Feasibility of photoacoustic/ultrasound imaging of synovitis in finger joints using a point-of-care system

(1)

Research

article

Feasibility

of

photoacoustic/ultrasound

imaging

of

synovitis

in

ﬁnger

joints

using

a

point-of-care

system

Pim

J. van

den

Berg

a

,

Khalid

Daoudi

b

,

Hein

J. Bernelot

Moens

c

,

Wiendelt

Steenbergen

a,

*

a_Biomedical_Photonic_Imaging,_MIRA_Institute_for_Biomedical_Technology_and_Technical_Medicine,_University_of_Twente,_PO_Box_217,₇₅₀₀_AE,_Enschede,_The

Netherlands

b

MedicalUltrasoundImagingCenter,departmentofRadiology,RadboudUniversityMedicalCenter,POBox9101,6500HBNijmegen,TheNetherlands

c

ZiekenhuisgroepTwente,DepartmentofRheumatology,Postbus546,7550AMHengelo,TheNetherlands

ARTICLE INFO

Articlehistory: Received29March2017

Receivedinrevisedform21July2017 Accepted28August2017

Availableonline31August2017 Keywords: Rheumatoidarthritis Synovitis Photoacoustic Optoacoustic Medicalultrasound Echography Ultrasonography Proofofprinciple Feasibilitystudy ABSTRACT

Weevaluateaportableultrasoundandphotoacousticimaging(PAI)systemforthefeasibilityofa point-of-careassessmentofclinicallyevidentsynovitis.Inflamedandnon-inflamedproximalinterphalangeal jointsof10patientswereexaminedandcomparedwithjointsfrom7healthyvolunteers.PAIscans, ultrasoundpowerDoppler(US-PD),andclinicalexaminationwereperformed.Wequantifiedtheamount ofphotoacoustic(PA)signalusingaregionofinterest(ROI)drawnoverthehypertrophicjointspace.PAI responsewasincreased4to10foldwhencomparinginflamedwithcontralateralnon-inflamedjoints andwithjointsfromhealthyvolunteers(p<0.001forboth).US-PDandPAIwerestronglycorrelated (Spearman’sr=0.64,with95%CI:0.42,0.79).Hence,PAIusingacompacthandheldprobeiscapableof detectingclinicallyevidentsynovitis.ThismotivatesfurtherinvestigationintothepredictivevalueofPAI, includingmultispectralPAI,withotherestablishedmodalitiessuchasUS-PDorMRI.

(http://creativecommons.org/licenses/by/4.0/).

1.Introduction

In rheumatoid arthritis (RA), imaging of synovitis with ultrasound power Doppler (US-PD) and magnetic resonance imaging(MRI)canpredictdiseaseprogressionandboneerosion

[1–3]. In clinical remission, detection of subclinical synovitis indicatesdiseaseprogressionandincreasestheriskofdiseaseﬂare

[4–7].US-PDhasgainedaplaceintheclinicalworkflowbasedon these qualities. However, US-PD has inherently high operator dependencyandsuboptimal reproducibility[8,9].Specific com-plicationsofUS-PDareitsdependencyontheanglebetweenthe flowvectorandthesoundbeam,andthedisturbanceoftheblood flow by the probe pressure. MRI is rather costly, specificity is modest and it requires contrast agents [10]. Optical imaging methodswerestudied inrecent years aspotentialalternatives. Opticalspectral transmission(OST) forexample,hasshown fair performanceatdetectingsynovitiswhilebeingpresumablylowin cost[11–13],however,sensitivityandspecificityaremodestand thelowspatialresolutionlimitsdifferentiationbetweensynovitis andtenosynovitis.Fluorescenceopticalimaging[14–17]appearsto

havehigherperformancethanOST,butalsohaslowresolutionand inadditionrequiresinjectionofcontrastagents.

Photoacousticimaging(PAI),ahybrid optical-and-ultrasound imaging technique, may offer a good balance in features, combining thesensitivityto haemoglobinof opticaltechniques withtheresolution ofclinicalultrasound[18–21].To formaPA image,shortlaserpulsesareshoneontheskinandsubsequently enterthetissue,wherethelightisscatteredbycellsandbecomes diffuse.Thelightpulseisthenabsorbedbydarktissueconstituents suchashaemoglobinandmelanin.Theabsorptionslightlyheats structurescontainingtheseconstituents whichleadstoa small pressurebuild-up,generatingsoundwavesthatcanbepickedup byultrasoundtransducers.PAIisthereforesimilartosonography, exceptthattheultrasoundisgeneratedwithintissue,insteadof reﬂected('backscattered')byit.

PAI differs significantly from US-PD in three aspects. First, movementoferythrocytesisnotrequiredforsignalgeneration, sincethegenerationofPAsignalsreliesonlyonthepresenceof haemoglobin (or other chromophores) [19]. Second, there is a largerconcentrationofhaemoglobinwithinvasculaturethanin surroundingtissue,leadingtomoresignalgeneration,whereasin US-PD, erythrocytes reflect comparatively less signal than the surroundingtissue[22,23].Awallfilteristhereforenotrequiredin *Correspondingauthor.

E-mailaddress:w.steenbergen@utwente.nl(W.Steenbergen).

http://dx.doi.org/10.1016/j.pacs.2017.08.002

Photoacoustics

(2)

PAI,and‘ﬂash'artefactsormotionclutterarenotpresent.These propertiesimplythatslowbloodﬂowinsynovialmicrovasculature poses no problem to PAI. As a result, we expect PAI to be particularlysensitivetosubclinicalsynovitis.Finally,thePAIsignal islessaffectedbytheorientationofthebloodvesselthanUS-PD. PAI has been investigated in other medical areas involving angiogenesis,forinstanceinclinicalstudiesintomammography

[24–27].PAIhasalsobeeninvestigatedinpre-clinicalstudiesof synovitis [28–31], and several setups have been proposed for humanﬁngerjoints[32–37].In addition,a fewearly feasibility studieshavebeenperformedwithRApatients[38,39].However, these studies used large lasers, not suited for routine clinical application,letalonepoint-of-careimaging.

InordertobringPAI tooutpatientclinics,a handheldPA/US probewasdeveloped[33],whichinthisstudyisinvestigatedfor possibleuseinassessingsynovitis.Theobjectiveofthisstudyisto investigatewhetherthisPA/USprobecandetectclinicallyevident synovitisandtocomparetheresultswithUS-PD.

2.Methods 2.1.Patientinclusion

Patients undergoing care in the Ziekenhuisgroep Twente hospitalwereaskedbytheirrheumatologisttoparticipateinthis study.Healthyvolunteerswererecruitedinpersonorviaﬂyersat theUniversityofTwente.

Patientsagedover18yearswithrheumatoidarthritisfulfilling6 or more ACR/EULAR criteria (ACR/EULAR=American College of Rheumatology/European LeagueAgainst Rheumatism) were in-cluded[40].Specificinclusioncriteriawere:swellingofatleastone proximalinterphalangeal(PIP)joint,2,3or4jointswithatleast grade1 power-DopplersignalonUSexamination.Test subjects (healthyorpatient)wereexcludedfromparticipationiftheyhad clinicallysignificantbonedeformationand/orosteoarthritisinthe jointofinterest.Allsubjectsreceivedwritteninformationandgave informedconsent,resultinginadelayof3to8daysbetweenthe inclusionbyarheumatologistandtimeofmeasurement. 2.2.Imagingsystem

The imaging study is performed using a dual modality photoacoustic/ultrasoundsystem. The system relies ona probe that houses both a small diode laser together withultrasound transducers (see Fig.1).The diode laser is pulsed to generate photoacousticwaves,whicharethendetectedbytheultrasound transducers. Thesetransducers are alsoused totransmit ultra-soundtogenerate high-qualityb-mode ultrasound images.The probein thisstudyis asecondgenerationprototypedeveloped fromtheprobedescribedearlierindetail[33].Theoriginalprobe

contained diode lasers producing 130ns pulses at a 805nm wavelengthandapulseenergyof0.56mJ.Aswillappear,themain changeisadoublingofthepulseenergy.

The diode laser source (Quantel Laser, les Ulis, France) is controlled bya short pulselaser driver (Brightloop Converters, Paris,France)and generates1mJ pulsesof120ns duration.The pulsesareformedintoarectangularshapeof2.2mmby17.6mm (1/e2₎ _by _a _diffractive _optical _element _(SILIOS _{Technologies,} Peynier, France),afterwhich thelightexitstheprobe underan angle via a prism. The laser emission is at 808nm, which corresponds to the isosbestic point of oxy-haemoglobin and deoxy-haemoglobin, which leads toPAsignalamplitudes inde-pendentofthebloodoxygenation.

TheultrasounddetectionisbasedonanESAOTESL3323probe. Transducersareplacedinanarrayof128elements.Eachelement hasabandwidthfrom2.5MHzto10MHzwitha7.5MHzcentre frequency.Anacousticlens(focallength:24mm)isplacedinfront of the transducers to moderately focus the detection in the elevationalplane.

The probe is connected to a MylabOne ultrasound scanner (ESAOTEEurope),whichcanbeusedintwomodes.Intheﬁrstit transfers the collectedtime-pressure data from the middle 64 elements directly to a laptop. This mode is used to acquire photoacoustic data. In the second mode the scanner operates regularlyandisusedtoacquireb-modeultrasoundusingall128 elementsinaline-by-linetransmissionandacquisitionscheme.

TheUS-PDexaminationisdoneusinganidenticalMylabOne scanner(inthesecondmodeasdescribedabove)incombination witha14MHzcentrefrequencylineararray(SL3116,ESAOTE).The PRF wassetat 750Hz, and thewall ﬁlteratitslowest andthe sensitivityatitshighestsetting.

2.3.Scanprotocol

Persubject examination,a minimumof twoPIP jointswere scanned:oneclinicallyinflamedjointandanuninflamedjoint– preferablythesamejointcontra-lateral.Acompleteexamination ofonesubjectincludedaseriesoflongitudinalimagesusingpower Doppler ultrasoundforeach applicablejointand anotherseries usingthePA/USsystem.Bothexaminationstookplacewiththe subject’sarmplacedinawaterbathfittedwithsupportsforthe arm, handand thefingertobescanned(see Fig.1).Thewater temperaturewascontrolledto29–31_C_during_the_examination. DuringmeasurementstherewasnocontactofthePA/USand US-PDprobeswiththeskininorder toavoidpressureartefacts.In addition,thePA/USprobewasplaced4–5mmfromtheskinsuch that the laser beam intersects with the ultrasound elevational planeattheskinsurface.

ForthePA/USexaminationthePA/US probewas placedona motorizedstageforbettercontrolofthemeasurement.Theprobe

Fig. 1.ThePA/USprobe(left)withviewofthefrontendshowingthelightdeliverywindow(darkaperture)andacousticlensinmediumgray.Thepatient’shandissubmerged inwater(right)whereitrestsonaseriesofsupports.Theprobeismountedona2-axismotorizedstageandpositionedabovethejoint.

(3)

wasalignedlongitudinaltothefingerandonthedorsalside.The stagewasmovedorthogonaltothefingerin0.5mmstepsforover 6mm.Ateachstep,aPAimagewasaccumulatedover500laser pulsesfor0.25s.Takingintoaccounttheangleofincidenceof52 withtheorthogonalontheskinandthebeamsizeof2.2mmby 17.6mm,thelightexposureis3.2mW/cm2_,_which_is_below_the_IEC 60825-1safetylimitof5mW/cm2_for_this_wavelength_and_pulse train.Inaddition,100framesofplanewaveultrasound(onefixed angle)wererecordedeachstep.Eachscanwasrepeatedwiththe sameprobeandatidenticalsteps,butthenwithhigh-quality line-by-lineb-modeultrasound.Onescanyieldedtherefore13PA,13 planewaveand13b-modeimagesatidenticallocations.In our scan protocol there was approximately 1min between a PA acquisitionandthesubsequentb-modeUSimage.

US-PD examinationwaseitherperformedbyanexperienced rheumatologistorbyplacingtheUS-PDprobeinthemotorized stage.Foreachjoint,3–5imagesarerecorded.

2.4.ScoringofUS-PDimages

Representative US-PD images were digitally stored and anonymized. They were graded (0–3) according to Szkudlarek et al. [40] by two rheumatologists who were blinded to the allocation of the images. The widely used semi-quantitative gradingsystem is basedonvisual assessment ofblood flow as indicatedbypower-Dopplersignals:nosignals(score0),upto3 singlevesselsignals(score1),confluentvesselsignalsinlessthan halfoftheareaofthesynovium(2)orvesselsignalsinmorethan half of the area of the synovium (3). Discrepant results were reviewedtoreachconsensusresultinginafinalPD-scoreforeach individualjoint.

2.5.Dataanalysis

The PAchanneldata–thepressureasa functionoftimeas measuredbythe transducers–is convertedinto a mapofthe originalpressuredistributionusingaFourierdomain reconstruc-tionalgorithm[41].Forthisreconstructionalgorithm,wefound anaxialresolutionof0.2mmandalateralresolutionof0.4mm

[33].Thealgorithmwasselectedforitscomputationalspeed.All dataanalysisisautomatedusingMatlab(Massachusetts,USA).To account for the light attenuation within tissue, a depth-dependentcorrection('gain')is applied.Sincetheﬁngerinthe longitudinalorientationisfairlyﬂat,abasicexponentialgainof 1=expð

m

effzÞisusedwith

m

eff=1/mmtheeffectiveattenuation coefﬁcientandzthedepthintissue[42,43].Adifferentz=0isset foreveryaxiallineinthePAimage,suchthattheﬂuencecorrection startsattheskinlevel.Determiningthepositionoftheskinsurface wasdonevisuallyusingthePAresponsefromthemelaninlayerin theskin.

Forimageformation,thePAdataiscompressedlogarithmically atadynamicrangeof40dBor18dB,withthesameminimumand maximum amplitude for inflamed and non-inflamed images. These dynamic ranges were selected based on the noise level (40dB) and the amplitude of healthy joint’s background PA signals(18dB)respectively.Pixelswithinthedynamicrangeare colorcodedinMatlab’sred-and-yellowcolormap‘hot’andfinally overlaidonab-modeultrasoundimage.

Foreachjointscan,aregion-of-interest(ROI)isdrawntoselect the hypertrophic joint area. The ROI is drawn on the b-mode ultrasoundimage,wherethehypertrophic areais deﬁnedas to includeanypixelsbetweenthetendonandthebonesurface.The ROIisthentransferredtothePAimage,fromwhichthenumberof PApixelsiscalculatedthatfallwithinthe18dBdynamicrange.A secondary quantiﬁcation metric is provided by the mean

amplitudeofnon-compressedPAsignalswithintheROI.Incase ofhealthyjointsthereisnohypertrophicareaandtheROIselection willincludemoretissuesthanjustthesynovialspace.

2.6.Statisticalanalysis

Mann-WhitneyU-test (left-sided)is used for comparingthe control group (either joints from healthy volunteers or non-inflamed joint from the same subject) with inflamed joints. Spearman’s rank correlation is used when comparing the PD gradingwithPAquantification.

3.Results

3.1.Subjectcharacteristics

7healthyvolunteersand10RApatientswereincludedinthe study.AllsubjectshadCaucasianskin.Thecharacteristicsofthese subjectsareshowninTable1.TheRApatientshadameandisease duration of 117 months (range 5–133), all were positive for rheumatoidfactorsand7werepositiveforanti-cyclic-citrullinated proteinantibody(anti-CCP),andthemeanC-reactiveprotein(CRP) levelspriortothemeasurementwere6.3(SD5.6).

3.2.Photoacoustic/ultrasoundimaging

Fig.2depictsexamplesoffluencecorrectedPA/USandUS-PD imagesforaninflamedjointandthecontra-lateralnon-inflamed jointof anRA patient.The reconstructedPAsignals areshown rangingfromdarkred(lowsignalamplitudes,startingat40dB) tolightyellow(high/abnormalsignalamplitudes,upto0dB);the dataisoverlaidonthegrayscaleUSb-modeimage.ThePAimages inFig.2Ashowasuperficialbloodvesselinboththeinflamedand non-inflamedjoint,withadditionalPAfeaturesunderneath,above thebonesurface.Largeramplitudesandmoreconfluentfeatures arerecordedfortheinflamedjoint,ascanbefurtherobservedin

Fig.2Bwhere only highamplitudes (18dBdynamic range) are plotted.Withthisthreshold,almostnoPAfeaturesarevisiblefor thenon-inﬂamedjoint.

3.3.QuantiﬁcationofPAandUS-PDimaging

ThenumbersofhighamplitudePApixels(suchasthosevisible inFig.2B)werecomputedforinflamedandnon-inflamedjoints, andofjointsfromhealthyvolunteers.Theresult(Fig.3A)indicates alargernumberofhighamplitudePApixelsforinflamedjoints, compared to healthy and non-inflamed joints. In addition, an alternativequantificationmethodforPAIalsoshowsalargervalue forinflamedjoints:themean(non-compressed)pressure ampli-tudeofPAfeatures(Table2).Bothquantificationmethodsshow4 to10-foldincreasedcounts(p_<0.001)whencomparingin_flamed jointswiththosefromcontrolgroups.Notealsothatthefingersare swollen:thesizeoftheROIasdrawnonthegrayscaleUSimagesis signi_ficantly larger in in_flamed joints compared to healthy (p<0.001) and compared to non-inflamed joints (p <0.05). Grading of US-PD images shows a strong agreement (

r

=0.64, 95% CI: 0.42, 0.79, p<0.001) of the PA pixel count with the Table1

Subjectcharacteristics.

Characteristic Healthyvolunteers RApatients (N=7) (N=10) Age:mean(range) 56(49–62) 63(49–80) Gender(%female) 43% 50% Valuesarethesubject’smean(standarddeviation,SD)or(range).

(4)

consensusPDscoreassignedtotheimagesbytworheumatologists (Fig.3BandTable3).

Toobtainanearlyimpressiononthediagnosticaccuracyofthe method,ReceiverOperatingCharacteristicshavebeenconstructed for themean PA amplitude in the regions of interest, and the number of high amplitude PA pixels, given in Fig. 4A and B, respectively.Separatecurvesandareasunderthecurvearegiven forinﬂamedjointsvs.non-inﬂamedcontralateraljointsinpatients, andvs.jointsinhealthysubjects.

4.Discussion

WefoundthatPAI–inthefirststudywithahandheldcombined photoacousticprobe–wassensitivetoclinicallyevidentsynovitis as demonstrated by the significant difference in PA features Fig.2. PA/USandUS/PDimagesofaninflamed(upperrow)andnon-inflamedcontra-lateraljoint(bottomrow)ofanRApatient.PA/USimagesin(A)showadifferencein colorbetweeninflamedandnon-inflamedcorrespondingtoanincreaseinamplitudelevels.WhendiscardinglowPAamplitudesin(B),onlyfeaturesintheinflamedjointare visible.CorrespondingUS-PDimagesareshownin(C).ThebluelineinthePA/USimagesindicatestheROIusedforquantificationofPAfeaturesinthesynovialspace.The0dB levelisthemaximumPAamplitudefromtheinflamedjoint.d=dermis;dv=dorsalvein;pp=proximalphalanx;pip=proximalinterphalangealjoint;mp=middlephalanx; s=synovium;t=extensortendon.

Fig.3. PAquantiﬁcationwith(A)comparingthenumberofhighPApixelsforeachjointgroupand(B)comparingthesamequantiﬁcationfordiscretePDscore(0,1,2or3, offsetonthex-axisistovisualizeindividualmarkers);Spearman’sr=0.64(95%CI:0.42,0.79),p<0.001.Onetrianglerepresentsonejointandhorizontalbarismedianofone group.

Table2

PDscore,PAquantiﬁcationandhypertrophicarea(ROIsize).

Parameter Healthy Non-inflamed Inflamed (N=12) (N=11) (N=11) PDscore 0.1(0.3)*** 0.5(0.7)** 1.7(0.9) NumberofhighPApixels 225(299)*** 444(694)*** 2792(1742) MeanPAamplitude 13.2(4.4)*** 14.9(11.7)*** 56.7(36.0) ROIsize(pixels) 4540(1318)*** 7900(3690)* 12468(4554) Quantificationvalues:mean(standarddeviation).Ranktestp-valuesfortesting inflamedjointsversuseitherofthecontrolgroups (healthyornon-inflamed): ***p<0.001,**p<0.01or*p<0.05.

(5)

between inflamed and control joints. In addition, the PA quantificationagreedwellwiththecorresponding semi-quantita-tivePDscores.TheROCsandareasunderthecurverevealagood separation of photoacoustic image characteristics between in-flamedandnon-in_flamedjoints.Thisobservationmustbetreated with care, because of the small size of the study and the methodological limitations discussed below. Nevertheless, the resultsdoencouragefurtherresearchinphotoacousticimagingof earlyinflammations.

Hyper-vascularization and angiogenesis are hallmarks of rheumatoid arthritis and are markers for imaging with US-PD, astheincreaseinbloodflowisdetectableusingultrasoundflow imaging. In joints that are close to the skin, the increase in vascularityisanattractivetargetforPAI.Itshouldberealized,that US-PD and PAI do not provide an identical representation of vascularity,synovialorotherwise.Ononehand,US-PDisexpected tohighlight largerfeedingvessels and,theoretically,the move-mentofotherstructuressuchasvilloussynovialfoldswithinthe hypertrophic region. On the otherhand, PAI is expected tobe particularly sensitive to increased blood volume in smaller vasculaturewithinthesynovial membrane. PAI typically works bestforsmallvessels,networkedmostlyparalleltotheprobe; US-PDrathervisualizeslargevessels,angledtotheprobe.Theunique photoacoustic probe that we used in this study is sensitive to vessels, or vascular networks of 0.2mm in size and larger. Interestingly,theappearanceofsynovitisin PAIis quitesimilar foralltheclinicallyinflamedjointsthatwereimagedinthisstudy unlike that of the US-PD representation, which varied considerably.

These fundamental differences betweenPAI andUS-PD may help explain the variation between the PD score and the PAI

quantiﬁcation(Fig.3B).Thereareafewdatapointsthatfalloutside the‘natural’spread:Fig.3Bshowstwograde1jointswithavery highphotoacousticsignalandthreegrade1–2jointsthathardly showaPAIsignal.Theformer(“toohighPAIsignal”)mayoriginate from a different source, as the shape of these corresponding structures was decidedly different from the regularly seen representation of the synovium in PAI. The latter offsets (“too lowPAIsignal_”)mayinfactbeduetofalsepositivePDscoringa resultofartefacts:notesfromoneofthetwoblindedexaminers conﬁrmthispossibility.

WhilethisworkprovidesevidenceofPAIdetectingsynovitis, thereexistafewmethodologicallimitationstothisstudy.Forone, theselectionofpatientstookplaceapproximatelyaweekbefore thePAexamination.ThismayexplainpartlythevarianceinthePA quantification of inflamed joints (Fig. 3A), as some patients’ synovitissubsidedafterselection,butwerestillincludedinthe inflamedgroup.Inaddition,US-PDishardtostandardize,which may have caused the PD artefacts explained earlier. Also, the researcherinchargeofdrawingtheROIswasnotblindedtothe jointinflammation,whichmayhavebiasedtheinterpretation.This issuewasmoderatedhowever,sincetheROIwasdrawnontheUS imagewithoutshowingthePAoverlay.Atechnicallimitationofthe systemwastheinabilitytoco-acquirehigh-qualityb-modeandPA images. The short delay between both may have resulted in inaccuracy due to accidental movement of the finger. This limitationofoursetupwillbesolvedinafutureversion,leading toalmostsimultaneousacquisitionofPAandb-modeUSimages. Despite these limitations this study shows positive and highly significant findings in PAI. Fluence correction appeared to be necessaryin ouranalysis.Variations of theappliedexponential Table3

PDscoreversusotherparameters.

Parameter PD-0 PD-1 PD-2 PD-3

(N=19) (N=7) (N=6) (N=2)

NumberofHighPApixels 252(367) 2368(2494) 1909(1219) 2741(472) MeanPAamplitude 12.2(4.1) 43.8(39.8) 50.1(38.8) 53.6(6.5) ROIsize(pixels) 5263(2115) 11075(5265) 12162(3868) 14013(4445) Quantiﬁcationvalues:mean(1s).

Fig.4. ReceiverOperatingCharacteristics(ROCs)forthemeanPAamplitude(A)andthenumberofhighPApixelsexceeding18dB(B)withintheregionsofinterest.Separate comparisonsandareasunderthecurvearegivenofinﬂamedjointswithjointsinhealthysubjects(‘healthy’)andcontralateraljointsinpatients(‘control’).

(6)

fluencedecayrateinarealisticintervalaroundtheassumedvalue of1/mm,hadnocriticalinfluenceontheoutcomeofouranalysis. This is the first clinical study with a compact and fully integratedPA/USimagingprobe.Itmeansanimportantstepfrom existingPAIsystems,wheresizableandcostlyexternallasersare used,towardspracticaluseinclinicalsettings.Furthermore,our systemreliesonanearinfrared(NIR)lightsourceat808nm,in contrasttopreviousstudies,which usedvisiblelightof 580nm

[44].WhilehaemoglobinabsorbslessNIRlightthanitdoesinthe visiblerange,lightattenuationinthesurroundingtissueis also lower.ThismeansthatwithNIRlightthePAoutcomedependsless on the exact tissue composition. In addition, absorption by superﬁcial structures would be much more pronounced with visible light, for instance in the melanin layer and of regular vessels.Absorptionlikethisisknowntocausepronouncedclutter when these PA signals also travel down and reﬂect on lower structures.

Previousstudiesshowedthatlineararray-basedsystemssuch as used in this study are susceptible to clutter and reﬂection artefacts [45]. Future studies should therefore include clutter reductionandartefactremoval[46,47].Wewereabletorejectthe possibilityofmosttypesofartefactsbymovingtheillumination positioninrelationtotheﬁnger–formosttypesofartefactsthe appearanceofPAfeatureswouldmoveinrelationtotheUSimage

[48], but this did not happen in the cases investigated here. However,cluttermayhavecausedthebaselinePAsignalascanbe seeninFig.2,andalsosomeoftheoutlyingdatapointsinFig.3. FutureapplicationsofPAItosynovitiscantakeadvantageofits multi-spectralimagingcapabilities,allowingtheestimationofthe oxygenationsaturation(sO2)ofthesynovium.Multi-spectralPAIis expectedtoimprovethespecificityofthetechnique.TargetedPA contrast agents [49] with specific spectral signature linked to molecular markers also deserve investigation, as they could provideinformation aboutinflammationsimilartofor example positronemissiontomography.Thenext prototypeofourprobe includesdiodelasersofvariouswavelengthsforthispurpose.This prototypemeritsfurtherinvestigationofsubclinicalsynovitisina largerpatientpopulation, and itspredictive valuefor a disease flare.Inaddition,comparisonwithMRIangiographywillallowa closerlookatwhich specificvascularstructuresaredepictedby PAI. A current limitation of the handheld probe is its low penetrationdepth(15mm)comparedtootherPAIsystems,which meansfutureapplicationswilllikelyfocusonperipheraljointsthat areclosetotheskin.

5.Conclusion

PAIisauniquemodalityduetoitsopticalimagingcontrastin combinationwithultrasound-based resolution. Wehaveshown that PAI with a handheld probe can detect clinically evident synovitis,whichis a_firststeptowardtheapplicationof PAIfor diagnosis and monitoringof inflammation in peripheral joints. Theseresultsprovideabasisforfurtherresearchtoinvestigatethe potentialbene_fitsofPAIoverothermodalities.

Contributions

Allauthorstookpartintheconceptionanddesignofthestudy. PJB,KDandHJBMperformedthemeasurements.PJBprocessedand analysedthedataandwrotethemanuscriptdraft.HJBMtookpart ingradingtheUS-PDimages.Eachauthortookpartineditingthe manuscript,readandapproveditsﬁnalversion.

Patientconsent

Writteninformedconsentwasobtainedpriortoinclusion.

Funding

Theresearchleadingtotheseresultshasreceivedfundingfrom theEuropeanCommission’sSeventhFrameworkProgramme(FP7/ 2007-2013)undergrantagreementno.318067,andtheEuropean H2020programundergrantagreementno.731771.

Ethicsapproval

TheethicalcommitteeMETCTwentegaveitsapprovalofthe studyprotocol.

Conﬂictofinterest

Theauthorsdeclarethattherearenoconﬂictsofinterest. Acknowledgements

WethankDr.CeesJ.Haagsmaforhissupportandgradingofthe US-PDimages.

References

[1]P.G. Conaghan,P.O’Connor,D. McGonagle,P. Astin,R.J. Wakeﬁeld, W.W. Gibbon,etal.,Elucidationoftherelationshipbetweensynovitisandbone damage—arandomizedmagneticresonanceimagingstudyofindividualjoints inpatientswithearlyrheumatoidarthritis,ArthritisRheum.48(1)(2003)64– 71.

[2]P.P.M. Reynolds, C. Heron,J. Pilcher, P.D.W. Kiely, Prediction of erosion progressionusingultrasoundinestablishedrheumatoidarthritis:a2-year follow-upstudy,SkeletalRadiol.38(5)(2009)473–478.

[3]T.Funck-Brentano,F.Gandjbakhch,F.Etchepare,S.Jousse-Joulin,A.Miquel,C. Cyteval,etal.,Predictionofradiographicdamageinearlyarthritisby sonographicerosionsandpowerdopplersignal:alongitudinalobservational study,ArthritisCareRes.(Hoboken)65(6)(2013)896–902.

[4]F.Gandjbakhch,P.G.Conaghan,B.Ejbjerg,E.A.Haavardsholm,V.Foltz,A.K. Brown,etal.,Synovitisandosteitisareveryfrequentinrheumatoidarthritis clinicalremission:resultsfromanMRIstudyof294patientsinclinical remissionorlowdiseaseactivitystate,J.Rheumatol.38(9)(2011)2039–2044. [5]F.Gandjbakhch,E.A.Haavardsholm,P.G.Conaghan,B.Ejbjerg,V.Foltz,A.K.

Brown,etal.,Determiningamagneticresonanceimaginginﬂammatory activityacceptablestatewithoutsubsequentradiographicprogressionin rheumatoidarthritis:resultsfromafollowupMRIstudyof254patientsin clinicalremissionorlowdiseaseactivity,J.Rheumatol.41(2)(2014)398–406. [6]A.Krabben,W.Stomp,J.A.B.VanNies,T.W.J.Huizinga,D.VanDerHeijde,J.L. Bloem,etal.,MRI-detectedsubclinicaljointinﬂammationisassociatedwith radiographicprogression,Ann.Rheum.Dis.73(11)(2014)2034–2037. [7]H.Nguyen,A.Ruyssen-Witrand,F.Gandjbakhch,A.Constantin,V.Foltz,A.

Cantagrel,Prevalenceofultrasound-detectedresidualsynovitisandriskof relapseandstructuralprogressioninrheumatoidarthritispatientsinclinical remission:asystematicreviewandmeta-analysis,Rheumatology53(11) (2014)1–9.

[8]E.L.Rowbotham,A.J.Grainger,Rheumatoidarthritis:ultrasoundversusMRI, AJRAm.J.Roentgenol.197(3)(2011)541–546.

[9]S.Torp-Pedersen,R.Christensen,M.Szkudlarek,K.Ellegaard,M.A.D’Agostino, A.Iagnocco,etal.,PowerandcolorDopplerultrasoundsettingsfor inﬂammatoryﬂow:impactonscoringofdiseaseactivityinpatientswith rheumatoidarthritis,ArthritisRheumatol.67(2)(2015)386–395. [10]I.K. Haugen,H.B. Hammer, Aneed for new imaging modalityto detect

inﬂammationinrheumatoidarthritisandosteoarthritis?Ann.Rheum.Dis.75 (3)(2016)479–480.

[11]A.J.Meier,W.H.Rensen,P.K.deBokx,R.N.deNijs,Potentialofopticalspectral transmissionmeasurementsforjointinﬂammationmeasurementsin rheumatoidarthritispatients,J.Biomed.Opt.17(8)(2012)081420. [12]A.M.Glimm,S.G.Werner,G.R.Burmester,M.Backhaus,S.Ohrndorf,Analysisof

distributionandseverityofinﬂammationinpatientswithosteoarthitis comparedtorheumatoidarthritisbyICG-enhancedﬂuorescenceoptical imagingandmusculoskeletalultrasound:apilotstudy,Ann.Rheum.Dis.75 (3)(2016)566–570.

[13]M.vanOnna,D.F.TenCate,K.L.Tsoi,A.J.Meier,J.W.Jacobs,A.A.Westgeest, etal.,Assessmentofdiseaseactivityinpatientswithrheumatoidarthritis usingopticalspectraltransmissionmeasurements,anon-invasiveimaging technique,Ann.Rheum.Dis.75(3)(2016)511–518.

[14]M.Krohn,S.Ohrndorf,S.G.Werner,B.Schicke,G.R.Burmester,B.Hamm,etal., Near-infraredﬂuorescenceopticalimaginginearlyrheumatoidarthritis:a comparisontomagneticresonanceimagingandultrasonography,J. Rheumatol.42(7)(2015)1112–1118.

[15]R.Meier,K.Thuermel,P.B.Noël,P.Moog,M.Sievert,C.Ahari,etal.,Synovitisin patientswithearlyinﬂammatoryarthritismonitoredwithquantitative

(7)

analysisofdynamiccontrast-enhancedopticalimagingandMRimaging, Radiology270(1)(2014)176–185.

[16]V.S.Schäfer,W.Hartung,P.Hoffstetter,J.Berger,C.Stroszczynski,M.Müller, etal.,Quantitativeassessmentofsynovitisinpatientswithrheumatoid arthritisusingﬂuorescenceopticalimaging,ArthritisRes.Ther.15(5)(2013). [17]S.G.Werner,H.E.Langer,S.Ohrndorf,M.Bahner,P.Schott,C.Schwenke,etal.,

Inﬂammationassessmentinpatientswitharthritisusinganovelinvivo ﬂuorescenceopticalimagingtechnology,Ann.Rheum.Dis.71(4)(2012)504– 510.

[18]L.V.Wang,S.Hu,Photoacoustictomography:invivoimagingfromorganelles toorgans,Science335(6075)(2012)1458–1462.

[19]P.Beard,Biomedicalphotoacousticimaging,InterfaceFocus1(4)(2011)602– 631.

[20]L.V.Wang,J.Yao,Apracticalguidetophotoacoustictomographyinthelife sciences,Nat.Methods13(8)(2016)627–638.

[21]V.Ntziachristos,Goingdeeperthanmicroscopy:theopticalimagingfrontierin biology,Nat.Methods7(8)(2010)603–614.

[22]P.J.vandenBerg,K.Daoudi,W.Steenbergen,Reviewofphotoacousticﬂow imaging:itscurrentstateanditspromises,Photoacoustics3(3)(2015)89–99. [23]Z.J. Guo, Z. Xu, L.H.V. Wang, Dependence of photoacoustic speckles on

boundaryroughness,J.Biomed.Opt.17(4)(2012).

[24]T.Kitai,M.Torii,T.Sugie,S.Kanao,Y.Mikami,T.Shiina,etal.,Photoacoustic mammography:initialclinicalresults,BreastCancer21(2)(2014)146–153. [25]E.Fakhrejahani,M.Toii,T.Kitai,S.Kanao,Y.Asao,Y.Hashizume,etal.,Clinical reportontheﬁrstprototypeofaphotoacoustictomographysystemwithdual illuminationforbreastcancerimaging,PLoSOne10(10)(2015).

[26]M.Heijblom,D.Piras,M.Brinkhuis,J.C.vanHespen,F.M.vandenEngh,M.van derSchaaf,etal.,Photoacousticimagepatternsofbreastcarcinomaand comparisonswithMagneticResonanceImagingandvascularstained histopathology,Sci.Rep.5(2015)11778.

[27]M.Heijblom,D.Piras,F.M.vandenEngh,M.vanderSchaaf,J.M.Klaase,W. Steenbergen,etal.,ThestateoftheartinbreastimagingusingtheTwente PhotoacousticMammoscope:resultsfrom31measurementsonmalignancies, Eur.Radiol.(2016).

[28]X.Wang,D.L.Chamberland,P.L.Carson,J.B.Fowlkes,R.O.Bude,D.A.Jamadar, etal.,Imagingofjointswithlaser-basedphotoacoustictomography:ananimal study,Med.Phys.33(8)(2006)2691–2697.

[29]J.R. Rajian, G. Girish, X. Wang, Photoacoustic tomography to identify inﬂammatoryarthritis,J.Biomed.Opt.17(9)(2012).

[30]J.R. Rajian, X. Shao, D.L. Chamberland, X. Wang, Characterization and treatmentmonitoringofinﬂammatoryarthritisbyphotoacousticimaging: astudyonadjuvant-inducedarthritisratmodel,Biomed.Opt.Express4(6) (2013)900–908.

[31]N.Beziere,C.VonSchacky,Y.Kosanke,M.Kimm,A.Nunes,K.Licha,etal., Optoacousticimagingandstagingofinﬂammationinamurinemodelof arthritis,ArthritisRheumatol.66(8)(2014)2071–2078.

[32]G. Xu,J.R. Rajian,G. Girish,M.J.Kaplan, J.B.Fowlkes, P.L. Carson,etal., Photoacousticandultrasounddual-modalityimagingofhumanperipheral joints,J.Biomed.Opt.18(1)(2013).

[33]K.Daoudi,P.J.VanDenBerg,O.Rabot,A.Kohl,S.Tisserand,P.Brands,etal., Handheldprobeintegratinglaserdiodeandultrasoundtransducerarrayfor ultrasound/photoacousticdualmodalityimaging,Opt.Express22(21)(2014) 26365–26374.

[34]C.Lutzweiler,R.Meier,E.Rummeny,V.Ntziachristos,D.Razansky,Real-time optoacoustictomographyofindocyaninegreenperfusionandoxygenation parametersinhumanﬁngervasculature,Opt.Lett.39(14)(2014)4061–4064. [35]P.VanEs,S.K.Biswas,H.J.B.Moens,W.Steenbergen,S.Manohar,Initialresults ofﬁngerimagingusingphotoacousticcomputedtomography,J.Biomed.Opt. 19(6)(2014).

[36]L.Xi,H.B.Jiang,Highresolutionthree-dimensionalphotoacousticimagingof humanﬁngerjointsinvivo,Appl.Phys.Lett.107(6)(2015).

[37]Z.Deng,C.Li,Noninvasivelymeasuringoxygensaturationofhuman ﬁnger-jointvesselsbymulti-transducerfunctionalphotoacoustictomography,J. Biomed.Opt.21(6)(2016)61009.

[38]J.Jo,G.Xu,A.Marquardt,G.Girish,X.Wang,Photoacousticevaluationof humaninﬂammatoryarthritisinhumanjoints,Proc.SPIE(2017) (1006409-1006408).

[39]G.Xu,D.Chamberland,G.Girish,X.D.Wang,Photoacousticandultrasound dual-modalityimagingforinﬂammatoryarthritis,PhotonicTher.Diagn.X (2014)8926.

[40]M.Szkudlarek,M.Court-Payen,S.Jacobsen,M.Klarlund,H.S.Thomsen,M. Østergaard,Interobserveragreementinultrasonographyoftheﬁngerandtoe jointsinrheumatoidarthritis,ArthritisRheum.48(4)(2003)955–962. [41]M.Jaeger,S.Schüpbach,A.Gertsch,M.Kitz,M.Frenz,Fourierreconstructionin

optoacousticimagingusingtruncatedregularizedinversek-space interpolation,InverseProb.23(6)(2007)S51–S63.

[42]B. Cox, J.G. Laufer, S.R. Arridge, P.C. Beard, Quantitative spectroscopic photoacousticimaging:areview,J.Biomed.Opt.17(6)(2012).

[43]S.L.Jacques,Opticalpropertiesofbiologicaltissues:areview,Phys.Med.Biol. 58(11)(2013)R37–61.

[44]J.Jo,G.Xu,A.Marquardt,S.Francis,J.Yuan,D.Girish,etal.,Photoacoustic imagingofinﬂammatoryarthritisinhumanjoints,Proc.SPIE9689(2016). [45]P.Stefan,H.Gerrit,G.A.Hidayet,J.Michael,F.Martin,Studyofclutteroriginin

in-vivoepi-optoacousticimagingofhumanforearms,J.Opt.18(9)(2016) 094003.

[46]M. Kuniyil, Ajith Singh, W. Steenbergen, Photoacoustic-guided focused ultrasound(PAFUSion)foridentifyingreﬂectionartifactsinphotoacoustic imaging,Photoacoustics3(4)(2015)123–131.

[47]H.M.Schwab,M.F.Beckmann,G.Schmitz,Photoacousticclutterreductionby inversionofalinearscattermodelusingplanewaveultrasound

measurements,Biomed.Opt.Express7(4)(2016)1468–1478.

[48]G.Held,S.Preisser,H.GünhanAkarçay,S.Peeters,M.Frenz,M.Jaeger,Effectof irradiationdistanceonimagecontrastinepi-optoacousticimagingofhuman volunteers,Biomed.Opt.Express5(11)(2014)3765–3780.

[49]J.Lemaster,J.V. Jokerst,What’snewinnanoparticle-basedphotoacoustic imaging,WIREsNanomed.Nanobiotechnol.9(1)(2016),doi:http://dx.doi.org/ 10.1002/wnan.1404.

PimvandenBergisaPhDresearcherattheUniversityof Twente,theNetherlands.HeisworkingontheEuropean projectFullphase,whichaimstodevelopanaffordable andportableultrasound/photoacoustic (US/PA)system forearlydiseasedetection.Hismainresearchinterests areﬂowimagingusingphotoacousticsandthe applica-tionofUS/PAimagingfortheassessmentofrheumatoid arthritis. Beforestarting his PhD, Pimdidhis master studiesonOpticsandBiophysics,buildingaSTORMsuper resolutionmicroscopeanduseditforcharacterizationof proteinaggregationinParkinson'sdisease.Interestsalso includehighschoolsciencepromotion,having partici-patedinamediapusharoundphotoacousticimagingfor thepopularizationofappliedsciences.

KhalidDaoudi,receivedhisPhDdegreeinAppliedOptics fromuniversityPierreetMarieCurie(ParisVII)ofParis, FranceonhisworkonOptical-ElastographyatLangevin Institute(ESPCI).Afterhegraduatedhestartedapost-doc position at Institute for Biomedical Technology and TechnicalMedicine,BMPIgroupatuniversityofTwente in Netherlands. His research focused on optical and hybridacousticalandopticalimagingmethodssuchas photoacoustics and acousto-optics and modeling of sound/lighttissueinteraction.Recentlyhejoined Rad-boud University Medical Center (Nijmegen, The Netherlands)atthedepartmentofRadiologywherehe isworkingonthedevelopmentofphotoacousticimaging techniqueatMedicalUltrasoundImagingCenter(MUSIC)group.

HeinBernelot Moensisarheumatologist inhospital ZiekenhuisgroepTwente,theNetherlands.Hespecialized ininternalmedicineandrheumatology,andreceivedin 1991hisPhDincomputerassisteddiagnosisofrheumatic diseasesattheuniversityofAmsterdam.Since2005heis qualifiedinultrasoundofthemusculoskeletalsystem, andusesultrasoundroutinelyinpatiëntcare.Nextto clinicalwork,hejoinedresearchprojectson computer-assistedimagingofhandradiographs.Since2011heis involvedinthedevelopmentofclinicalapplication of photoacousticimagingofsynovialinflammationatthe BMPIdepartmentoftheUniversityofTwente.Since2015 heisPresidentoftheDutchSocietyforRheumatology. Wiendelt Steenbergen obtained a Master degree in AerospaceEngineeringattheDelftUniversityof Technol-ogy (1988), a PhD degree in fluid dynamics at the EindhovenUniversityofTechnology(1995)andjoined theUniversityofTwente,Enschede(theNetherlands).In 2000hewasappointedassistantprofessorinbiomedical optics and broadened his scope to low-coherence interferometry and photoacoustic and acousto-optic imaging.In2010 hebecamefull professorandgroup leaderoftheBiomedicalPhotonicImaginggroupofthe UniversityofTwente.Hiscurrentresearchinterestsare specklebasedperfusionimaging,photoacousticimaging formammographyandrheumatology,andquantification ofphotoacousticimagingusingacousto-optics.