epo on
he use o smoking cessa ion sma phone applica ions, which
ep esen a signi ican s ide in le e aging echnology o aid in
smoking cessa ion. Ou s udy, s uc u ed ollowing he PICO
o ma [25], ocused on indi iduals who smoke (P), examining
he applica ion o digi al heal h echnologies, such as sma -
phone apps, wea able de ices, senso s, and machine lea ning
echniques, o de ec smoking e en s and suppo smoking
cessa ion (I). These inno a i e app oaches we e compa ed o
adi ional me hods, including manual sel -moni o ing, s an-
da dbeha io al he apies,o heabsenceo in e en ion(C).The
ou comes o in e es included imp o emen s in he au oma ic
de ec ion o smoking e en s, such as ecall a es and accu acy,
inc eased smoking cessa ion a es, usabili y and accep ance o
hese echnologies, and hei success ul in eg a ion in o clinical
p ac ice (O).
B. Sys ema ic Sea ch o Pa en s
A sys ema ic sea ch was conduc ed on Google Pa en s o
iden i y pa en s ela ed o smoking cessa ion sys ems and ech-
nologies. Google Pa en s was chosen as he sea ch engine
due o i s comp ehensi e co e age o pa en da abases om
mul iple ju isdic ions, including he Uni ed S a es Pa en and
T adema k O ice (USPTO), Eu opean Pa en O ice (EPO), and
Wo ld In ellec ual P ope y O ganiza ion (WIPO). The sea ch
s a egy was designed o be b oad o cap u e as many ele an
pa en s as possible. The sea ch e ms used we e combina ions
o he ollowing keywo ds: (“smoking cessa ion sys em” OR
“au oma ed smoking de ec ion”) AND (“pa en ” OR “appli-
ca ion” OR “me hod”). The sea ch, un es ic ed by da e o
ju isdic ion, sc eened all esul s o ele ance based on i le
and abs ac . Pa en s de ailing smoking cessa ion sys ems o
echnologies we e u he analyzed. Addi ional ele an pa en s
we e iden i ied h ough sc eening he e e ence lis s o hese
pa en s.Da a ex ac ed omeachpa en , including i le,numbe ,
iling and publica ion da es, in en o s, assignees, abs ac , and
claims, p o ided an o e iew o he la es echnology in au o-
ma ed smoking de ec ion and cessa ion sys ems. The sys ema ic
sea ch esul s we e inco po a ed in o he PRISMA low diag am
(Fig. 1), isually ep esen ing he sea ch and selec ion p ocess
o anspa ency and ep oducibili y o he s udy.
C. Li e a u e Sea ch
An upda ed sea ch o smoking de ec ion echnologies and
smoking cessa ion applica ions was conduc ed in Decembe
2023 using PubMed and ACM Digi al Lib a y da abases. The
ollowing sea ch s a egy was used: (“smoking” AND “de ec-
ion sys em”) OR (“smoking” AND “senso ”) OR ((“smoking”
AND “de ec ion sys em”) OR (“smoking” AND “senso ”))
OR ((“smoking cessa ion”) AND (“applica ion” OR “app” OR
“sma phone app”)). The ull sea ch s a egy is p o ided in
Suppl. Ma e . 2. All he s udies published since 2019, yea
o publica ion o he p e ious e iew, we e included. The e
we e no limi a ions based on language. The e e ence lis s o
he included s udies unde wen addi ional sc u iny o iden i y
addi ional po en ial s udies. We manually sea ched key pee -
e iewed scien i ic jou nals in he ield o obacco esea ch
(speci ically, Nico ine & Tobacco Resea ch, Tobacco Con ol,
Ca cinogenesis, Heal h Educa ion Resea ch, and Con ibu ions
o Tobacco and Nico ine Resea ch). Two au ho s o he e iew
independen ly examined and chose s udies om he conduc ed
sea ches. Any disag eemen s we e esol ed h ough discussion
o , i necessa y, wi h he in ol emen o a hi d e iewe .
1) Web-Based Sea ch o Smoking Cessa ion Applica ions:
Fo smoking cessa ion applica ions, an addi ional web-based
sea ch was ca ied ou . The selec ion p ocess was conduc ed as
ollows: we pe o med mul iple sea ches using Bing, Google,
CASU e al.: SMOKING DETECTION AND CESSATION: AN UPDATED SCOPING REVIEW OF DIGITAL AND MOBILE HEALTH TECHNOLOGIES 5193
Fig. 1. PRISMA-ScR (P e e ed Repo ing I ems o Sys ema ic e iews and Me a-Analyses ex ension o Scoping Re iews) low diag am
ep esen ing he a icle selec ion p ocess in acco dance wi h he guidelines o upda es o sys ema ic e iews [24].
and DuckDuckGo sea ch engines using he que y “qui smoking
apps.”F om he ini ialsea ch esul s,weexcludeden ies ela ed
o sponso ed links p omo ing speci ic apps. Ins ead, we ocused
on links associa ed wi h blogs dedica ed o heal h opics (e.g.,
heal hline.com). F om his esul ing lis o apps, we chose hose
wi h high a e age use a ings in he And oid and iOS app
s o es (i.e., a ings o 4/5 o highe ). Then, a u he li e a u e
sea ch was conduc ed on Google Schola combining he names
o app iden i ied ia web as ollows: (“name o he applica ion”
AND “applica ion”). This addi ional s ep, which was pe o med
o each applica ion, enabled e i ica ion o which apps we e
clinically assessed.
D. Eligibili y C i e ia
The eligibili y c i e ia o he inclusion o s udies in his
scoping e iew we e as ollows:
S udies ha epo ed on de elopmen , e alua ion, o appli-
ca iono adigi al o mobileheal h echnology o smoking
de ec ion o cessa ion.
Technology ha in ol ed sma phone, sma wa ch, wea -
able de ice, o o he senso -based sys em.
S udies in he o m o o iginal esea ch a icles (includ-
ing andomizedcon olled ials),c oss-sec ional, coho s,
b ie epo s, case epo s, case se ies communica ions,
me hodologies, and me hods.
S udies published in pee - e iewed jou nals o con e ence
p oceedings.
S udies published be ween 2019 and 2023.
E. Exclusion C i e ia
The exclusion c i e ia we e he ollowing:
S udies ha did no ocus on smoking de ec ion o cessa-
ion as a p ima y o seconda y ou come.
Technology ha did no in ol e senso o mo ion da a
collec ion o analysis.
S udies no w i en in English.
Apps ha we e no a ailable in English.
S udies in he o ms o abs ac , p ep in , edi o ial, com-
men a y, le e , o e iew.
S udies published be o e 2019 (excep hose included in
he p e ious e sion o his e iew).
F. Da a Ex ac ion
Two e iewe s independen ly pe o med da a ex ac ion. Any
inconsis encies we e esol ed h ough discussion o wi h he
assis ance o a hi d e iewe . In ou analysis, we ca ego ized he
e iewed echnologies in o wo main g oups: smoking de ec ion
echnologies and smoking cessa ion applica ions.
1) Smoking De ec ion Technologies: Fo each s udy, he ol-
lowing elemen s we e sys ema ically ex ac ed and compiled
5194 IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS, VOL. 29, NO. 7, JULY 2025
in s udy ables: p oduc , echnology, ope a ing sys em (OS),
goal, pa icipan s, hou s, ecall, F-sco e and A ea Unde Cu e
(AUC).
2) Smoking Cessa ion Applica ions: Fo each s udy, he ol-
lowing i ems we e ex ac ed and adap ed in o app op ia e ables:
p oduc , echnology, mobile ope a ing sys ems, scien i ic e alu-
a ion (Sc. E al.), public a ailabili y, and lis p ice. Wi hin each
ca ego y, we also highligh ed whe he smoking cessa ion appli-
ca ions ha e been suppo ed by scien i ic a icles and whe he
hese ha e unde gone assessmen , o example, on a clinical
le el. Sma phone applica ions we e di ided in o wo subsec-
ions based on whe he hey ha e been subjec ed o clinical
assessmen o no , and he e o e a leas one phase o es ing
on eal samples.
III. RESULTS
A. S udy Cha ac e is ics
The sea ch yielded a o al o 37 dis inc documen s, o be
added o 12 s udies e ie ed om he p e ious e sion o
he e iew (6 ela ed o smoking de ec ion echnologies, and
6 ela ed o smoking cessa ion applica ions). Gi en ha he
u iliza ion o wea able de ices o smoking de ec ion is el-
a i ely ecen , mos o he loca ed wo ks pe ain o p oduc s
cu en ly unde going e alua ion o s ill in he expe imen al
p o o ype phase. The a icle selec ion p ocess is epo ed in he
PRISMA-ScR, compiled in acco dance wi h he guidelines o
upda es o sys ema ic e iews (Fig. 1). The ull lis o included
s udies is epo ed in Table I. A o al o 26 s udies we e iden i ied
o smoke de ec ion echnologies and 20 o smoking cessa ion
applica ions, 2 o which we e subsequen ly added o sugges ions
ob ained h ough he e iew p ocess. Fu he mo e, 3 s udies
ela ed o gene ic smoking de ec ion we e included. The main
cha ac e is ics o he included s udies a e explained below. The
emaining 60 e e ences ci ed h oughou his pape p o ide
con ex ual backg ound, heo e ical aming, o supplemen a y
discussion bu we e excluded om o mal analysis o main ain
ocus on he co e esea ch ques ions. This app oach aligns wi h
scoping e iew me hodologies, which p io i ize dep h on key
hemes while acknowledging b oade schola ly discou se.
B. Technologies o Smoking E en s De ec ion
The echnologies discussed in his pa ag aph aim o de ec
smoking e en s in eal- ime, elimina ing manual acking. Some
a e ma ke - eady, o he s a e unde e alua ion. They ypically
use a wea able de ice and sma phone app o iden i y smoking-
ela ed mo emen s like hand- o-mou h ac ions.
The de elopmen o smoking de ec ion sys ems has un-
de gone signi ican ad ancemen s, pa icula ly in le e aging
wea able echnologies and machine lea ning. Lopez-Meye
e al. [26],[27] laid ea ly ounda ions using espi a o y induc-
i e ple hysmog aphy (RIP) senso s and w is -wo n de ices o
de ec smoking ges u es (see Fig. 2). Thei app oach u ilized
Suppo Vec o Machines (SVM) and h eshold-based algo-
i hms, achie ing ecall a es o 80–90%. Howe e , hei sys em
was limi ed o con olled se ings, equi ing o line p ocessing
Fig. 2. Senso s o he sys em depic ed in Lopez-Meye e al.’s
wo k [26].
and p o iding minimal adap abili y o ee-li ing en i onmen s.
Mo ing o wa d, sys ems like SmokeBea [28] enhanced de-
ec ion by inco po a ing accele ome e s and gy oscopes in o
comme cial sma wa ches. SmokeBea combined p obabilis ic
models wi h ges u e segmen a ion, yielding p ecision and ecall
a esexceeding85%.Simila ly,RisQ[29]le e agedCondi ional
Random Fields o sequence smoking ges u es in ee-li ing con-
di ions,whileS opWa ch[30]adop edRandomFo es classi ie s
o dis inguish smoking om o he ac i i ies. These sys ems
demons a ed he po en ial o low-cos , use - iendly pla o ms,
achie ing ope a ional accu acies be ween 70–90%.
Pos -2020 s udies demons a ed ema kable ad ances in
smoking beha io de ec ion h ough inc easingly sophis ica ed
me hodological app oaches. Senyu ek and colleagues [31] de-
eloped a wea able sys em in eg a ing espi a o y induc i e
ple hysmog aphy (RIP) and ine ial measu emen uni (IMU)
senso s, employing a hyb id deep lea ning amewo k combin-
ing con olu ional neu al ne wo ks (CNN) and long sho - e m
memo y (LSTM) ne wo ks. Thei esea ch u ilized a comp e-
hensi e da ase e alua ed h ough lea e-one-subjec -ou c oss-
alida ion, achie ing an F1-sco e o 78%. Simila ly, Ki mizis
e al. [32] employed an a i icial neu al ne wo k wi h con olu-
ional and ecu en laye s, u ilizing he Smoking E en De ec-
ion (SED) and Smoking E en De ec ion F ee-Li ing (SED-
FL) da ase s. Thei wo-s ep me hodology le e aged sma -
wa ch da a o de ec indi idual pu s and localize smoking
sessions, achie ing imp essi e weigh ed accu acies o 0.968
and F1-sco es o 0.878. Agac e al. [33] ad anced senso u-
sion me hodologies, u ilizing accele ome e s and gy oscopes
om sma wa ches (LG Wa ch R, LG Wa ch U bane o Sony
Wa ch 3) and sma phones (Samsung Galaxy S2 o S3). Thei
amewo k inco po a ed use -speci ic ea u es, such as body di-
mensions, in o a Random Fo es classi ie o achie e 83% ecall
in dis inguishing smoking ges u es om o he hand- o-mou h
ac i i ies. They alida ed he model using a comp ehensi e
da ase collec ed unde ee-li ing condi ions, demons a ing
he impo ance o pe sonaliza ion in wea able sys ems. Mo e
ecen ad ancemen s, such as Hnoohom e al.’s [34], ad anced
CASU e al.: SMOKING DETECTION AND CESSATION: AN UPDATED SCOPING REVIEW OF DIGITAL AND MOBILE HEALTH TECHNOLOGIES 5195
TABLE I
SUMMARY OF THE STUDIES INCLUDED IN THIS SCOPING REVIEW,LISTEDINTHEORDER THEY APPEAR IN THE TEXT,REFLECTING THEIR RELEVANCE TO
DIFFERENT TOPICS AND SECTIONS OF THE ARTICLE
smoking ges u e de ec ion h ough a sophis ica ed ResNe SE
amewo k, in eg a ing deep esidual ne wo ks wi h a en ion
mechanisms. By analyzing he UT-Smoke da ase collec ed
om 11 olun ee s o e h ee mon hs, he esea che s com-
pa ed hei app oach agains i e baseline models (CNN, LSTM,
BiLSTM, GRU, and BiGRU; see Fig. 3). The ResNe SE model
demons a ed excep ional pe o mance, consis en ly achie ing
op accu acy and F1-sco es o 98.65%, 98.39%, and 98.63%
ac oss mul iple scena ios, highligh ing i s supe io capabili ies
in eal- ime ges u e ecogni ion.
Thaku and colleagues [35] de eloped a obus ac i i y ecog-
ni ion amewo kusinga6-axisine ialmeasu emen uni (IMU)
senso , explo ing mul i-class classi ica ion models including
Logis ic Reg ession, k-Nea es Neighbo , Adap i e Boos ing,
Random Fo es , Suppo Vec o Machine, and Decision T ee.
Magui e e al. [21] in oduced a pa icula ly inno a i e mul-
imodal sys em combining a sma wa ch (wi h accele ome e s
and gy oscopes) and a wea able inge senso , and an An-
d oid app (Fig. 4), using a Tenso Flow Li e model o ac i i y
classi ica ion. Thei sma wa ch-only sys em achie ed accu acy
imp o emen s om 75.8% o 85.5% by in eg a ing he inge
senso . Fu he mo e, Sha ma e al. [22] ad anced he ield wi h
a mic ocon olle -based sys em employing a con olu ion-based
ne wo k and Neu al A chi ec u e Sea ch (NAS) o de elop cus-
om Deep Neu al Ne wo k (DNN) models. Mukhopadhyay’s
esea ch [36] u ilized ein o cemen lea ning o op imize CNN
5196 IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS, VOL. 29, NO. 7, JULY 2025
Fig. 3. ResNe SE model included in Hnoohom and colleagues’ wo k [34].
TABLE II
SUMMARY OF THE SMOKING DETECTION TECHNOLOGIES DESCRIBED IN THIS ARTICLE
Fig. 4. Diag am o smoking cessa ion echnology desc ibed by
Magui e and colleagues [21].
a chi ec u es, achie ing a pu de ec ion F1-sco e o 0.81, while
Alha bi e al. [37] in oduced SmokeMon, a ches -wo n he mal-
sensing sys em ha demons a ed high-p ecision smoking e en
de ec ion ac oss labo a o y and eal-wo ld en i onmen s.
A summa y o hese s udies and hei ecall a es can be seen
in Table II, while a summa y o he sma wa ches employed in
hem can be ound in Table III.
TABLE III
SUMMARY OF THE COMMERCIALLY AVAILABLE-TO-PUBLIC SMARTWATCH
EMPLOYED IN THE STUDIES INCLUDED IN THIS SCOPING REVIEW
C. Smoking Cessa ion Applica ions
The applica ions delinea ed in his sec ion a e ounded upon
he anno a ion o smoking beha io s, accomplishmen s, and in-
s ances o c a ing. These applica ions ha e a ained exceedingly
ele a edmean eedback a ings omuse sand ha e expe ienced
subs an ial a es o downloads wi hin he applica ion ma ke s.
1) Clinically Assessed Applica ions: Smoking cessa ion ap-
plica ions highligh s a obus and inno a i e landscape o digi al
in e en ions, each employing unique s a egies o suppo use s
in hei qui ing jou neys. Th ough scien i ic hi d-pa y e alua-
ion o endo semen , hese applica ions demons a e a commi -
men oin eg a inge idence-basedme hodologies,pe sonalized
CASU e al.: SMOKING DETECTION AND CESSATION: AN UPDATED SCOPING REVIEW OF DIGITAL AND MOBILE HEALTH TECHNOLOGIES 5197
ea u es,andad anced echnologies oenhanceuse engagemen
and e icacy.
Chen e al. [38] showcased an And oid-based sys em combin-
ing wea able senso s and ailo ed qui ing plans, emphasizing
pe sonaliza ion h ough demog aphic and beha io al da a. This
app oach aligns closely wi h mind ulness-based s a egies, such
as he RAIN me hod [39], o aid in managing c a ings, while
in eg a ing suppo i e messaging o bo h use s and hei social
ne wo ks. Pi o , a widely e alua ed p og am, s ands ou o i s
inco po a ion o an FDA-clea ed b ea h senso ha o e s eal-
ime physiological eedback, alongside a comp ehensi e app
p o iding cus omized lessons, p og ess acking, and coaching.
I s compa ibili y wi h iOS and And oid pla o ms, coupled wi h
mul iple clinical alida ions [40],[41],[42],[43], unde sco es
i s scalabili y and e ec i eness. Simila ly, Cu eApp Smoking
Cessa ion (CASC) in eg a es beha io al and pha macological
he apies wi h a mobile exhaled CO checke , demons a ing
signi ican imp o emen s in abs inence a es and educ ions in
nico ine dependence compa ed o con ol g oups [44]. This app
exempli ies he po en ial o hyb id digi al and pha macological
app oaches.
Apps like Sma S op and C a ing To Qui ! ocus on combin-
ing echnology wi h beha io al science. Sma S op le e ages
a p og ammable nico ine pa ch synch onized wi h a sma -
phone app o add ess peak c a ing pe iods, while C a ing
To Qui ! in eg a es cogni i e beha io al he apy (CBT) and
mind ulness p ac ices o dis up smoking pa e ns [45],[46].
These in e en ions highligh he in e connec ed physiological
and psychological dimensions o smoking cessa ion. By exam-
ining how s ess comp omises p e on al co ex unc ion and
inc eases smoking ulne abili y, esea che s illumina e he neu-
ological unde pinnings o addic ion [47]. Mind ulness he apy
o e s a p omising app oach o modula ing desi e and ciga e e
use, e ealing he complex mechanisms ha sus ain obacco
dependency [48].
Apps employing gami ica ion and in e ac i e ea u es, such
as Clicko ine, Smoke F ee, Kwi , and Qui Genius, ha e demon-
s a ed e ec i eness in enhancing sel -e icacy and mo i a ion
h ough ewa ds sys ems, p og ess acking, and engaging chal-
lenges [49],[50],[51],[52],[53],[54],[55],[56]. In his ega d,
Qui START exempli ies ano he ace o smoking cessa ion
suppo , combining p og ess acking wi h s a egies o manage
c a ings and nega i e moods. The app employs use da a o
o e pe sonalized challenges, ad ice, and mo i a ion, ensu ing
an in e ac i e and engaging cessa ion jou ney [57]. Gami ied
elemen sappea pa icula ly in luen ialin os e ing use engage-
men and add essing cogni i e ac o s c i ical o qui ing.
No ably, Accep ance and Commi men The apy (ACT) has
eme ged as a ecu en heme, unde pinning he design o 2Mo -
owQui , Sma Qui , and iCanQui [12],[58],[59]. These apps
le e age ACT p inciples [60] o build psychological lexibili y,
mi iga e c a ings, and p omo e mind ulness, wi h p omising
ou comes in abs inence a es and beha io modi ica ion. Finally,
Qui Genius and o he CBT-based apps demons a e a holis-
ic app oach, add essing no only smoking cessa ion bu also
b oade addic ion challenges. Thei in eg a ion o pe sonalized
plans, ex ensi e CBT exe cises, and suppo i e communi ies
e lec s a comp ehensi e s a egy aimed a sus aining long- e m
change [61],[62].
2) Applica ions No Ye Clinically Assessed: Se e al pub-
licly a ailable apps aid in smoking cessa ion and do no , ye ,
p o ide clinical assessmen .The LIVESTRONG MyQui Coach
and Qui Smoking: Cessa ion Na ion o e goal-se ing and
communi y suppo . Qui Now! p o ides mo i a ional messages
and suppo s mul iple languages. The Qui Smoking wi h An-
d ew Johnson app uses sel -hypnosis, while Bu Ou p o ides
insigh s o a smoke- ee li es yle. Ge Rich o Die Smoking
mo i a es h ough mone a y incen i es, and SmokeF ee—Qui
Smoking Slowly o e s op ions o qui ab up ly o g adually. The
Qui Smoking NOW—Max Ki s en app uses hypnosis and NLP
echniques, and he Qui T acke : S op Smoking app displays
inancial sa ings and heal h bene i s. Qui I Li e helps use s se
pe sonalized goals, and Qui Smoking Hypnosis o e s daily
hypnosis sessions. Qui e ’s Ci cle suppo s smoking cessa-
ion wi h esou ces and a qui und ool. EasyQui p o ides
a pe sonalized qui plan and a dis ac ion game. Flamy o e s
pe sonalized plans and ewa ds, and Smoking Log helps educe
ciga e e consump ion. All hese apps a e a ailable on iOS and
And oid, wi h some o e ing p emium ea u es. This subsec ion
p o ided a b ie summa y o some sma phone apps designed
o smoking cessa ion ha lack published pee - e iew. A mo e
comple e lis can be ound in Table IV.
3) Sma phone Apps Limi a ions: Engagemen wi h sma -
phone apps, pa icula ly hose designed o smoking cessa-
ion, aces se e al limi a ions. One key issue is he lack o
pe sonaliza ion and adap abili y o he use ’s changing needs
and con ex s, which can lead o dec eased engagemen o e
ime [63],[64]. Mo eo e , many apps do no adequa ely as-
sess he use ’s eadiness o qui smoking o a ange ollow-up,
which a e c ucial o main aining engagemen [65]. Imp o ing
engagemen could in ol e inco po a ing mo e use -cen e ed
design p inciples, such as eal- ime messaging wi h suppo
ne wo ks and educing ba ie s o access [63]. Fu he mo e,
he use o assessmen ools like he Mobile Applica ion Ra ing
Scale (MARS) can p o ide aluable insigh s in o app quali y,
includingengagemen , unc ionali y,aes he ics,andin o ma ion
quali y [66],[67],[68]. Howe e , i ’s impo an o no e ha com-
me cializa ion o apps does no necessa ily imply widesp ead
a ailabili y o he gene al public. Fo ins ance, he Pi o App
ope a es on a B2B model [69],[70],[71], which may limi i s
accessibili y o only ce ain o ganiza ions o g oups. The e o e,
while comme cial apps may be widely ma ke ed, hei ac ual
accessibili y may be mo e limi ed [72],[73].
4) Apps Usabili y: The usabili y and con enience o he de-
sc ibed applica ions play a c ucial ole. While a gene ic ap-
plica ion may achie e high pe o mance in e ms o smoking
de ec ion, i could p o e incon enien o use. A undamen al dis-
inc ion exis s be ween apps ha p o ide in o ma ion in a s and-
alone manne , wi hou he need o addi ional de ices, and hose
ha ope a e wi hmul imodal in o ma ion ommul iple sou ces.
Table IV illus a es ha mos o he desc ibed applica ions do
no equi e addi ional de ices, making hem use - iendly ools.
Howe e , i is e iden ha applica ions u ilizing supplemen a y
in o ma ion, such as Sma S op, Pi o , and Cu eApp Smoking
5198 IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS, VOL. 29, NO. 7, JULY 2025
TABLE IV
SUMMARY OF THE QUIT SMOKING APPLICATIONS DESCRIBED IN THIS STUDY,INCLUDING THOSE WITH “FREEMIUM”MODELS,WHERE THE APP IS FREE TO
DOWNLOAD AND USE,BUT OFFERS ADDITIONAL PAID FEATURES OR CONTENT
Cessa ion, achie e highe pe o mance, ollowing he p inciple
ha mo e da a equa es o g ea e knowledge and, consequen ly,
be e pe o mance. On he o he hand, howe e , he use o
addi ional de ices is in some cases an un easible and in o he s an
incon enience ha discou ages hei long- e m use. The e o e,
al hough i will always be easie o use a s and-alone applica ion
and simple o use a mul imodal sys em, he co ec ade-o
depends on he scena ios whe e i is o be used.
D. Nico ine De ec ion in Smoke De ec ion Sys ems
Tai e al. [74] in oduced he “s-band,” a wea able nico ine
senso employing a gold nanodend i e-modi ied wo king elec-
ode and a sel -assembled monolaye , enabling high sensi i i y
and s abili y in de ec ing nico ine om human swea . Valida ed
in bo h bu e solu ions and eal-wo ld samples om smoke s,
he senso eliably iden i ied nico ine le els consis en wi h
ciga e e nico ine con en , highligh ing i s po en ial o public
heal h and pe sonalized medicine applica ions. Rani e al. [75]
ad anced he ield wi h a me al-o ganic nano ube (MONT)
senso capable o selec i ely de ec ing nico ine in ciga e e
smoke a concen a ions below 23.3 µM. The MONT’s po ous
s uc u e, combined wi h apid esponse imes (20 seconds)
and sunligh s abili y a oom empe a u e, allows o e icien
nico ine de ec ion h ough isible ligh -d i en binding o me al
ions. I s eusabili y a e hea ing a 110◦C unde acuum en-
hances cos -e ec i eness and p ac icali y ac oss gaseous and
solu ion-phase applica ions. Meanwhile, Rahman e al. [76]
de eloped awi eless, ba e y- ee,skin-moun ednico inesenso
using anadium dioxide (VO2) echnology o de ec nico ine
apo om e-ciga e es. By le e aging elec on ans e be ween
nico ine molecules and he VO2 su ace, his senso achie es
p ecise apo ized nico ine de ec ion, suppo ed by densi y
unc ional heo y (DFT) calcula ions and composi ional analy-
sis. I s ligh weigh design acili a es con inuous moni o ing o
bo h pe sonal and en i onmen al use.
E. Non-Wea able Smoking De ec ion: Deep Lea ning,
Wi eless Signals, T ials, Da ase , and Ges u e De ec ion
1) Smoking De ec ion Th ough Vision: Recen ad ance-
men s in non-wea able smoking de ec ion ha e made signi -
ican s ides, le e aging deep lea ning echniques and no el
sys em designs o enhance accu acy and e iciency. Macalisang
e al. [77] de eloped a smoking de ec ion sys em using a da ase
o 300 images and he YOLO 3 model, achie ing high ain-
ing and alida ion accu acies o 98.10% and 98.22%, hough
challenges wi h de ec ion angles and ideo quali y emained,
wi h accu acies a ying om 63% o 98% in eal-wo ld es ing.
Weie al.[78] expanded his wo k by building a la ge da ase
o 9,424 smoking images and employing da a augmen a ion
echniquessuchasMosaicenhancemen ,whichimp o edgene -
aliza ion. Thei model, op imized wi h he DIoU_Loss unc ion
and adjus ed lea ning a es, showed enhanced pe o mance, pa -
icula ly in A e age P ecision (AP) and In e sec ion o e Union
(IoU), unde sco ing he model’s obus ness. Zhang e al. [79]
in oduced CBAM-Tiny, a ligh weigh a en ion mechanism de-
signed o imp o e small a ge de ec ion by e ining spa ial
ea u es wi h global pooling and u ilizing a cus om DenseBlock
module o be e g adien low. Thei model achie ed an mAP
o 86.32% and a ame a e o 55 ames pe second, demon-
s a ing bo h p ecision and speed, which is c ucial o eal- ime
applica ions. Finally, Chong [80] de eloped a eal- ime sys em
u ilizing he Real-Time S eaming P o ocol (RTSP) o cap u e
ideo ames and p ocess hem h ough a cus om model ained
on he Tsinghua-Tencen 100 K da ase . This sys em employed
CASU e al.: SMOKING DETECTION AND CESSATION: AN UPDATED SCOPING REVIEW OF DIGITAL AND MOBILE HEALTH TECHNOLOGIES 5199
Non-Maximum Supp ession (NMS) and a con ex in o ma ion
co ela ion algo i hm o imp o e de ec ion accu acy and p o-
cessing speed, ou pe o ming models like YOLO 3, SSD, and
Re inaNe .
2) De ec ion o Non-Ciga e e Smoke: Gau e al.’s e-
iew [81] explo es smoking de ec ion, discussing challenges
wi h smoke obscu ing da a and he ea u es used in algo i hms.
They highligh he need o da ase es ing and ad anced me h-
ods like qua e nionic wa ele ea u es, Kalman il e ing, and
ansmission-based de ec ion. Xu and Xu [82] combined s a ic
anddynamic ea u es o AI-basedde ec ion.Sapona ae al.[83]
used deep lea ning o eal- ime i e and smoking de ec ion,
le e aging he NVIDIA Je son Nano’s CPU (Cen al P ocessing
Uni ) and GPU o pa allelize neu al ne wo ks. They ocused on
he YOLO 2 de ec o , achie ing a de ec ion a e o 21 FPS.
Gu e al.’s s udy [84] e alua es a Deep Dual-Channel Con olu-
ional Neu al Ne wo k (DCNN) o smoking de ec ion, which
ou pe o ms o he models in e ms o s abili y and e iciency.
The DCNN su passes p ocessing imes o o he models and
excels a ex ac ing de ailed and basic ea u es. These s udies
collec i ely ushe in a new e a in algo i hm-d i en i e and
smoking de ec ion.
3) Smoking De ec ion Using Deep Lea ning: Jeong and
Ha [85] explo ed a deep lea ning-based sys em o smoking de-
ec ion using CCTV oo age, in eg a ing OpenPose-based skele-
onanalysiswi hspecializedha dwa e o enhanced ecogni ion.
Thei sys em p ep ocesses image da a o ecognize smoking
beha io , coupling i wi h senso -equipped de ices o de ec
smoke componen s, igge ing wa nings o non-smoking a eas.
A neu al ne wo k buil wi h Tenso Flow and Ke as, op imized
wi h MobileNe V2, achie es 75% accu acy o smoking im-
ages and 70% o non-smoking images, o e ing a p omising
eal- ime smoking de ec ion amewo k. On a di e en on ,
Lai e al. [86] ocused on smoking cessa ion by le e aging da a
om a p og am in no he n Taiwan spanning om 2010 o 2018.
Using machine lea ning models like a i icial neu al ne wo ks
(ANN), suppo ec o machines (SVM), andom o es s (RF),
ando he s, heyaimed o p edic smokingcessa ionp obabili ies
based on ac o s such as pa ien cha ac e is ics, smoking habi s,
and nico ine dependence sco es. The ANN model ou pe o med
o he s wi h an accu acy o 0.640 and an ROC alue o 0.660,
o e ing a aluable p edic i e ool o smoking cessa ion p o-
g ams. Bo h s udies con ibu e o he unde s anding o smoking
beha io and cessa ion, wi h Jeong and Ha’s wo k enhancing
eal- ime de ec ion h ough image p ocessing and ha dwa e
in eg a ion, while Lai e al.’s esea ch p o ides insigh s in o
machine lea ning’s po en ial in p edic ing success ul smoking
cessa ion.
4) Human Beha io De ec ion Wi h Wi eless Signals: Song
e al. [10] de eloped a con ac less AI echnology using Channel
S a e In o ma ion (CSI) om wi eless signals o de ec human
mo ion, ocusingon dis inguishingbe weensi ingand s anding.
Theyused USRPde ices ocollec CSI da a om olun ee s and
analyzed i using MATLAB and sciki -lea n. Machine lea n-
ing models we e buil and es ed, wi h Random Fo es (RF)
pe o ming well and K-Nea es Neighbo s (KNN) being less
e ec i e. An ensemble classi ie imp o ed pe o mance, and he
CSI da ase ou pe o med a benchma k da ase . The model was
e ec i e inp ac ical applica ions,wi h local es sp o idingGUI
p edic ions and eal- ime es s o e ing CSI ampli ude g aphs
and web in e ace p edic ions.
5) Smoking De ec ion T ials: The smoking de ec ion ials
conduc ed ac oss a ious s udies demons a e he po en ial o
in eg a ing eal- ime, pe sonalized in e en ions in smoking
cessa ion. Ba alio e al. [87] u ilized a Jus -in-Time Adap i e
In e en ion (JITAI) model o help smoke s manage s ess, a key
igge o elapse. The s udy inco po a ed mul iple senso s, in-
cluding ches bands and w is bands, o ga he physiological and
beha io al da a o eal- ime analysis. By using s ess-de ec ion
algo i hms, he sys em p o ided indi idualized ea men op-
ions, such as s ess managemen p omp s, o p e en smoking
episodes du ing high-s ess momen s. In a simila ein, He nan-
dez e al. [88] ocused on he easibili y and e ec i eness o
mind ulness-based in e en ions deli e ed ia wea able senso s
ha acked physiological indica o s associa ed wi h nega i e
a ec ,sel - egula ion,andsmokingbeha io s.Usingdeeplea n-
ing echniques, he s udy pe sonalized in e en ions based on
eal- ime da a, o e ing a mo e dynamic and ailo ed app oach o
smoking cessa ion. Ho a h e al. [89] explo ed he e ec i eness
o a sma band-based sys em ha p o ided au oma ic smoking
de ec ion and mind ulness in e en ions, including he RAIN
echnique, which was ailo ed o help pa icipan s ecognize and
manage c a ings. In his ial, da a collec ed om he wea able
de ices we e used o assess ea men ideli y, adhe ence, and
use sa is ac ion, wi h smoking beha io and abs inence a es
also being acked.
6) Ges u e De ec ion: Ges u e de ec ion has e ol ed
h ough a ious app oaches, each con ibu ing o he
accu acy and e iciency o ac i i y ecogni ion sys ems.
Hnoohom e al. [90] de eloped an inno a i e Human Ac i i y
Recogni ion (HAR) wo k low inco po a ing da a collec ion
omwea ables,p e-p ocessing,model aining,andassessmen .
They in oduced he A -BiLSTM model, which in eg a ed a
BiLSTM laye , an a en ion laye , and a ully connec ed laye ,
demons a ing supe io pe o mance on he WISDM-HARB
Da ase . This model achie ed highe accu acy and F1-sco es
when combining w is -wo n accele ome e and gy oscope
da a wi h a 20-second window size, e alua ed using me ics
such as F-Sco e, Recall, P ecision, and con usion ma ices. In
con as , Agac e al. [91] ocused on a dynamically adap able
pa ame e selec ion me hod wi h he Conawac algo i hm o
ac i i y ecogni ion, which ailo ed senso pa ame e s based
on ac i i y complexi y. This dynamic app oach signi ican ly
imp o ed he F1-sco e by 7% o complex ac i i ies and
by 6% o e all, while also educing ene gy consump ion by
38%, main aining memo y size, and lowe ing CPU usage
by 15%. Thei me hod p o ed o be pa icula ly e ec i e
o ac i i ies like “smoking in a g oup” and “d inking while
si ing down,” showing imp o emen s o o e 20%. Meanwhile,
Pa el e al. [92] explo ed 3D ges u e ecogni ion h ough
wea ables, emphasizing he in eg a ion o senso da a om
sma wa ches and a mbands wi h image/ ideo da a. Thei
wo k aimed a imp o ing human-machine collabo a ion,
wi h a ocus on ges u e and pa e n ecogni ion o enhance
