ContentslistsavailableatScienceDirect
Information Fusion
journalhomepage:www.elsevier.com/locate/inffus
Online heart monitoring systems on the internet of health things environments: A survey, a reference model and an outlook
Marcus A.G. Santos
a, Roberto Munoz
b,c, Rodrigo Olivares
d, Pedro P. Rebouças Filho
e, Javier Del Ser
f, Victor Hugo C. de Albuquerque
a,∗aUniversity of Fortaleza, Fortaleza, CE, Brazil
bSchool of Informatics Engineering, Universidad de Valparaíso, Valparaíso, Chile
cCentro de Investigación y Desarrollo en Ingeniería en Salud, Universidad de Valparaíso, Valparaíso, Chile
dPontificia Universidad Católica de Valparaíso, Valparaíso, 2362807, Chile
eFederal Institute of Education, Science and Technology of Ceará, Fortaleza, CE, Brazil
fTECNALIA, Donostia-San Sebastián, Spain. Department of Communications Engineering, University of the Basque Country, Bilbao, Spain. Basque Center for Applied Mathematics, Bilbao, Bizkaia, Spain
a r t i c le i n f o
Keywords:
Internet of health things Heart
Bio sensors Online monitoring Reference model
a b s t r a ct
TheInternetofHealthThingspromotespersonalizedandhigherstandardsofcare.Itsapplicationisdiverseand attractstheattentionofasubstantialsectionofthescientificcommunity.Thisapproachhasalsobeenappliedby peoplelookingtoenhancequalityoflifebyusingthistechnology.Inthispaper,weperformasurveythataimsto presentandanalyzetheadvancesofthelateststudiesbasedonmedicalcareandassistedenvironment.Wefocus onarticlesforonlinemonitoring,detection,andsupportofthediagnosisofcardiovasculardiseases.Ourresearch coverspublishedmanuscriptsinscientificjournalsandrecognizedconferencessincetheyear2015.Also,we presentareferencemodelbasedontheevaluationoftheresourcesusedfromtheselectedstudies.Finally,our proposalaimstohelpfutureenthusiaststodiscoverandenumeratetherequiredfactorsforthedevelopmentof aprototypeforonlineheartmonitoringpurposes.
1. Introduction
Recently,cardiovasculardiseases (CVD)havebecometheleading causeofdeathintheworld,accountingforabout17.7million,or31%, of alldeaths in2017[1]. CVDsareassociatedwithdisorders ofthe heartandbloodvesselswhichmightoftenleadtocardiacarrhythmia, stroke,hypertension,andheartfailure.CVDsareclassifiedaschronic non-communicablediseases(NCD)whichalsoincludecancer,diabetes, andchronicrespiratorydiseases,forexample[2].NCDsareassociated withaslow,long-term,orevenlifelongadvancement,andareusually silentorsymptomatic,therebyaffectingthequalityoflife[3].
Severalfactorscontributeto thedevelopment of CVDs,but poor lifestylemanagementsuchasinadequatediet,sedentarylifestyle,stress, andlegalandillicitdrugusearedeterminingfactorsfortheprogression ofthesepathologies[4].Therefore,propercarecansignificantlyreduce thelikelihoodof heartproblemsaswellasroutineexaminationsand follow-upsbyacardiologistastreatmentexpendituresbecomeexces- sivewhencardiovascularriskfactorsarepoorlycontrolled.
∗Correspondingauthor.
E-mail addresses: [email protected] (M.A.G. Santos), [email protected] (R. Munoz), [email protected] (R. Olivares), [email protected](P.P.R.Filho),[email protected](J.D.Ser),[email protected](V.H.C.d.Albuquerque).
The electrocardiogram (ECG),even withmore thanone hundred yearsofuse,isstillthefirstnon-invasivetestusedforthediagnosisof CVDsandshouldbeinterpretedaccordingtotheclinicalinformation obtainedduringtheanamnesisandphysicalexamination.TheECGcan be recordedatrest andduringwork(ergometrictest)asthepatient mighthavearestingECGwithoutchanges,butwhensubmittedtostress, significant changes might occur [5]. Similar totheECG, the Holter monitoralsorecordstheelectricalsignalsoftheheart.Thedifferenceis thatthehandhelddevicemonitorscontinuouslyfor24hoursormore, andthus,itdetects cardiacarrhythmias thatmightoccur atvarying times during thestipulated periodof the examination.Moreover, in thecaseofdiseasesalreadydiagnosed,itcanevaluatetheeffectofthe treatment.
In this context, computational techniques can be used as com- plementary alternatives toconventional approaches. These methods analyze data or clinical treatments related to a cardiac evaluation fasterandmoreaccurately.Amongthesetechniques,wecanhighlight the following: Feature Extraction [6,7], Image Analysis and Image
https://doi.org/10.1016/j.inffus.2019.06.004
Received27December2018;Receivedinrevisedform21May2019;Accepted2June2019 Availableonline3June2019
1566-2535/© 2019ElsevierB.V.Allrightsreserved.
Processing [8,9], Signal Analysis and Signal Processing [10], and Prediction[11,12].Anotherrecentapproachusedformonitoringand informationmanagementistheInternetofThings(IoT),whichinvolves storingandtransmittingreal-timeinformationfromaphysicalnetwork byusingspecificsensors. Furthermore,in itsapplication inthefield ofhealth,IoTisreferred toastheInternet ofHealthThings (IoHT).
It is a rapidly progressingfield with various investments relatedto improvementandusefromIoT.Itisestimatedthatin2020,IoHTwill haveaneconomicimpactofUS$170billionaccordingtotheMcKinsey study[13].IoHTcancontributetotheexpansionofaccesstoquality healththroughdynamicmonitoringofthehumanbeingwithinhis/her environment.Inthisway,IoHT canimprovetheeffectiveness ofthe treatments,preventrisk situations,andassistthepromotionof good health.Furthermore, IoHT enhancesthe efficiencyof resourceman- agement through flexibility andmobility using intelligent solutions.
However, this requires a transition from clinical-centered treatment to patient-centered medical care such that each agent, such as the hospital,thepatients,andservices,areperfectlyconnected[14].
Researchersandinstitutions arededicatedtothe developmentof IoT-basedtechnologiesforhealthapplications.Researchessuchas[15–
21]discussvariousservices,applications,andsystemsthathavebeen developedbasedonwearablemedicalsensors(WMS),illustratingtheir objectivesandchallenges.Thedifferentialofthissurveyistopresent andanalyzetheadvancesofrecentstudiesinhealthsystemsbasedon IoT,focusingonsolutionsformonitoring,detection,andsupportofthe diagnosisofCVDsbycapturingandprocessingbiosignalsgeneratedby thehumanbody.
Wecansummarizecontributionsthatthissurveycoversfromthelit- eratureofthelastthreeyearsinthefield.Ourresearchprovidesamore detailedaccount ofthemostwidely usedandrelevantprotocolsand standardscomparedtootherresearchpapersinthefield.Moreover,our workprovidesextensiveknowledgeabouthowsecurityissuesarebeing addressedintheprojectsselectedforanalysis.Furthermore,wepresent howthisinformationcanbeusedforcompaniesthatwanttoexplore thecurrentlyunderdevelopedmarketforIoTsolutions forhealth.To determinewhetherelectrocardiography(ECG)orphotoplethysmogra- phy(PPG)isthebesttechnologyforthedevelopmentofitsapplication, wepresentastudyevaluatingtherespectivepointsassessedintheworks ontheselectedtechnology.
Another essential point that describes this work is that it is a potentialreferencematerialforallprofessionalsinvolvedinthehealth sector, such as physicians, nurses, physical therapists, and physical educators.Theyneedtoknowaboutthemodelsandmetricsforadopt- ingtheproposed solutionsin patientsassociatedwithcardiovascular diseases.Currently,therearenotsufficientnumberofarticlesdevoted exclusivelytopresentingevaluationsofthestudiesontheuseofIoHT tovalidate thediagnosisandmonitoring of CVDs;this indicatesthe valuablecontributionof thiswork. This studywouldaid any health careenthusiastinterestedintheuseofIoTsolutionsinthecollection, decisionsupport,andmonitoringofthecardiacfactorsdevelopasys- temicviewonthissubject.Finally,themanuscriptpresentsareference modelaimedathelpingfutureenthusiasts discoverandsubsequently enumeratethefactorsnecessaryforthedevelopmentofaprototypefor onlineheartmonitoringpurposes.
Therestof thisresearchisstructuredasfollows:InSection2,we presentthemethodologyofthissurvey,focusingontheresearchques- tions,searchprocess,inclusionandexclusioncriteria,qualityevalua- tion,anddatacollection.Section3describestheresultsof usingthe proposedmethod,analyzingandsubsequentlycitingtherelevantstudies withadescriptionofthetopicinterlinkedwiththeresearch.InSection4, wepresentadiscussionhighlightingcertainchallengesstilltobeover- come,andlater,wepresentamodelthatcanbeadoptedinprojects involvingthedevelopmentofexperimentsrelatedtotheIoTforheart monitoring.Finally,inSection5,openquestionsandfuturedirections regardingtheinternetfieldofthingsoftheheartareprovided.More- over,wehavesummarizedtheimplicationsofthissurvey.
2. Methodology
Themethodologyadoptedtoconductthissurveyisbasedonthree stages: (1) Todevelop research termsrelated toCVDsand theIoT.
The primaryobjective wastoobtainthemost significant numberof researchesofthesespecificapplicationsaddressedinourstudy.There- fore,thesearchstringdefinedwas:(“InternetofThings” or“IoT”)and thetermassociatedwithcardiovasculardiseases.Theexpressionsused inthisresearchrelatedtoCVDsareHeart,Cardiac,Electrocardiogram, Blood Pressure,Cardiovascular, Beats, Arrhythmia, Electrocardiogra- phy,Cardiology,HRV,Hypertension,Coronary,Myocardial,andCVD.
(2)Theexpressionswereadaptedtothesearchenginesaccordingtothe rulesdefinedbyeachdatabaseandbyfilteringthesearchonlybymag- azinesandcongresses.Forthearticleswefindtwoormoreexpressions of,weattempttoexcludethesamearticlesthatgeneratedduplicity.(3) Apply inclusioncriteria:Works relatedtowearabletechnologiesthat focusonusingIoTtoimprovetheprognosisorpossibledecisionsup- portbasedonthecaptureandprocessingofbio-signalsgeneratedby thehumanbodybymonitoringheartrateandbloodpressure.
Thissurveyincludesworkspublishedsince2015,giventhatwein- tendtofindthemostrecentresearchonthedevelopmentofhealthap- plicationsforCVDsbasedontheIoTinfrastructure.Theselectionofthe studieswasmadeinthebasesofIEEEXplore,Springer,ScienceDirect, SAGE,PLOS,MDPI,InderscienceOnline,Hindawi,Frontiers,andACM.
Otherpublicationsthatwerenotassociatedwithoneofthementioned databaseswereselectedusingGoogleScholar.
Finally,weobtainedatotalof115works,55fromIEEEXplore,19 fromSpringer,9fromScienceDirect,6fromMDPI,3fromACM,3from PLOS,1fromFrontiers,1fromHindawi,1fromInderscience,1from SAGE,and16 otherpaperswereselecteddue totheircontentbeing necessaryforthepresentationofthisresearchwork.
3. Onlinecardiacmonitoring
ThispaperpresentsIoT-basedcardiovascularhealthresearchofthe lastfouryearsanduncoversnumerousissuesrequiredtobeaddressed to transformhearthealth care technologies through IoTinnovation.
AsmentionedinSection2,therelevantworkisdividedintodifferent groupsorcategories,whichrepresenttheresearchers’focusinaddress- ingthepotentialofIoTinthefieldofhearthealth,consideringseveral practicalchallenges.Therefore,therearenowseveralapplications,ser- vices,processesandprototypesinthefieldofourresearch.Theresearch trendswefoundarerepresentedinFig.1andhavebeendiscussedinthe following subsections.General observations,includingcomputational resourcesusedinprototypes,arediscussedinSection4.
3.1. Monitoringsignalstotheheart
Making a medical diagnosis is an imperfect process based more on probability than on certainty. According toSackett et al. [22],
“Evidence-basedmedicineistheconscious,explicit,andjudicioususeof thebestavailableevidencethatiscapableofjustifyingdecision-making bycaringforindividualpatients”.Diagnostictestsareimportanttools tofacilitatethehealthprofessionals’decisionsregardingthetreatment thatwillbeadministeredtothepatient.Inthefaceofalistofsymptoms andsigns,somediagnostichypothesesgetdelineatedinthemindofthe healthprofessional.
3.1.1. HeartRate(HR)
Heartratereferstothenumberoftimesaperson’sheartbeatsper minute.Thenormalheartratevariesfrom persontoperson,butthe normalrangeforadultsis60to100beatsperminute(BPM)[5].How- ever,anormalheartratedependsontheindividual’sage,bodysize, heartconditions,whetherthepersonissittingormoving,andtheuseof medication;evenphysicalconditioningreducestheoverallrestingheart rate,makingtheheartmusclesworkmoreefficiently.
Fig.1.IoToftheHeart.
Anarrhythmiacausesthehearttobeattooquickly,tooslow,orwith anirregularrhythm.Tachycardiaisgenerallyconsideredtobeaheart rategreaterthan100BPMwhentheindividualisinarestingstateandis usuallycausedwhentheelectricalsignalsintheupperchambersofthe heartfunctionabnormally[4].Aconditionknownassupraventricular tachycardia(SVT)referstotheconditionwhentheheartrateiscloserto 150BPMormore;thisoccurswhentheheartratecontrollingelectrical systemisoutoftune,requiringmedicalcareinconjunction withthe patient.
3.1.2. BloodPressure(BP)
Bloodpressureisoneofthevitalsignsthatphysiciansmeasureto evaluateone’soverallhealth.Itreflectsthepressureofthebloodagainst thewallsofthearteries,giventhatwhentheheartcontracts,forcing bloodintothebody,thebloodpushesagainsttheinsideoftheblood vessels.Ahighbloodpressure,aconditionknownashypertension,with- outpropermedicalmonitoringcanresultinheartproblems,stroke,and othermedicalconditions.
Severalfactorscanincreasebloodpressure,includingstress,smok- ing,caffeine,excessivealcoholuse,certainprescriptiondrugs,lowtem- peratures,andmore.Moreover,aspeopleage,theyaccumulateplaques inthebloodvessels,andtheflexiblewallsofthearteriesbecomestiff,re- sultingintheheartmakinggreaterefforttopumpblood.Consequently, itstartstofail[5].
Bloodpressureisrecordedbasedontheratiobetweensystolicand diastolicpressure.Thefirstdescribesthepressurewhentheheartbeats.
The second is theperiod of contraction of theheartbefore beating again.Accordingtotheguidelinesannouncedin November2017by theAmericanHeartAssociation(AHA),measurementsofpeople’sblood pressureareorganizedintothefollowingcategories:Normal:Lessthan 120mmHg(mmHg) forsystolicand80mmHg fordiastolic;Elevated:
between120–129mmHgforsystolicandlessthan80mmHgfordias- tolic;Stage1Hypertension:between130–139mmHgforsystolicor80–
89mmHg fordiastolic;Stage2Hypertension:at least140mmHg for systolicoratleast90mmHgfordiastolic.
Sometimes, people confuse high blood pressure with high heart rate.Bloodpressureisthemeasureof theforceof thebloodagainst thewallsofthearteries,whileheartrateisthenumberoftimesthe heartbeatsperminute.Thereisnodirectcorrelationbetweenthetwo, andhighbloodpressure,orhypertension,doesnotnecessarilyresult in a high pulserate andvice versa. Heartrate risesduring strenu-
ous activity,butvigorous trainingcan only modestlyincreaseblood pressure.
3.1.3. BloodOxygenSaturation(SpO2)
Oxygeninthebloodisoneoftheimportantindicatorsofahealthy body,sinceitisdirectlyrelatedtobodymetabolism.Lowoxygensatura- tioninthebloodcancausefailureinorganfunctionsofthehumanbody, causingseriouscomplicationssuchasnecrosisandlossoffunction.
SpO2 is the measure of the amount of oxygen bound to the hemoglobin in thecells withinthe circulatory system.This measure providesanestimateofarterialoxygensaturationintheblood.Itispre- sentedasapercentageofoxygenatedhemoglobincomparedtothetotal amountofhemoglobinintheblood(oxygenatedandnon-oxygenated hemoglobin).NormalSpO2valuesvarybetween95and100%[5].
Mostbloodoxygenconcentrationmetersusethefingerasameasur- ingpoint,asthefingersarethinnerandfullofcapillariesandblood.
In [23], IoT-based heart disease monitoring system for generalized health serviceis proposed.Thissystemmonitorsthephysical signals ofpatients,suchasbloodpressure,SpO2,andECG.
3.1.4. BodyTemperature(BT)
Thestudiespresentedin[24–26]indicatethatthefallorincrease ofbodytemperaturecanbeassociatedwiththedevelopmentofcardio- vasculardiseases.Suddendeathfromcardiaccausesinathletesshould not be mistakenforabruptdeathrelatedtoheatstroke ormalignant hyperthermia.Inthesecondcase,thevictimhasusuallyexercisedex- cessivelyinwarmweather,oftenwithsportingequipmentthatdisrupts heatdissipation.Inaddition,duringthesetrainings,peoplecanconsume substancesthatcauseincreasedbodytemperatureandvasoconstriction, hamperingthechangeinbodytemperature.Thissituationleadstocol- lapsebecausewithhighbodytemperatureirreversibledamageinthe organsystemscandevelop[5].
Thesestudiesusebodytemperaturemonitoringasoneoftheparam- etersmeasuredintheirresearch.Theinvestigations[24]and[25]pro- poseremotemonitoringofthehumanbodyusingheartrate(HR)and body temperature(BT)asparameters.Thework[26]focuseson the earlydetectionofheartdiseasewhereBTisusedasoneofthepredic- torsforevaluation.
3.1.5. Electrocardiogram(ECG)
Inahypotheticalsituation,a31-year-oldpatientseeksacardiolo- gist’shelpinresponsetopainthatafflictshimfor2days.Thepatient
doesnotpresentwithanyrelevantmedicalhistory,deniestheuseof licitandillicitdrugs,presentsno symptomsofoverexertion,andthe paindoesnotimprovewithrest.Thepainbeganafteranargumentwith hiswife.Therefore,theprobability ofcoronaryarterydiseasein this patientisverylow[27].
Theinformationprovidedbyanormalelectrocardiogram(ECG)has arelativelymodestvaluebecauseitcannotgreatlyreducethepre-test probability[28].However,iftheECGrevealschangesinrepolarization, itismorereasonabletoassumethattheyareduetohyperventilation generatedbyanxietythandue tomyocardialischemia.Therefore,an ECGalteredinthispatientalsodoesnotmodifythepre-testprobability ofcoronaryarterydisease.Furthermore,alterationsofrepolarizationin theyoungpatientmightbeafalsepositiveresult.Theadvantagesofthe ECGarelowcostandhighavailability.However,forthediagnosisof anginapectoris,theErgometertest(ECGrecordedduringexertion)is moreaccurateandtheAngiocoronarographycanconfirmsuspicionsif obstructivelesionsareidentifiedinthecoronaryartery.
TheECGconstitutestherecordoftheelectricalactivitygeneratedby thecardiactissue[29].ThetraditionalECGusestwelveshuntstorecord theelectricalactivityoftheheart,involvingthreebipolarderivationsof thelimbs(D1,D2,andD3),threeamplifiedunipolarlimbshunts(AVL, AVR,andAVF),andsixprecordialunipolarshunts(V1-V6).TheECG cansupportahealthprofessional’sdecisioninidentifyingarrhythmias, ischemia,myocardialinfarction,cardiacchamberoverloads,inflamma- toryprocesses,drugeffects,metabolicdisturbances,andevendiseases causingtheriskofsuddendeath.
Theemergenceof theIoTasoneofthemost importantintercon- nected technologies has succeeded in transforming the health care sector.IoTallowsECGdevicestobeincorporatedintocustomizedand interconnectedsoftware,buttheycanalsototracktheirownactivity and results along with the activity of other devices connected to them[30].Thestudiesnotonlydiscusstherelevanceandapplicability ofIoTmonitoringforECG,buthavealsodevelopedsystemsforECG monitoringbasedonIoT.Thisdevelopmentcomprisesaportablewire- lessacquisitiontransmitterandawirelessreceiverprocessor[31–33], and[34].
3.1.6. Photoplethysmography(PPG)
Photoplethysmography(PPG)is anothermethodused tomeasure theheartcycleandobtainheartrateandbloodpressuremeasurements.
PPGmeasures thevolumetric change of theheartthrough themea- surementoftransmissionorreflectionoflight.Astheheartcontracts, thebloodpressureintheleftventricle,theprimarypumpingchamber, increases.Thisincreaseforcesapressurized“pulse” ofbloodintothe arteriesofthebodywhichswellalittlebeforereturningbacktotheir previousstate.ThebrightnessofaLEDlightsourceplacedonaspoton theskinincreasesthepulsepressure,causingameasurabledifference intheamountoflightreflectedbackortransmittedwhichiscaptured throughasensor.
Theamplitudeofthissignalisdirectlyproportionaltothepulsepres- sure;thehigherthepeak,thestrongerthepulse.Althoughthesignal mightbelessapparentifmeasuredontheskinatapointmoredistant totheheart,thechangeofpressureissufficienttoexpandthosearteries toameasurableextent.Eachpeakintheresultingsignalcanbeidenti- fiedbyaspecialheartratealgorithmthatcandeterminetheamountof timebetweeneachsuccessivepeak,therebyofferinganothermeansof measuringheartrate.
Photoplethysmographyisanindirecttechnique;thatis,itdoesnot recordthetrueactivitythatoccursintheheartsoitisnotalwaysac- curatewithmillisecondmeasurements.However,thestudiesdescribed in[35,36],and[37]focusonbetteraccuracyinobtainingtheresults.
IfwecomparethesemethodstotheECG,itispossibletodemonstrate theeaseofmeasuringtheheartrateorbloodpressureofanindividual.
Aspresentedinthesurveyof[38],thistopicthathasbeenstudiedin- creasinglyinthelasttwentyyears,hasshownprogress,andseemsto demonstratetheclinicalapplicabilityofPPGinthediagnosisofCVDs.
3.1.7. Wearable
Constant monitoring might present results in the investigative processofadiseasethatanECGmightnotidentify.Forinstance,an ergometer testor exertion ECGareequivalentdenominations ofthe test that evaluates cardiac work during a physical, scheduled, and progressiveeffort.Itisusedinthestudyofcoronaropathy,arterialhy- pertension,cardiomyopathy,arrhythmias,evaluationofdrugefficacy, cardiacrehabilitation,andphysicalconditioningrelatedtosports.
For heartmonitoring over long periods, the dynamic electrocar- diography, also known as the Holter system, involves, as a basic principle, the recordingof theelectrical activityof theheartfor 24 hours.Toperformthetest,theelectrodesareplaced inthethoraxof thepatientwhilewiresareconnectedtoanexternalrecordingdevice.
Holter is utilized in thediagnostic study of ischemic heartdisease, inpost-infarction evaluation,pacemakers, sinusnodedisease,Chagas cardiomyopathy,Wolff-Parkinson-Whitesyndrome,Transientischemic attacks,andinthediseaseevaluationoftheeffectofseveraldrugs.
Portable medical sensors (WMSS) are attracting more and more attention from the scientific and industrial communities. Driven by technological advances in detection, wireless communication, and machine learning,WMS-based systemshave begun totransformour dailylives[39].WMS,initiallydevelopedtoallowlow-costsolutions forcontinuoushealthmonitoring,nowextendsfarbeyondhealthcare.
Most of theconventional sensors such as ECG andPPG detect and collect biomedicalsignals. However, some sensors, i.e.,accelerators, collectrawdatathatcanbeusedtoextracthealth-relatedinformation.
Withcontinuousperformanceandimprovedefficiencyinreal-timesig- nalprocessing,thenumberandvarietyofWMSsincreasedsignificantly fromsimplepedometerstobloodpressuremonitors.
TheadoptionofWMScanofferalong-termhealthmonitoringalter- nativeindependentofthepatient’slocationthroughintegrationintothe moreextensiveinfrastructureofIoTthatwillprovideabetterdynamic inthecareanddiagnosisofpatients[40].
3.2. Wirelessandmobileheartmonitoringsystems
Theincreasedpopularityofvisibletechnologiesopenedthedoorto aIoT-basedsolutionsforhealthcare.Portablehealth-trackingdevices areexcellentcandidatestominimizethedistancebetweenthepatient andthedoctor.
Oneoftoday’smostprevalenthealthproblemsisthelowsurvival rateofsuddencardiacarrests.Allexistingsystemsforpredictingcar- diacarrestprimarilyconsiderheartrateparameters.Someresearchad- dressesthisproblembyconcentratingontheanalysisanddetectionof ECGsignalsthateventuallyleadtoarisk forecast,asabnormal ECG patternsindicatepotentialheartattack[41].
The use of thesmartphone asa resource in the development of an integration platform to monitor andeven predict a heartattack is quitebeneficial asitnaturallycombinescomponentsof detection, communication, and popular consumption. Currently, this device is widelyusedtorecognizephysicalactivitybasedonintegratedsensors withhighaccuracyandreliability.Forinstance,in[31],thefocusis tomeasureECGinrealtimewithelectrodesplacedonasmartphone.
ThecollectedECGsignalisstoredandanalyzedinrealtimethrough asmartphoneapplicationforprognosisanddiagnostics.Theinvestiga- tiongivenin[42]describesanECGmonitoringsystemcomposedofa sensor forremoteandcontinuousECGmonitoringforthelongterm.
Formonitoring,a smartphoneisused thatcombinestheinformation suppliedbythebuilt-inkinematic sensors(accelerometer,gyroscope, andmagneticsensor),torecognizethephysicalactivitiesofauserand, betteryet,theaccuracyfortheidentificationofabnormalECGpatterns.
Portable wireless monitoring systems cannot work without an application. This application is responsible for data collection and transmissionof messages tothoseinvolved in thecare process.The use of cloud computing inspired numerous projects in health care services[43].Withinthecloudsystem,medicaldatacanbecollected
andtransmittedautomaticallytomedicalprofessionalsfromanywhere andreal-timefeedbackcanbeprovidedtothepatient.In[44],amethod wasproposedforECGmonitoringbasedontheIoTCypressplatformfor integrateddevicesthatcollectsECGdatausinganon-intrusivewearable monitoringsensorandinwhichthedataistransmitteddirectlytothe cloudusingWi-Fi.
3.2.1. Bodyareanetwork(BAN)
Alsoknownasabodysensornetwork(BSN),BANisawirelessnet- workofwearabledevicesthatallowsthesynchronizationofdifferent bio-signalstoobtainanintegrateduserprofile.BANoperatesinclose proximitytothehumanbodyandcan beincorporatedintothebody throughimplantsormountedonthesurfaceofthebodyinafixedposi- tion.TheIEEE802establishedataskforcecalledIEEE802.15.6forthe standardizationofthebodyareanetwork.
IoT-basedcaredistributionmodelsareproposedtomonitorhuman biomedicalsignalsinactivitiesinvolvingphysicalexertion,incorporat- ingflexibilityincalculating thehealth application,andusing device resourcesavailablewithintheuser’sbodyareanetwork.Forinstance, in[45],acasestudywasconductedmonitortheheartfrequenciesof playersduringafootballmatchthroughasmartwatchandthereal-time dataacquiredbythisdeviceshows,asforeseen,notonlypossiblesitu- ationsofsuddendeathbutalsoinjuries.
Wirelesssensor technologiesassistinthesolutionofvariousneg- ativeaspects relatedtowired sensors, which arecommonlyused in hospitalsandemergencyroomstomonitorpatients,i.e.,aconventional electrocardiogram[46].Thevarioussetsofassociatedthreads,cables, andconnectorscausediscomforttothepatient,restricttheirmobility, increasetheiranxiety,andhindermanagementforhealthprofession- als[47].Furthermore,theuseofwirelessconnectionsislessnoticeable, reducespatientanxiety,anddecreasesthepossibilityoferrorsindata acquisition.Forinstance,thesystemproposed in[48]monitorsECG wavessimilartoaconventionalHoltermonitor;itanalyzesandsupports thedecisiononthediagnosisofarrhythmia.Ifarrhythmiaisdetected, thesystemsendsanalerttothedoctor.In[34],anultrasoundprototype wasdevelopedasawearabledeviceandexperimentalresultsshowed itspromisingperformancecomparedtothatofanelectrocardiogram.
3.2.2. PersonalAreaNetwork(PAN)
PersonalAreaNetworkisasetofambientsensorssurroundingthe patient.Inhealthcarenetworks,PANscanbeusedtoconnectBANs.The chiefuseofPANistodelivertheinformationmonitoredbythesensors toexternalnetworksortothenextdirectnodewiththosesensors.PANs arebeingwidelyusedinIoTapplicationssuchaslaptopsconnectedto smartphones.
BluetoothandZigBeearesomeof thetechnologiesused tomake PANavailableforhealthcarenetworks.Toincreasethescopeofsuch localnetworks,therewasamodificationoftheIPstacktofacilitatelow- powercommunicationthroughtheIPprotocol.Oneofthesolutionsis 6LoWPAN,whichincorporatesIPv6withlowpowerPANs.
TherangeofaPANwith6LoWPANissimilartothelocalareanet- worksandtheenergyconsumptionismuch lower[49].Studiessuch as[50]propose, for example,aplatform in which6LoWPAN nodes areattachedtobiomedicalsensors(ECG,bloodpressure,accelerometer, bodytemperature,etc.)tomeasureandtransmitdataviaPUtoRPUfor processingandanalysis.
3.2.3. WirelessLocalAreaNetwork(WLAN)
WhenoneormorePANnetworksareconnectedtoalocalareanet- work(LAN)wireless,they forma wirelessLAN betweenthem. WiFi technologyisusedbycertifiedproductsthatbelongtotheclassofwire- lessLANdevicesbasedontheIEEE802.11standard.Medicalmonitor- ingapplicationsrequiretheaccountofemergencyeventsaswellasthe physiologicalinformationmeasuredperiodically.Incriticalconditions, theemergencydatamusthaveaguaranteedaccountwithanacceptable delayforthespecificsignaltype.
Thehealthsystemisincreasinglyenhancingefficiencyandresults.
Theadoptionanddissemination ofIoTcanplayasignificantrolein thetenureofhealthcarecostswithoutaffectingthequalityofthecare providedtopatients.Forinstance,asthepopulationages,thecognitive, physical,andpsychologicalcapacitiesoftheelderlyvary,affectingthe qualityoflife.Remotemonitoringcanhelpinaccuratediagnosisand recoveryofpatients,primarilyifitiscarriedoutintheleastintrusive waypossibleandeveninthepatient’sownresidence[25].
3.2.4. Cloud
TheconceptofCloudisbasedontheofferingofITresourcesover theInternet.Presently,oneofthegreatestchallengesfacingthefieldof healthcareistheneedtostore,process,andanalyzeever-increasingvol- umesofmedicaldata,images,andstudiescontainingtheinformation ofeachpatient.Thiscrossoverofdatainthecloudmakesitpossibleto monitorpatientsoutsidehospitals,especiallywhenattemptingtomon- itortheminrealtime.
Patient remotemonitoringstructures(PPHM)proposetocombine IoT, cloud,andWLANtechnologies forefficient andhigh-quality re- motehealthstatusmonitoring.Inthecasestudypresentedin[51],a patientsufferingfromcongenitalheartfailure(ICC)whoneedsregular careathomedemonstratestheadequacyofthePPHMinfrastructure.
Intheconventionalhospital-centeredhealthsystem,patientsoftenuti- lizemultiplemonitors.Theproposedworkaddressesthechallengesof healthexpenditure,substantiallyreducinginefficiencyandwasteaswell asallowingpatientstostayintheirownhomesandgetthesamecare orevenbetter.
Amajorcurrenttrendinthehealthcarefieldisthedevelopmentof solutionsforremotemonitoringoftheECGsignalusingwirelesstech- nology. Thegatewaycan beauser-chargedmobiledevice, suchasa smartphone,whichwillserveasabridgetoconnectBANsandPANsto awide-rangenetwork(WAN)asshownin[31,52–56],and[57].
Wealsofindothertrendssuchasthestudies[58,59]and[39]that presentcloudapplicationsortoolkits.Theseworksfocusonmonitoring heartratevariation(HRV)in realtime, integratingportablesensors, cellphones,andaWEBsystem.Additionally,studiessuchastheone describedin[60]demonstratetheclinicalprocessofhypertensionbe- inganalyzedthroughamodelbasedonBusinessProcessManagement (BPM).Thisprocessisdescribedbyamonitoring-basedarchitectureof bloodpressuresensors,whichinturnincludesamobileapplicationre- sponsiblefortheanalysisofthepatient’shealthmonitoring.
3.3. MachineLearningfortheheart
Generally,clinicalstaff useadiagnostictesttoanticipatethepres- enceofadisease.Adiagnostictestisbasedonclinicalrecord,physical examination, laboratory sample or other information source.Result analysismustbecarefulbecausethiscanincreaseordecreasetheaccu- racyofthefinalmedicaldiagnostic.Techniquesofartificialintelligence arebeenappliedinthisfield[5],oneofthemisMachineLearning[61]. The Internet of Things (IoT) provides the modernization of the health caresector through thecontinuous,remote, andnon-invasive monitoring of cardiovascular diseases. In [62] an IoT platform for predicting cardiovascular disease using an IoT-enabled ECG system acquirestheECGsignal,processesthesignal,andalertsthespecialist toanemergency.TheresourcesareextractedfromtheECGsignalsand classifiedbyusingSVM.
Anotherexampleispresentedin[63],whichproposesthatnewin- formationrelatedtothephysicalstateofthepatientcanberemotely captured through abracelet.Theauthorsproposed theuse of linear regression(LM)andClassificationandRegressionTrees(CART)algo- rithmstoanalyzetheheartbeatandthedetectionofheartrate.
Machinelearningcanbeusedtocreateimages,identifyabnormal- ities,supportareasthatrequireattention,andimproveaccuracy,effi- ciency,andreliabilityofallprocessesinvolvedinamedicaldiagnosis.
Inthestudy[64],theuseofIoTandaneuralnetwork-basedhealthrisk
predictionframeworkisproposed,presentinganalternativetomonitor thehealth ofelderlycitizensandpreventtheonsetofcardiovascular diseases,forexample.Theresultsobtainedindicatethattheproposed structureusingIoThasagreataddedvaluetoensurethatelderlycitizens liveahealthylife.In[65],authorsuseartificialneuralnetworks(ANN) forpredictinghypertensiveinpatientsandthus,evaluatinglevelofrisk ofaheartfailure.Theyproposegeneratingcontinuousalertmessages reportedbycriticalchangesinbloodpressureusingonlysmartphones.
VitalpatientparameterssuchasECGandtemperaturearecontin- uouslymonitored by medical sensors in the hospital. Replacing the traditional methods of parameters being regularly monitored by a nursewithan autonomoussystemavoids humanerrors inmanually collectingpatientdataandconsequentlycanreducefatalitiesoccurring inhospitalsandhealthcentersduetohumandelayandneglect.In[66], an intelligent health care system is proposed, which can regularly monitorpatients’conditionsandminimizehumanneglect.Thesystem incorporatesNeuro-Fuzzy,whichisthecombinationofNeuralNetwork andFuzzylogic.TheNeuralNetworkadaptsandchangestoobtainthe desiredoutputvalueforinputresponsechanges,whichincreasesthe reliabilityofthemonitoringsystem.Fuzzylogicdealswithuncertainty and simulates human reasoning in the diffused data obtained. It logicallygeneratesfuzzyrulestomakethesystemrobust.
Theuseof physiologicalsensorstomonitor andassesshealth be- comesubiquitousinIoTaspresentedinthestudies[67,68],and[69]. Anexampleisprovidedthroughtheresearchof[70],whichwasbased on the heartbeat being recorded in wearable devices and sent to a server that links with other patient information and, furthermore, applyingSVMandlogisticregressiontotheinformation;studieshave beenconductedtopredictconditionswithandwithout stress.When peoplearestressedornervous,thereisanincreaseinheartratethat can lead toa heart attack.Machine learninghas also beenused to developnewapproachestocountertheIoTlimitations,whichwewill seeinmoredetailinthesectionMitigatingKeyChallenges.
3.3.1. Datafusionfortheheart
Additionally, more complex cases have been observed in which a single medical professional does not have enough information, experience,orbothtomakeadiagnosis.Here,themedicaladviceof healthprofessionalswithdifferentspecializationsshouldbediscussed andcorrelatedwiththeirindividualfindingstodiagnosetheissuemore accurately.Therefore,theboardheadcanmakethefinaldecision,based ontheinformationreceived.Fromthisperspective,theproposedmed- icaladviceresemblesahierarchicalorganizationinwhichlower-level membersprovideinformationforhigher-leveldecision-making[71].
AccordingtoBoström,datafusionisthestudyofefficientmethods toautomaticallyorsemi-automaticallytransforminformationfromdif- ferentsourcesanddifferentpointsintimeintoarepresentationthatpro- videseffectivesupportforhumanorautomateddecision-making[72].
Differentdatafusionstrategieshavebeenproposedovertheyearsto reducetheeffectofdisplacementandincreasetheaccuracyofhumanac- tivityrecognition,e.g.,inmobileandwearablesensors[73].According to[74],theseapproachescanbecategorizedintoProbabilisticmethods, whichincludetheanalysisofBayesiansensorvalues,statespacemodels, maximumlikelihoodmethods,possibilitiestheory,probativereasoning, morespecifically,theoryofevidence,KNN,andleastsquaresestimation methodsasKalmanfiltering.Ontheotherhand,statisticalmethodsin- cludecross-covariance,covarianceintersection,andotherrobuststatis- tics.Moreover,methodsofknowledgebasetheory,whichincludeintel- ligentaggregationmethods,suchasRNA,geneticalgorithms,andfuzzy logic.Finally,therearemethodsofevidencereasoning,whichinclude Dempster–Shafer[75],theoryofevidenceandrecursiveoperators.
Featurefusion strategies provide an excellent meansto combine heterogeneous sensor data using machine learning algorithms. Ma- chinelearningalgorithmssuchas VectorSupportMachine,Artificial Neural Networks [71], Decision Trees and Hidden Markov Model have beenused to extract featuresfrom different sensor modalities.
Besides,toreducethesolvingtimeandtoselectanoptimumfeature vector, different methods of feature selection have been proposed.
Filter,wrapper, andembeddeddatabaseapproachesaremethodsfor selectingfeatures. Thesehavebeencriticallyreviewed.However,the handcrafted features aretime-consuming andapplication-dependent.
Recently,deeplearningalgorithmssuchasDeepBoltzmannMachine, Autoencoder, Convolutional Neural Networks [76], and Recurrent NeuralNetworks[77]wereproposedfortheautomaticrepresentation offeaturestoreducethedependenceofmanuallyprojectedresources andthetimespentintheselectionofresources.
In[78],theclassificationoffouraffectivestateswasperformedin differentdatasets.Theauthorsusedk-meansastheclusteringalgorithm tovisualizepeople’semotions.Theskintemperature,heartrate,and RMSvalueswereusedforgroupingtheemotionalstateunderhappy, sad, neutral, and stressed categories. This work is an example of unsupervisedlearningthathelpstofindtheintrinsicconstitutionofthe data.In[79],theproposedwearableIoTsystemintroducestheuseof anewdatafusionsensor fordatamanagement. Forthis,theauthors apply aBayesianapproach, makingitpossible todetectandclassify differenttypesofphysiologicalstatesofpeople.
3.4. Mitigatingkeychallenges
Inthissubsection,wepresentabriefdescriptionofsometechnolo- giesappliedtomedicalprocesses.
3.4.1. Energyandsignals
Thetechnologyappliedinhealthcarehasrevolutionizedthepro- cessesofdiagnosisandtreatmentofdiseases,reducingthechancesof errorsandallowinggreateraccuracyinchoosingthemostappropriate treatmentforaspecificpathology.However,similartohowanewdrug isevaluateduntilitobtainsapprovalandgetsreleasedinthemarket,a newdeviceorsystemassociatedwithhealthmustbeexaminedthrough carefulstudiesandsimulationssothatitcanbereeasedasanewre- sourceforthebenefitofdecisionmakersandespeciallythepatients.For instance,foraccessiblemonitoringofcardiachealthviaphotoplethys- mography(PPG),itisnecessarytoguaranteeanaccuratedetectionof thecardiacconditionfromthesignalsextractedfromasmartbandor anyothersimilardevicethatintendstocapturetheparameters.
The presence of noise primarily due to motion strongly affects theresultof thePPGanalysis. Thestudy[80]statesinits proposed methodthatphysiologicalsignalinvolvespreprocessing,specificallythe breakup,andcansubstantiallyimprovetheoverallperformanceefficacy andclinicalutilityasdemonstratedinthecasestudy,whichshowsa significantimprovementinefficiencywhenidentifyingcoronaryartery disease(CAD)fromthePPGsignal.
Heart ratemonitoring using pulsed PPG signals during intensive exercise is a subject studied in [35,36], and [26]. As the signals get corrupteddue toextrememovementdisorders causedbyabrupt hand movements,ZhilinZhang, SeniorMemberof IEEE,in hisstud- ies [35] and [36] proposes and compares with techniques used, a general framework called TROIKA, comprising a signal decomposi- tionforthedismemberment,signalreconstructionspectrum,spectral analysis, andverifiedspectral data analysis. Alreadyin [26], anew techniquewaspresentedtoaccuratelydeterminetheheartrateduring excessivemovementbyclassifying PPGsignalsobtainedfrom smart- phonesorwearabledevicescombinedwithmotiondataobtainedfrom accelerometer sensors. The approach uses theIoT cloud connection fromsmartphonestoPPGsignalselectionusingdeeplearning.
Recently,wearabledeviceswereusedforlong-termmonitoringof ECGsignals.However,thecomfortofmonitoredpatientswasnotsat- isfactory,andthereforeawell-designedtextileelectrodewithexcellent signalqualityandcomfort(SQC)isessentialforacquiringECGsignals.
In[81],inordertoprovideoptimizationoftheSQC,atextileelectrode withmeshstructureandconductivematerialcomprisingsilvercoated
withcotton andnylonfiberisstudied.This textilecompoundisinte- gratedintoablousewithaminiaturizedwirelessdetectionplatformto collectECGsignalsfromthevolunteer.In[82],theauthorsfocusedon theevaluationofthesignalqualityin realtime(SQA)andtheslight detectionofQRSfortheapplicationofavisibleECG.Forthat,thestudy proposesamethodofcombiningseveralSQIsandlearningSVM-based machinetoautomaticallyclassifythesignalqualityofECGsacquired inrest,outpatient,orphysicalactivityenvironments.
Previouslyselected researchesdescribemonitoring thevitalsigns of patients topredict its health status by using applications of the InternetofHealthThings.Inordertoreachthis objective,electronic healthsystemsareprincipallybasedonsensorsintheenvironment.The work[83]proposesamethodtopredictboththeECGsensordataand themostlikelyhealthconditionofthepatient,whichdoesnotneeda commonmethodofactivityrecognitiontopredictthepatient’shealth situation. Theproposed approach provides for future mobile sensor dataandoverallpatient health statususinga semi-Markov(HSMM) modelwithtwooutputs.
Acommonissuewithwearabletechnologyisthatsignalcaptureand transmissionrequirespower,andassuch,devicesoftenrequirefrequent recharging,whichisalimitationtothecontinuousmonitoringofvital signs.Tomitigatethis,[84],and[85]advocatedtheuseoflossysignal compressiontodecreasethedatasizeofthecollectedbiosignalsand consequentlyincreasethebatterylifeofwearabledevices.[86]presents anewwearablewristECG,proposingheartmonitoringwithlowenergy consumptionandcustomizableelectrodesthroughtheuseofsilverpaint appliedonthehousingofthedevice.Thedesignemploysafewactive componentstominimizeenergy.TheECGfrontendisintegratedwithan existingIoTbackendforuseonthewrist,andthedevice’sperformance hasbeencomparedtoothercommercialdevices,showingitsaccuracy toafewbeatsperminute.Assumingmoderateperiodsofuser-activated sampling,thesystemcanlastformorethanamonthusingthesupplied batteryandweighslessthan30gincludingthebatteryandthewrist strap.
3.4.2. Securityandprivacy
ThefactorofBSNscollectingsensitivedataandgeneratingvaluable informationforcaregiversandusersmakesthemattractivetargetsfor technologycriminals.One suchthreatis sensorcompromise,suchas unauthorizedmodificationofthesensoroutput(i.e.,measurement)to relayincorrect patienthealth data tothebase station. Forinstance, tampering with the ECG sensor can have severe consequences for theuserasitmonitorsthecardiacprocess.Onestudy[87]examined attacksassociatedwithhandhelddevices.proposingwaystodelegate usagetoawidevarietyof restrictedhandhelddevicestoensurethe privacyandintegrityoftheirdata.
Traditionalencryptiontechniques aredisadvantageouswhen pre- ventingdataanalysis in acloud environment(unless thedecryption keyshavebeensenttothecloudprovider).In[39],thisdisadvantage isovercomebyimplementingHomomorphicEncryption(HE).Inthis study,theresearchersproposedanewsystemthatanalyzesECGdata, aggregatesdataonthePC,uploadstothecloudserviceprovider(such asAmazonWebServices),examinesdatainthecloudenvironment(and forwardsthedata/resultstothedoctorinquestion),andalsoprotectsthe entireroutewiththehelpoftheHEtechnique.Subsequently,thecardi- ologistrequeststoobservethepatient’sHRVresultforaspecifiedinter- valandtheresultsareextractedfromthecloudserver,decrypted,and displayed.Finally,thecardiologistreviewstheresults,diagnosesthepa- tient,anddecidesonanappropriatetreatmentstrategy.Pulse-basedran- dombinarysequences(RBSs)arethebackboneofvarioussafetyaspects ofwirelessbodysensornetworks(WBSNs).Toimprovetheefficiency of time, a technique of generating biometric RBSs using interpulse intervals(IPIs)oftheheartrateisdevelopedin[88].Accordingtothe experimentalresults,128-bitgeneratedRBSsfromhealthyindividuals andpatientscanpotentiallybeusedaskeystoencryptionorentityiden- tifierstoprotecttheWBSNs.However,thestudyofLiangetal.[89]in-
corporates thesafe andconvenient sharing of personal health data, confrontedwithprivacyissuesandexistingvulnerabilitiesinthestorage ofcurrentpersonalhealthdataandsystems,aswellastheconceptof ownershipofself-patient-basedhealthdatasharingsolutionssupported byblockchaintechnology.In[90],theproposedprocessorreusesthe ECG featuresused during classificationtogenerate a keytoprotect the proposed IoT platform against hardware- and telemetry-based attacks.
Theuseoftechniquestoconcealtheexistenceofamessagewithin anothersuchassteganography,demonstratedby[91]tohidepatient diagnostic information on the ECG signal, based on the intelligent compensation coefficient,provedtobe asafealternative.Toprovide ahighconcealmentcapability,theproposedmethodincorporatestwo bits of data intoa coefficient packetfrom the host ECGsignal ata time.Thesimulationspresentedindicateasuperiorabilitytohide,and besides,themethodistolerantofattackssuchasinversion,translation, andadditionalattacksbyGaussiannoise,whichisrareinconventional steganographicECGschemes.
3.5. OpportunitiesfromtheIoTforheart
Portable devicesandsmartphones areatrend allovertheworld.
Amongtheirusesarestorageofimages,soundsorpersonaldocuments.
Inadditiontothedevice’sinternalstorage,userdatacanbestoredinser- vicesinthecloud,thusfacilitatingaccesstoalllocations.[92]explores theuseofECGrecordstoencryptuserdataandaddressesthischallenge byproposingEbH,amechanismforgeneratingsymmetrickeysbased onECG.
Inrecentyears,theadoptionofbiometricmethodsforauthentication andaccesscontrolhasbeenincreasing,for example,theadoptionof biometricreadingcapturedbythepalmtoaccessbankingdata.In[93], acomputationalmethodispresentedidentifyingapersonusingonly threefeaturesbasedonECGMorphologyfromasingleheartbeat,which aimsforscenariosoflowcomputationalcost.
TheadoptionofIoTinheartcarebringsarangeofnewopportunities inthemanneritaddressesthediagnosisandtreatmentofheartdisease.
Ifpreviouslythedoctorhadtowaitforthepatient’svisit toperform theinitialanamnesis,today,throughareal-timemonitoringnetwork, onecanobtainthepatient’scurrentstatusaswellaspredictextreme situations,forinstance,aheartattackthatcaninmanycasesleadto deathorleavesequelaethatarecostlybothtothehealthcarestructure andespeciallytothepatient.
Researchessuchas[24,94–96],and[55]focusondevelopingsys- temsfordetectingaheartattackwithIoTsupportand[41]proposes anearlypredictionsystemforcardiacarrestusingthesameveinofre- search,thatis,collectingfromsensorssuchasapulsesensor(∗);using anIoTdevicesuchasanArduinoUnoorevenasmartphonetotransmit thedatacollectedbythesensorsoutsidetheBANtopology,(thepatient themself),andindirectsystemscorrelatedwithemergencycareprofes- sionalssothattheycanactonthedetectedhealthincidentinthefastest possibleway.
Presently,avehiclecontainsnumeroussensorsthathelptomonitor itsoperationandifanabnormalityisidentified,thedriverisinformed thatitrequirestechnicalassistanceforathoroughcheckconductedbya specializedprofessionalteam.Thissituationsuggeststhatthemonitor- ingisdonecorrectlyandobeysthestandardsstipulatedinstudiesand throughdetailedanalysishasensuredsubstantialreliabilityintheuse ofthismeansoftransport.Inthehumanbody,thisisnodifferent.Ifone practiceshealthyhabits,performsconstantmonitoring,andisassisted byskilledprofessionals,thelikelihoodoffurtherhealthproblemsinthe futuredecreases.Thisdoesnotmeanthatissueswillneveroccur,similar tohowevenwithmonitoring,avehiclecanundergoanincidentwith apothole,causingitsdisrepair.Therefore,ahealthproblemcanoccur duringthejourneyoflifeevenwithallthemonitoringtechnologyand treatmentswehavetoday;ouraimistoreducetheprobabilitythrough stochasticprocesses.
Studiessuchas[42,48,88,97–102],and[103]elucidatethebenefits ofusingIoTtomitigatediseasesassociatedwithanabnormalheartrate.
IniCarMa[97],anon-demandcardiacmonitoringschemedistinguishes cardiacconditionsusingonlythePPGsignalextractedfromasmart- phonecamera or other sensors andalso classifies abnormal cardiac conditionssuchasasystole,extremebradycardia,extremetachycardia, ventricularflutter,andventriculartachycardiatoindicateseverity.The useoftheECGsignalcapturedbywearablesdevicestodeveloplearn- ing,classification,andpredictionmodelsfortheautomaticdetection ofcardiacarrhythmiasisthepillarofthestudies[98–101],and[88]. TheIOT-BEAT, whichcallsitselfanintelligent nurseforthecardiac patient,[48]proposesasystemthatmonitorsECGwavessimilartothe existingHoltermonitorandalsoattemptstopredicttheoccurrenceof arrhythmia.Inthecasewherethewavesareofnormalsignificance,the heartisfunctioningproperly;oursystemwouldsendaregularreport tothedoctoratregularintervals.Furthermore,in[103],alight-based sensorisusedtocheckheartratevariationsbecausethestudyfocuses onatrial fibrillation,whichis amajorcause ofstroke,heartfailure, and other heart-related complications. Hence, a reliable method of detectionisrequiredtointervenebeforetheconditionworsens.Oneof thechiefgoalsofthisresearchprojectistobuildalow-costPPGsensor, andthisisachievedbyusingacombinationofLEDphotodiodestoemit lightandexaminethereflectedlight.
Inadditiontoheartrate,monitoringofbloodpressurehasbecome the focusin several studies related to the use of IoTfor health as demonstrated in [60,104,105], and[106]. Chronic diseases such as hypertension andhypotension, in the long run, can lead to serious illnesses.They arealready identified andsubsequentlyaccompanied bycarriers ofthe samethroughreadings ofdiastolicblood pressure (the minimum pressure)and systolic blood pressure (the maximum pressure)generatedby sensorsthat can be interconnectedin aPAN networkandlaterintegratedintoamedicalmonitoringstructure.
Theuseofsensorsandelectroniccircuitsarerequiredtoderivethe readingsobtainedfromtheECGinconjunctionwithPPGsignalstocap- turebloodpressurefromasmartphoneasdemonstratedintheprototype developedby[104].Thisstudyillustratesthatalow-costdevicecanbe easilyextended tomonitor cardiaccharacteristics suchaspremature ventricularcontractionandvenouspulsations.Thestudiesof[60,105], and[106] demonstratethat monitoring blood pressure through IoT bringssocialandeconomicbenefitssuchasimprovementinthehealth process that requires interdisciplinary cooperation and coordination amongmedicalteams,clinicalprocesses,andhypertensivepatients.
Theoptimalopportunitytoimprovequalityoflife(QoL)standards hasbeendemonstratedin[105]and[107].Thesestudieshaveadded medicalvaluetoasmartwatchsuchasmakingitpredictingadropin bloodpressure(BP)whichmightcausedizzinessandincreasetherisk offallingintheelderlyaswellasinyoungergroups[105]byproviding suggestionsregardinghabitsconsistentwiththeircurrentphysicalstate.
Theseincludedietsorsportsthroughnon-invasivemonitoringofblood pressure[107].
AccordingtotheEuropeanUnion,SmartCitiesaresystemsofpeople interactingandusingenergy,materials,services,andfundingtocatalyze economicdevelopmentandimprovethequalityoflife.Theseinteraction flowsareconsideredintelligentforstrategicallyusinginfrastructureand servicesandinformationandcommunicationalongwithurbanplanning andmanagementtomeetthesocialandeconomicdemandsofsociety.
TheelectronichealthapplicationsinIoTnetworks,especiallyforintelli- gentmanagementofmedicalresourcesinthecity,havebeenattracting manyresearchers.
In[94],theauthorsproposeanewIoTelectronichealthservicefor detectingreal-timeheartattacks(RHAMDS)throughvoicecontroland gesturecontrolusingsmartclocksinconjunctionwithSoftwareDefined Network(SDN).SDNanetworkparadigminnovativeapproachthatal- lowsnetwork programmingbyseparating thedata planandcontrol plane,therebyprovidingglobalintelligenceforthenetwork.RHAMDS aimstoimproveemergencyresponsetimeforheartattackpatientsin
particularvehiclenetworksandtopreventpossiblecollisionsofvehicles generatedbythisincident.[108]focusesondevelopingareliableand low-costhealthmonitoringsystemforautomotivedrivers.Itisbasedon theprincipleofcontactlessECGandsendingdatatothecloud.Multiple- signalacquisitionECGelectrodesareplacedintheseatandseatbeltof theautomobileandthereceivedsignalsaresenttoawebserverwhich canbeanalyzedbyexpertdecision-makerstoprovideanimmediatere- sponseincaseofemergency.
Remote monitoringis extremely usefulfor hospitalsfacing insuf- ficientinfrastructuretoadmitnewpatients.Studiessuchas[48]and [56] aim to disseminate cardiac care to urban and rural locations using the advantagesof IoTin healthto remotelymonitor the ECG signalsoftheafflictedcardiacpatients,alertingthoseresponsiblefor healthincaseofanemergency,thus solvingtheproblemofstocking hospital structures as the patient could be monitored from his or her residence.Thework[109]presents asmartstethoscope thatcan visualize sound signals from the heart and simultaneously measure theelectrocardiogram(ECG).Thisdevicehasprovedtobesuitablefor elderlypeopleandforpeoplewhosufferfromchronicdiseases.
3.6. LoweringthecostinIoHT
The prototypes presented in the articles reviewed in this survey chosetouselow-costhardwaretoperform theirexperimentssuchas in [110] and[111]. This does not imply that the samecan not be usedinaproductionenvironment;[31,33,112],and[52]proposethe use of their infrastructure presentedfor future adoptionasproducts availableforanassistedlivingenvironment.Forthesesolutionstobe competitive, improvingthem forusabilityin sucha complex health environmentmustbeaccomplishedastheyneedtoprovidehighlevels ofrobustnessforalltheinvolvedresourcesprovided,fromcommunica- tionwithsensorsforautomaticmonitoringtovisualizationandstorage ofsensordata[113].
Forinstance,theobjectiveoftheworkof[110]istocreateadaily monitoringmodulewhichwillallowobtainingahigh-precisionECG, processing it, andtransmittingitthrough theBluetooth LowEnergy (BLE)wirelessinterface.Modernelectroniccomponents(ADAS1000-4 BSTZ,CC2650MCU)areusedinthisdevice.Thepatient’ssignalthrough theelectrodesissuppliedtotheADAS1000-4BSTZECGchipwhich producesscanning,filtering,andsignalamplification.Additionally,the data is sent to themicrocontroller via theSPI wireinterface. Here, thedataisprocessedandtransmittedthroughthepartofthenetwork viatheBLEtotheswitch.TheCC2650waschosen astheMCU.The real-timeoperatingsystemcreatedbyTexasInstruments(TIRTOS)was used asa basisfor writing software forthe CC2650microcontroller (MC). Furthermore, Code Composer Studio 7 (CCS 7) was chosen as a development environment. In[111], theresearchers attempted to adapt various short-range technologies for WBANs in ubiquitous health monitoring. For this, we used a BLE wireless module called BLE112fromBluegigawhichisbasedontheCC2540modelfromTexas Instruments.BLE112canhostcompleteend-userapplicationswithout usinganexternalhostormicrocontrollerandprovideshostinterfaces suchasUARTinapplicationswheretheexternalhostisrequired.
In[112],theauthorproposesatutorialondesigningasmartbracelet tomonitorheartrate.ThemonitoringhardwarecomprisesBLENano, PulseSensor,andabatterymountedinaboxcreatedina3Dprinter.
Heusedalow-costsensoreasilyfoundinthemarketandaspresented inTable002,manyofthesewereidentifiedtohavebeenusedinproto- typespresentedinthestudiesselectedforourevaluation.In[33],the Arduino-basedmicrocontrollerwasusedasagatewaytocommunicate withvarioussensorssuchasthetemperaturesensor,ECGsensor,pulse oximetersensor,etc.Themicrocontrollercollectsthesensordataand sends it tothe networkvia Bluetooth using the Bluetooth interface moduleforArduino.Thisarticleessentially explainsthedetectionof abnormalitiesintheheartthroughaspeciallydesignedECGcircuitand thetransmissionoftherelevantdatatoasmartphoneviaBluetooth.
Unlikemostexistingmethodsthatuseanopticalsensortomonitor heart rate, in [31] the approach is to measure real-time ECG with dryelectrodesplacedinthesmartphonecase.Theproposedhardware comprisesa single chipmicrocontroller (RFduino) built-in withlow power Bluetooth (BLE) which thus decreases the size and extends thebatterylife.Asmartphoneboxwascreatedusinga3Dprinterto validateandanalyzethesystem’sviability.In[52]thedevelopmentofa projectforECGmonitoringwaspresentedthatalsousedasmartphone, butthistime,itwasusedasahubbetweenthesensorsusedtopickup andtreattheECGsignalstosendthemtoacloudserver.
Thestudies[98]and[60]uselow-costsolutionsinIoTtodevelopthe proposalsoftheirresearch.[98]describesthemethodsusedtotrainthe ECGclassifiersandtheirpredictiveaspectbecauseonceabnormalbeats orwaveformsaredetected,theappropriatealarmscanbeissuedand senttothehealthcareagencyinchargeofpatientcare.Foracasestudy, areal-timeintelligentIoT(InternetofThings)systemthatcanbeinte- gratedwiththeGPConnectinfrastructureprovidedbytheNHS,UK,was proposed.AdataanalysissoftwarewasdevelopedwithGCCWFDB,the librarieswereprovidedbytheMITDBBIH,andarrhythmiawashosted onaLinuxIoTdevicethatcouldbeaRaspberryPiorBeagleboneBlack.
In[60],amodelofempowermentofhypertensivepatientsthrough BPM,IoT,andremotesensingwasproposed forwhichawebsystem connectedwithhealthmonitors,environmentsensors,andamobileap- plicationwasutilized.Tocapturetheenvironmentalinformation,apro- totypeofasystemfortemperatureandatmosphericpressuremonitoring wasdeveloped;thesearetwovariablesthataffecthypertensivepatients.
ARaspberryPIv3unitwasusedasahardwareplatformbecauseitisro- bustandhaslowpowerconsumption,whichmakesitsuitabletosupport aservicethatworkscontinuously,allowstheconnectionofsensorssuch astheBPM180thatisusedintheexperiment,andhaswiredandwireless communicationinterfacesrequiredtotransmitmonitoredinformation.
In addition to the acquisition of low-cost hardware, many open sourcesolutionsforthedevelopmentofwebapplications,orAppscre- atedforthearchitectures,proposedintheselectedstudieswereused;
see[56,60,114].In[60],tosupportthelifecycleoftheproposedBPM model,theBonitaCommunityEditionversion7.4softwarewasused.
Bonitaisamodular,opensourceBPMSthatimplementsBPM’scomplete lifecycle.Withthistoolandincollaborationwithamedicalteam,all proceduresassociatedwiththeclinicalprocessoftheproposedredesign weredesigned.Forwebdevelopment,AngularJSwasusedwhichisan open-sourceJavaScriptframeworkmaintainedbyGoogle[60].
In [114], the prototype includes an alarm system that sends a messagetothephoneapplicationand,whenusingsocialnetworkssuch as Twitter, timelynotifies the attending physician andthe patient’s relatives whenever a critical level in the patient’s body signals is detected.TocreateanAndroidOSapptoviewflowdataandmanage alert messagesatcritical levels,MITApp Inventor 2was used. App Inventorisbasedonablockdiagramprogrammingthatallowsoneto manage different aspectsof the mobile devicesuch as notifications, webaccess,calls,andmessages.Emergencynotificationmanagement is possible due to the Twitter DeveloperApp linked tothe created app.Thus,whenanemergencywarningispostedtoaTwitterwall,an audiomessageandthedisplayofemergencyoptionsalsoappear.App Engine,aplatformforcreatingWebapplicationsandmobilebackends withautomaticescalation,wasusedinthetelemedicalhealthsystem, developed by [56]and based on a mobile communication platform calledECGforEverybody,fortheimplementationofthewebplatform.
To address the needs generated by increasing the variety and quantity of data generated by IoT-basedhealth monitoring devices, researchershave begun touse large data and NoSQL(Unstructured QueryLanguage)technologies.TheIoTframeworkproposedby[56]for earlydetectionofheartdiseaseusestheopensourcedatabaseNoSQL ApacheHBasetostorethehugedatavolumeofthesensors.Thesewere usedinthecloudwhichwasalsousedtoproposeasystemthatwill detect heartattacks withthe helpof different aspects,for example, monitoringheartrateandsmartbloodpressure[95].
IoTisascenarioin whichmost devicesareresourceconstrained, which meansthat intelligence will be delegated to a more capable entity.ThisentityisasoftwareidentifiedasIoTmiddlewareorIoTmid- dlewareplatform[54].Forinstance,studiessuchas[44]and[115]used theAmazonWebServices(AWS)IoTCloudandin[115]theMicrosoft AzureIoTSuite.
4. Discussionandreferencemodel
Thissectionpresentsadiscussionandcomparativeanalysisofthe proposals presented by theauthors of theresearch reviewed in this survey. Moreover,those thatweretested byvolunteersor simulated only fromdatacaptured inresearch baseswerealsoscored.Table 1 summarizes all the points covered in Section 3 and scored in the selected articles. This table includes signs, communication whether machine learningalgorithmswere used, whethercardiovasculardis- easeswereaddressed,howtheycollectedsignalsforvalidationofthe presentedproposal,andwhetherthecomputationalresourcesusedin theexperimentweredescribed.
As the proposal of this survey is also intended tobe a support material, we present a reference model based on the evaluation of the resources used by the authors of the studies selected in this research. This model aims to help future enthusiasts discover and subsequentlyenumeratethefactorsnecessaryforthedevelopmentof aprototypeforonlineheartmonitoringpurposes,includingfacilitating theidentificationoftheresourcesrequiredforfuturestudy.
4.1. Discussion
Next,wewilldiscusspointsforthedevelopmentofaprototypeto presenttheconcept,model,orsolutiontoaproblemthatwasdelimited intheresearchselectedinthissurvey.Tables2–4,asshowninFig.2, summarizethetechnologicalresourcesofthestudiesthatrespectively emphasized prototyping,applicationdevelopment,andhomologation environment.
Thefollowingconclusionscanbedrawnregardingtheresourcesused bytheauthors:
• Theauthorsemphasizedusinglow-costcomputingresourcessuch asthePulseSensorforsignalcollection.Forprojectsinvolving setsofhardwarelogicandnotrequiringtheuseofrobustsoftware behind,theArduinoplatformwasthemostchosen.RaspberryPi cards,whichcanalsoworkinapplicationsinvolvinghardware logic,havebeenexploredincasesrequiringoptimizedsoftware solutionssuchasinmachinelearningalgorithms;afterall,these arecompletecomputers.TheIoTplatformIntelGalileowasused promptlyinthestudyof[101].
• We have found that the integration of wearable sensors into mHealth applicationsisstillbeingexploredby researchersfor theregularuseofactivitytrackersandbiosignalsintheirexper- iments.Theuseofthesmartphoneassociatedwithasmartband suchastheMIBAND2thatjointlyformsaPersonalSensorDe- vice(PSD)wasusedinsomestudiesthataimedtodemonstrate theintegrationofpersonalhealthdevicesconvergedwithatten- dantnetworks,asillutrsatedin[120].
• Asemphasized inthis paper,theuse oflow-cost resourcesby theauthorswasevident,becauseoneoftheaspectsthatarefre- quentlyemphasizedintheadoptionofIoTinhealthisthepos- sibility ofreducing costs.Wecorrelatethis totheapprovalof Androidasanoperatingsystemusedintheproposalspresented bytheauthorsusingmobileapplicationsaspartoftheirresearch, suchas[31,52,58,96,114,122],and[37],andeventhoseusing platformssuchasMITAPPInventor2fordevelopment.However, foramorecomprehensiveadoptionofaprojectrelatedtotheuse ofamobileapplication,thedevelopmentofaniOSversionshould beconsidered.