ContentslistsavailableatSciVerseScienceDirect
Medical
Engineering
&
Physics
j o ur n a l ho m e p age : w w w . e l s e v i e r . c o m / l o c a t e / m e d e n g p h y
In
vivo
interactions
between
tungsten
microneedles
and
peripheral
nerves
Pier
Nicola
Sergi
a,∗, Winnie
Jensen
b, Silvestro
Micera
a,c, Ken
Yoshida
daBioRoboticsInstitute,ScuolaSuperioreSant’Anna,Italy
bDepartmentofHealthScienceandTechnology,AalborgUniversity,Denmark cInstituteforAutomation,SwissFederalInstituteofTechnology,CH,Switzerland
dBiomedicalEngineeringDepartment,IndianaUniversity-PurdueUniversityIndianapolis,USA
a
r
t
i
c
l
e
i
n
f
o
Articlehistory:
Received13December2010 Receivedinrevisedform28July2011 Accepted19September2011 Keywords:
Biomechanics
Peripheralnervoustissue Neuralinterfaces Indentation
Microneedle/livingnervedynamical frictioncoefficient
a
b
s
t
r
a
c
t
Tungstenmicroneedlesarecurrentlyusedtoinsertneuralelectrodesintolivingperipheralnerves. How-ever,thebiomechanicsunderlyingtheseproceduresisnotyetwellcharacterized.Forthisreason,the aimofthisworkwastomodeltheinteractionsbetweenthesemicroneedlesandlivingperipheralnerves. Asimplemathematicalframeworkwasespeciallyprovidedtomodelbothcompressionoftheexternal layerofthenerve(epineurium)andtheinteractionsresultingfrompenetrationofthemainshaftof themicroneedleinsidethelivingnerves.TheinstantaneousYoung’smodulus,compressionforce,the workneededtopiercethetissue,puncturingpressure,andthedynamicfrictioncoefficientbetweenthe tungstenmicroneedlesandlivingnerveswerequantifiedstartingfromacuteexperiments,aimingto reproducethephysicalenvironmentofrealimplantations.Indeed,abetterknowledgeofthe interac-tionsbetweenmicroneedlesandperipheralnervesmaybeusefultoimprovetheeffectivenessofthese insertiontechniques,andcouldrepresentakeyfactorfordesigningrobot-assistedprocedurestailored forperipheralnerveinsertion.
© 2011 IPEM. Published by Elsevier Ltd. All rights reserved.
1. Introduction
Thereareseveralmedicalproceduresthatinvolvetheinsertion ofneedlesintobiologicaltissues:needlesareusedtoplace radioac-tiveseedsclosetoclustersoftumourcells(brachytherapy)[1],to reachlessaccessiblesitesandextractbiologicalspecimens(biopsy) [2,3],andtoinoculatedrugswithhighaccuracyinsidethebody (severaltypesofinjections)[4].Theyarealsocommonlyusedin microneurography[5,6].
Since thesekinds of procedures canbe intrinsically difficult andcansometimescausecomplications,alsoafterintense train-ingbysurgeons,severalstudieswereconductedtoinvestigatethe behaviourofneedlesduringinsertioninsofttissues.Inthese stud-ies,theneedlesweremodelledas linearelasticwithgeometric nonlinearities[7–9],andsimulationswerevalidatedusing numer-icalmodelsor phantomtissues.Theworknot onlycontributed tounderstandingtheseinsertionprocedures,butalsoprovideda generalizationtoautomatethem. Aparticularclass ofinsertion involvestheuseofmicroneedleswhichareusedforplacingneural interfacesintoperipheralnerves.Inparticular,exvivoinsertionof
∗ Correspondingauthorat:ScuolaSuperioreSant’Anna,piazzaMartiridella Libertà,33,56127Pisa,Italy.Tel.:+39050883135;fax:+39050883101.
E-mailaddress:[email protected](P.N.Sergi).
theseneedleshavebeenexperimentallydescribedinporcine[10] andrabbit[11]nerves.
Althoughthebehaviourofperipheralnerveshasalreadybeen studiedinradialcompressionexperimentsaimingtosimulatethe applicationofcuffelectrodes[12],thebiomechanicsofinteraction betweentungstenmicroneedlesandlivingperipheralnerveshas notbeensufficientlycharacterizedyet.
Indeed,thisknowledgecouldbeusefultoimprovethequality ofstandardinsertionprocedures,andcouldrepresentakeyfactor forphysicallybasedfeedbackaimedatmonitoringrobot-assisted procedures[4,13]inperipheralnervepunctureandinsertion tech-niques.
Therefore,theaimoftheworkwastoprovideasimple theo-reticalframeworkforanalysingtheexperimentalresultsachieved duringinvivoinsertionoftungstenmicroneedles.Theframework wasabletoaccountforthemainstepsoftheinsertion process, consideringontheonehandboththematerialofthemicroneedles andtheirtipgeometry,andontheotherthebiomechanical-based responseofthelivingnerves.Inparticular,themodelwasableto quantify notonlypiercingpressures,forces, andwork,but also physical parameters,suchas theinstantaneousYoungmodulus of livingperipheralnervesand thedynamic frictioncoefficient, characterizingtissueresponseandinteractionwiththetungsten microneedles.
Thiskindofdatacouldbeusefultoimproveinteractionmodels accountingforcomplexanddissipativeeffects.
1350-4533/$–seefrontmatter © 2011 IPEM. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.medengphy.2011.09.019
748 P.N.Sergietal./MedicalEngineering&Physics34 (2012) 747–755 2. Materialsandmethods
2.1. Animalpreparation
Insitu acuteexperimentswereperformedontheperipheral nervesofthreeDanishLandracepigs(∼8–9monthsold weigh-ingabout45kg).Allexperimentalprocedureswereapprovedby theAnimalExperimentsInspectorateundertheDanishMinistryof Justice.
The pigs were anesthetized using a mixture of Xylazine (Rompun®, 20ml/l), Ketamine (Ketaminol®, 50mg/ml),
Butor-phanol Tartrate (Turbugesic®, 10mg/ml) and a combination of
Tiletamin and Zolazepam (combined in Zoletil, 50mg/ml). The animalswereintubatedand placedona veterinaryanaesthesia ventilator(model2000,HallowellEMC,USA) at15breaths/min. Anaesthesiawasmaintainedusinga50–50%air/oxygenmixture with 1–1.2% Isoflurane. The animals received 0.9% NaCl saline bycontinuousIVinfusionthroughearveintoprevent dehydra-tion,androcuroniumbromide(Esmeron®,10mg/ml)andFentanyl
(“Hameln”50g/ml)toprovideanalgesiathroughouttheentire experiment.Theanimals’heartrateandoxygensaturationwere monitoredthroughouttheexperiment(Fig.1a).
2.2. Experimentalmethodsanddataacquisition
Surgical access was created in the upper forelimb to the ulnar and mediannerves, and thesite of insertion waschosen approximately50mmabovetheelbow.Preciseimplantationswere performedataconstantvelocity[14]of2mm/s,intheulnarnerves. Thecompressiveforcewasmeasuredusingaloadcell(Sensotec Inc.,Model31/1435-02,maxload0.1kg,resolution1.31mV/g).A malefittingontheloadcellwasusedtoacceptafemaleLuerlock ontowhichtheneedlesweremounted(Fig.1b).
The nerves were elevated using a plastic platform with a 4mmdiameterhole(seeFig.1b)centeredunderneaththem.The microneedletipswereperpendicularlyplacedslightlyabovethe externalsurfaceofthenerves,andwerethenadvancedintothem byamotor-controlled(MaxonDCMotor22-60-881,JVLIndustri ElektronikA/S,DK)hydraulicmicromanipulator(Narishige MMO-220)witharesultingmovementresolutionof0.25m.Aftereach insertion,theneedlewasrepositionedina differentpartofthe nervetoavoidmultipleinsertionpointsinthesamelocation.
Twogroupsofelectro-sharpenedtungstenmicroneedles,with differentshaftdiameters(∅)andconicaltipswereusedtoperform 32insertions:19and 13trialswereconductedusing micronee-dleswith∅=75mand∅=100m,respectively.Moreover,allof themwereassumedtohaveYoungmodulusandPoissonratioof En=411GPa,andn=0.28,respectively.
Round tungstenrods (no insulation, 75 and 100m diame-ters,A-MSystems)werecutinto2cmlengthsandthenmanually electro-sharpenedtocreateasharptip.Thetipsweresharpened byloweringoneendintoa2NKNO3 solution,placinga∼8VAC
potentialacrossit,andusingalargecarboncounterelectrodefor 20–40stoetchthem.Thetipswerevisuallyinspectedthrougha microscopeduringandaftertheelectrolysesprocedure.Finally,the openingangleofthetipwasmeasuredfollowingoptical photomi-crography(Fig.1c).Allelectrodeswerecleanedinde-ionizedwater beforeimplantationtoremoveanyresiduefrometchant.
The microneedles were inserted according to a ramp–hold–ramp profile (Fig. 1d), and the force and position signals(Fig.1e)werefilteredbeforesampling(1storderlowpass filter, sampling frequency=2.5kHz, NI DAQ card-6204E). The sampleddatawerealsofilteredoffline,tominimizesuperimposed physiologicalandinstrumentalnoise.Inparticular,therelaxation phase wasfilteredusinga 3rdorderButterworthlowpassfilter
(F3dB=0.325Hz),whilethecompressionandinsertionphaseswere
filteredwitha3rdorderButterworthlowpassfilter(F3dB=12.5Hz).
2.3. Modellingofcompressionandpuncture
Whenthetipof a microneedlecomesintocontact withthe outermostlayerofthenerve(epineurium),thepressureincreases (Fig.1e,phaseA)andthelayerdeforms:therelationbetweenthese setsofquantitiesisgovernedbyconstitutiveequations.
SinceanonlinearKelvinmodel[15]wasusedtoapproximate the generalbehaviour of the living nerves,and thevelocity of microneedleswasconstantforallphases(Fig.1e,phasesA–B–D), theinteractionforceswerewrittenas:
F(z,
v
s)=fs(z)+k(z)v
s 1−exp −v
z s (1) wherezisthedimplingofthetissue(which,inthiscase,equals themicroneedletipdisplacement),v
=dz/dtisthevelocityofthe needletip,fs(z)isanonlinearforce–displacementfunctionandk(z)anonlinearspringconnectedwithanonlineardamperb(z),witha characteristicrelaxationtimes=b(z)/k(z).Theexponentialtermof
Eq.(1)dependsonthedimplingofthetissuebeforethepuncture (z),aswellasonthecharacteristicrelaxationtimeofnerve(s)
andpiercingvelocity(
v
).Sinceinthisstudyvsz,theasymptoticexpansionofEq.(1)withrespecttovswasused:
F(z,
v
s)=fs(z)+k(z)v
s 1−exp −v
z s =fs(z)+k(z) z+O 1v
s (1.1) andEq.(1)wasapproximatedwithEq.(1.2):F(z) ∼= fs(z)+k(z)z (1.2)
Eq.(1.2)showsthatthetotalforceisindependentoftheviscous characteristicsofthetissueforanychoiceoffs(z)andk(z).
Aboveall,duetothesmallsizeoftheneedletip(100m) withrespecttothenervediameter(3–5mm),thisproblemwas treatedasanindentationofanincompressibleelastictissue(i.e.the livingnerve)performedwithaconicalindenter(i.e.the micronee-dletip).Consequently,Eq.(1.2),whichimplicitlyrelatesforcesand dimpling,wasmadeexplicitbymergingthetwotermsandusing Sneddon’sapproach[16,17]togetherwiththechangesduetothe realgeometryoftheconicaltip[18].Furthermore,theliving periph-eralnerveswereconsideredanonlinearelasticmaterial,andthe measuredloadF(z)waswrittenas:
F(z)=8tg(˛)3 M(E,z)z[z+ ¯ ()] (2) where E is the instantaneous Young’s modulus of the nerve, M(E,z)=Eexp(bz)modelsthenonlinearelasticityofthenerve[12] undergoingafinitedeformation(indentation)andassumed incom-pressible[19],˛isthehalfangleoftheneedletip,zisthedimpling ofthenerve(ortipdisplacement),and ¯ ()isafunctionofthe cur-vatureradiusofthetip,withthedimensionofalength.According to[18],forbothgroupsofmicroneedles,Eq.(2.1)canbewrittenas:
¯
()=c12+c2 (2.1)
wherec1=1.5×10−4m−1andc2=1.17×10−1.
Thehalfangles(˛)andthetipradii()ofthemicroneedletips weremeasuredfromdigitalpictures,andEq.(2)wasusedtofit experimentaldata(phaseA)throughnonlinearoptimizationwith parameteridentificationbasedonmeasuredcompressionforces (quasiNewtonianalgorithm,Scilab©INRIA).
Allexperimentsperformedwithmicroneedleswith∅=75m and∅=100mwereseparatelygroupedtoextractthemean val-uesofthequantities(E,b)characterizingthetissueresponse.
Fig.1.(a)Generalviewofexperimentalsetup.(b)Magnificationoftungstenmicroneedleandperipheralnervewithsupport.(c)Microneedletipgeometry,themeasures areindegreesandmicrometres.(d)Theramp-old-rampprofileofinsertion-rest-extractionofthemicroneedle.(e)Acharacteristiccurve(forcevs.time),inwhichdifferent phasesareunderlined.A:Compressionphase(frombeginningtopoint˛).˛:Pointofpuncture,wherecharacteristicdiscontinuityinthecompressionforcecanbeseen.B: Insertionphase.ˇ:Endoftheinsertionphase,wherethemovementofthemicroneedlecomestoastopinsidethetissue.C:Relaxationphase.:Retractionpoint,where theneedlebeginstoberetractedfromthetissue.D:Retractionphase.ı:Extractionofthemicroneedlefromthetissue.Thenegativevaluesoftheforces(traction)aredue tobindingofthetissueontheexternalsurfaceofthemicroneedle.
750 P.N.Sergietal./MedicalEngineering&Physics34 (2012) 747–755
Fig.2.Force–displacementcurvesduringthephaseofpiercing.Thisphasecanbe theoreticallydescribedasanindentation.(a)Experimental(lightgrey)indentations witha∅=75mtungstenmicroneedle,theoreticalfittingcurve(solidblack).(b) Experimental(lightgrey)indentationswitha∅=100mtungstenmicroneedle, theoreticalfittingcurve(solidblack).
Tointegratethisdirectbiomechanicalapproach,anindirect sen-sitivityanalysiswasperformedtofurtherinvestigatethedegreeof nonlinearityofthecompressionforce.Letq(z,n)beapolynomial functionofnorder,asshowninEq.(2.2):
q(z,n)=
n
j=1
nzn (2.2)
wherenaretheconstantcoefficientsofeachn-orderterm(the
0-ordertermissetidenticallytozerobecausetheforceof com-pressioniszeroforz=0).Thesetofexperimentalcurveswasfitted withthesephenomenologicalfunctionsbyvaryingthemaximum orderofpolynomialapproximation(n-index)inordertoevaluate boththeresidualsandR2 index(R2isthesquareofthe
correla-tionbetweentheresponsevaluesandthepredictedvalues).The orderofnforwhichtheimprovementofindicesstopsistheorder ofnonlinearityofEq.(2.2).
RecallingEq.(2),theglobalworkappliedtothenerveduring compressionuptothepuncturewaswrittenas:
W=
z00
F(z)dz (3)
where z0 is the coordinate where the tissue was pierced.The
nervewaspuncturedwhenacharacteristicmaximumpressurewas exceeded,thesurfaceofthenerveyieldedandacrackopenedin thetissue.Atthisstage,thecentralradiusofthecrackwasassessed withtheradiusofthemicroneedletip.Duringinsertion,thesizeof thecrackincreaseduptotheexternaldiameterofthemainshaftof themicroneedle:intheeventofyielding,thetheoreticalpressure atthemicroneedle-nerveinterfacewasroughlyassessedbyusing thefollowingequation[20]:
p(r)= 1−rt/r
(1−n) /En+3/[2Ets(εr)]
(4) where En and n are the Young and Poisson modules of the
needle, r and rt are the radii of the needle shaft and of the
needle tip, and Ets is the instantaneousYoung modulus of the
tissue.
Whentheinitialmicrocrackexpanded intoa hole,the non-linear behaviour of the tissue wasmodelled with thefunction Ets(r)=Eexp(ˇεr),whereεr=(rn−rt)/rt andˇ>0representan
exponentialconstantmodellingtheradialstrainhardeningofthe tissue[12].Eq.(4)isstrictlyvalidforapairofperfectlyelastic mate-rialsandusuallyprovideshighpressurevalueswhenther/rtratio
ishigh,asinthiscase.
Nevertheless,themaximumpressurethatcanbeobtainedat theinterfaceisyieldingpressure;viscousflowsarisebeyondthis thresholdwhichlowerthelocalfieldofcontactstresses.Eq.(4) showstheneedforyieldingattheinterfacebetweenthenerveand themicroneedlesbutisnotabletoquantifyit.Tothisaim,itcould beusefultorecallthebehaviourofthenerveunderthe micronee-dletip,beforeopeningofthehole:thetissueyieldsandviscous flowsarisebeforefracturingoftheoutermostlayer(viscoelastic fracture).
Asaconsequence,thesuperficialyieldingpressurewas reason-ablyapproximatedwiththepiercingpressure:
p(r)= ¯py=F(z0)
A (5)
whereA=[22+(r+)
(r−)2+L12]andistheradiusof
curvatureofthemicroneedletip,risthenominaldiameterofthe mainshaft,andL1istheheightoftheconicalregionattheendofthe
needle[15,21].Toachieveameanvalueofthisparameter,digital picturesofmicroneedleswereanalysedandmeasured.Itshouldbe notedthatEq.(5)providesthesuperficialyieldingpressure,which differsfromthemeanpressureoverthemicroneedlecrosssectional areagivenby:
¯pc=
4F(z0)
d2 (5.1)
wheredisthenominaldiameterofthemicroneedle.
Atotalforceisappliedonthemicroneedlesafterpuncturingof thetissue,whichderivesfromthesumoffrictionandaconstant cuttingforce[14].Aboveall,thetotalinteractionforceincreases almostlinearlyduringinsertiontogetherwiththecontactsurface ofthemicroneedle.
Forthesakeofsimplicity,thefrictionwasmodelledas Coulom-bian,thenthemeandynamiccoefficientwaswrittenas:
¯ d= ¯
¯py
sgn(
v
) (6)where ¯=m/2risthemeanshearstressonthelateralsurfaceof theinsertedmicroneedle,misthederivativewithrespecttozof
Fig.3. (aandb)Polynomialfunctionsfittingexperimentalcurvesofcompressionwitha∅=75m(a)anda∅=100m(b)microneedle.Allfunctionswithn≥2perform thesamefitandareshownsuperimposedtothesolidblackcurve.(candd)Residualsofpolynomialfitsofexperimentaldataachievedwitha∅=75m(c)and∅=100m. (d)Forn≥2residualsareexactlysuperimposedtotheblackarea.(e)ChangesofR2indexvaryingthen-index(maxdegreeofpolynomialfunctionandorderofnonlinearity
ofthecompression)forbothsetsofexperiments.Forn≥2R2indexremainsconstant,anddoesnotallowdirectassessmentofthen-orderofcompression.(f)Relevance
analysisofconstantcoefficientsofallfittingfunctions:forbothsetsofexperimentsonlythelinear(1)andquadratic(2)termsaredifferentfromzero(lightgrey=0,dark
grey>grey>0,white=notallowed),revealingaquadraticcoreofthecompressionphenomenon.Forsakeofsimplicityonlyfourpowersareshown.
thestraightlinefittingthetotalforce,andsgn(
v
)isthesign func-tionofthemicroneedlevelocityv
.Thepiercingpressureandthe slopeoftheinsertionforceareunivocallyrelatedforeachcurve. Asaconsequence,eachcurvewasseparatelyanalysed to quan-tifythefrictioncoefficientavoidingerrorscomingfromtheuseof meanvalues.Additionally,thisapproachallowedustoinvestigate thenatureofthedistributionofthedynamicalfrictioncoefficient amongexperiments.3. Results
3.1. Anelasticframeworktodescribeinteractionbetweennerve andmicroneedles
For insertion experimentsperformedwiththe firstgroupof tungstenmicroneedles(∅=75m)meancharacteristicrelaxation timeand meandimplingwererespectively ¯s75=28.9558sand
752 P.N.Sergietal./MedicalEngineering&Physics34 (2012) 747–755 ¯z075=2.6456mm,whileforexperimentsperformedwiththe
sec-ondgrouptheywere(∅=100m), ¯s100=28.8158sand ¯z0100=
3.6369mm.Inbothcases,
v
¯s75 ¯z075andv
¯s100>> ¯z0100,thenEqs.(1.2)and(2)wereusedandtheelasticframeworkwasableto correctlydescribetheexperimentalresults.
3.2. Geometryofmicroneedlestipsandinstantaneousresponseof thenerve
Thehalf angles(˛) and thetip radii ()of the microneedle tipsforthefirstgroupwere˛=10.25±0.5◦and=1.7±0.577m, while for the second group they were ˛=7.875±0.478◦ and =2.48±0.349m:thesevalueswereinsertedinEqs.(2)and(2.1) tofitdataresultingfromexperimentalindentations(phaseA).
Fig.2shows bothexperimentsandfittingcurves:Eq.(2) fit-tedexperimentaldataof thefirst group(R2=0.8591) totheset
{¯E75=22.2265kPa, b75=0, ¯ ()75=1.99×10−4mm}, and also
fittedexperimentaldataofthesecondgroup(R2=0.9649)tothe
set{¯E100=32.9663kPa,b100=0, ¯ ()100=3×10−4mm}.
Furthermore,theindirectanalysisofthecompressionclearly showedaquadraticnatureofthephenomenon(seeFig.3a–f).The analysisofbothresidualsandR2indexesdidnotsupportthis
con-clusion.Indeed,inbothcasesforn≥2therewasasuperimposition ofresidualsandaconstancyofR2asshowninFig.3c–e.
However,therelevance analysisof thecoefficients offitting functions(Fig.3f)showedthatonlythelinearandquadraticterms weredifferentfromzeroforallpolynomialswithn≥2andforboth groupsofmicroneedles.
3.3. Piercingandinsertionfeatures
TissuedimplingandpiercingwasestimatedwithEq. (3).To this aim, thecontact areasfor both groups were calculated as A75=0.03304±0.0038mm2 andA100=0.0663±0.0130mm2,and
Eq.(3)ledto ¯W75=28.4912Jand ¯W100=60.8844J.Moreover,
Fig.4a showstheboxplotsofbothquantitiesW75andW100.In
particular,thelinesinsidethecentralboxesrepresentthe medi-ans ˜W75=22.9210Jand ˜W100=54.0432J,theupperandthe
lowerlinesoftheboxesshowtheupperandlowerquartiles,and the error bars include outliers. Fig. 4b shows thebox plots of thepiercingforces:themeanvalueswere ¯Fp75=27.4301mNand
¯Fp100=54.1736mN,whilethemediansofthesamequantitieswere
˜Fp75=24.094mNand ˜Fp100=51.4703mN.
Eq.(4)providesamuchlargertheoreticalinterfacialpressure thantheyieldingpressureofanysofttissue.Indeed,forˇ=0.145 and ¯E75or ¯E100,thepressureswerefoundtobe ¯p75≈110MPaand
¯p100≈162MPa,whilethetractionyieldstressfora 1002Asteel
was131MPa.Itshouldbenoticed,accordingto[12],thatavalue of ¯ˇ=4.03wasfoundforacircularcompressionofrabbitsciatic nerves.Therefore,Eqs.(5)and(6)wereusedtoapproximatethe yieldingpressureandthedynamicfrictioncoefficient.
Indeed, Fig. 5 shows the box plots of the piercing pres-sures and dynamic friction coefficients for each specimen and for each kind of microneedle. The mean and median values of yield contact pressure, pressure on the cross sectional area anddynamicfrictioncoefficientformicroneedleswith∅=75m were: ¯py75=0.8302MPa, ¯pc75=6.2121MPa,¯d75=0.1626,and
˜py75=0.7292MPa, ˜pc75=5.4565MPa, ˜d75=0.1748. Similarly,
for the ∅=100m group, the same quantities were ¯py100=
0.8171MPa, ¯pc100=6.9010MPa, ¯d100=0.1900 and ˜py100=
0.7763MPa, ˜pc100=6.5567MPa,˜d100=0.1726.
Finally,Fig.6showsthequantile–quantileplotsofdistribution ofthedynamicfrictioncoefficientamongexperimentaltrialsfor bothgroupsofmicroneedles.Inparticular,theShapiro–Wilktest wasperformedto assessthe degreeof “normality” of distribu-tion(R:Alanguageandenvironmentforstatisticalcomputing.R
Fig.4.(a)Boxplotsofworktoreachthenervepunctureforbothgroupsof exper-iments:thelinesinsidethecentralboxesrepresentthemedians ˜W75=22.9210J
and ˜W100=54.0432J,theupperandthelowerlinesofboxesshowtheupperand
thelowerquartiles,theerrorbarsincludeoutliers.(b)Boxplotsofforceforpiercing thenervesusingdifferentdiametersofmicroneedles:centrallinesrepresentthe medians ˜Fp75=24.094mNand ˜Fp100=51.4703mN.
FoundationforStatisticalComputing,UniversityofVienna).Forthe firstgroup,thistestresultedinW=0.9311,p-value=0.1623,while forthesecondgroup,W=0.9452,p-value=0.5278,whereWisthe squareofthecorrelationbetweenexperimentaldataandnormal distribution,whilethep-valueistheprobabilityofobtainingatest statisticatleastasextremeastheonethatwasactuallyobserved, assumingthat thenullhypothesis is true.Inthis case, thenull hypothesiswasthatexperimentaldata(d)camefromanormally
distributedpopulation. 4. Discussion
4.1. Asimpleframeworktomodelsuperficialinteractions
In this manuscript, the interactions between tungsten microneedles and the nerve were described using a simple mathematical framework. Specifically, the first phase of com-pression was modelled with an elastic indentation. The more theviscous effects were negligiblein a standard Kelvin model, the more this simplificationapproached theexact solution. As previously shown, the viscous effects slightly influenced the responseoftheperipheralnervesaccordingtothechosenvelocity of punctureand insertion, whichapproximated thatof surgical implants. In particular, viscous effects mainlyaffected the end
Fig.5.(a)Boxplotsofpiercingpressureatpuncturefordifferentdiameterof microneedles:centrallinesrepresentthemedians ˜py75=0.7292MPaand ˜py100=
0.7763MPa.(b)Dynamiccoefficientoffrictionduringinsertionintolivingnervesfor bothdiameterofmicroneedles:centrallinesrepresentthemedians˜d75=0.1748
and˜d100=0.1726.
part of indentation; furthermore, maximum difference with a fully elastic description was approximately 2% at the end of compression,whileatthebeginningitwasnegligible.
Moreover,theclassicalSneddon’s approach wasnot directly used,sincethisapproachisstrictlyvalidforaninfinitelysharp con-icalindenter,whiletherealneedleshadtipswithafiniteradiusof curvature.Therefore,amodificationoftheclassicaltheorywasused accordingto[12]toextracttheinstantaneousYoung’smodulusof thelivingperipheralnervoustissue.
In particular, the mean values of this parameter (22.2265–32.9663kPa) were comparable with mean data from othercompressionexperimentsonrabbitnerves.Indeed,aYoung modulus of41.6±5kPa [22] wasfoundfor in vitro unconfined compressions,andavalueof66.9±8kPa[12]wasfoundforinvivo circularcompressions.
Furthermore,theachievedvalueswerecomparablewith physi-calpropertiesofotherbiologicaltissuesandshowedthattheliving peripheralnervesofpigswereasstiffasdermis(35kPa).Again, theirstiffnesshadthesameorderofmagnitudeofmusclestiffness (80kPa)[23],andtheyweremorecompliantthanarticularcartilage [24].
In addition, although peripheral nerves were consid-ered as nonlinear elastic (Eq. (2)), the achieved values of b (1×10−2<b<1×10−3) showed a practically linear behaviour
Fig. 6.Quantile–quantile plotsof dynamicfriction coefficients achievedwith microneedlesof∅=75mm(a)and∅=100mm(b).Inbothcasesthenormal dis-tribution(straightboldline)ishighlyconsistentwithexperimentalvalues(dotted linesshow95%confidencebounds).
duringlocalizedcompressions(M(E75,100,z)≈ ¯E75,100)with
lim-itedvalues ofz.Asaconsequence,theYoungmodulusofliving nerves waspracticallyconstant and equal totheinstantaneous Young modulus,andforsakeofsimplicitytheparameterbwas settozero.Thisapproachledtoapolynomial(2◦order) indenta-tionlawconsistentwiththeindirectphenomenologicalanalysis (Fig.3a–f)andwiththeapproachgivenin[14].
Finally,thissimpleframeworkwasconsideredcorrectsincethe characteristicrelaxationtimeofthenervemultipliedbythe veloc-ityofindentationwasgreaterthanthecharacteristicdeflectionof thenervesurface.Sinceperipheralnervesmayexperiencealimited dimplingunderthemicroneedletip,thistheoreticalframeworkis still validandusableforalltissueshavingsimilarcharacteristic relaxationtimeandequalorhigherinsertionvelocities. Neverthe-less, viscoelasticeffects havetobeconsideredfor lowinsertion velocitiesorforshortercharacteristicrelaxationtime.
754 P.N.Sergietal./MedicalEngineering&Physics34 (2012) 747–755 4.2. Piercingandinsertionofmicroneedlesintoperipheral
nerves:interactionfeatures
Compressionforceswereusedtogaugeinteractionsbetween livingnervesandmicroneedles.Inparticular,theachieved pierc-ingforcesweregreaterthanthosefoundinsimilarexperiments withexvivoporcine[10]andrabbit[11]nerves.Thiscouldbedue todifferencesinexperimentalsetup.In[10],indeed,thenervewas keptunderaxialtensioninasalinebath,whereasinthese experi-mentsthenerveswerekeptintheair,andaperforatedsupportwas placedunderthemtoeasepiercingandtheinsertionof micronee-dles.However,inbothcases,thepartsofthenerveinteractingwith themicroneedleswerefreetodimpleunderthetipforces.
Anotherreasonforthiswasthegreaterstiffnessoftheliving nerveswithrespecttosimilarexvivospecimens.Indeed,adense microcirculationsysteminsidetheliving nervesprovides nour-ishmenttoallinternalsubstructures,andpressurizedbloodflows insmallveinsandcapillaries.Moreover,thenervousfasciclesare filledwithaphysiologicalfluidunderpressure.Asaconsequence, theglobaltissueresponsetolocalizedindentationisaffectedby thisinternalpressurization.Furthermore,inexvivospecimens,the lackofinternalpressureinfascicles,smallveinsandcapillaries, andpost-mortemautolysis,whichsoftenstheconnectivetissues, worktogethertolowertheglobalresponseofthenervesto exter-nalmechanicalstimuli.Thedifferencebetweeninvivoandexvivo piercingforceswasmostlikelyduetothecomplexinteractionof thesetwocauses,whichgenerallyseemstobevalidforothertissues also[25–27].
Again,itcouldbeworthpointingoutthedifferencesbetween themeanyieldingpressuresandthemeancross sectional pres-sures.Indeed,fortheanalysedgroupsofmicroneedlesthemean crosssectionalpressureswere∼7.4825and∼8.4458timesgreater thanthesuperficialyieldingpressure.Inparticular,since micronee-dleshavetobearthemaximumpiercingforcewithoutanybuckling effect[10,11,28],thecrosssectionalpressureisrelevantfor design-ingbothgeometryandmaterials.
Thepreviousdifferencewasduetotheprocedureusedfor esti-matingthecontactarea.Inthiscase,theareawastheconicalarea ofthetipincontactwiththedimpledtissue,accordingto[21].This choiceledtoamorecloseassessmentoftheforceforthesurface unitofthedimpledtissue.Sincethetipofthemicroneedlesank intothesurrounding(soft)tissuebeforecompression,thecontact areawasassumedtobeconstantforthewholecompression.
Eq.(4)wasusedtoassesstheinterfacialpressurebetweenthe mainshaftofthemicroneedlesandthelivingperipheralnerves. Since this approach exactly assesses the interference pressure betweentwo elasticmaterials,Young’smoduliof tungstenand nerveswereusedinthiswork.Aboveall,aspreviouslydiscussed, amicrocirculationsystemnourishestheinnerpartofthenerve and thefasciclesare filledwithendoneural fluid:sincethe tis-sueissoakedinbiologicalfluids,theincompressibility(≈0.5) [19]ofthewholenerve isacceptableandEqs.(2) and(4)were usedinthesesimplifiedforms.Furthermore,thestiffness ofthe microneedleswasassumedtobemagnitudeshigherthanthatof thelivingperipheralnerves:experimentsconfirmedthis assump-tion(En>> ¯E75,100)allowingEqs.(2)and(4)tobeindependent
fromthemicroneedlematerial.Inaddition,Eq.(4)onlyshowedthe needfortissueyieldingaroundthemainshaftofthemicroneedle. Asaconsequence,Eq.(5)wasusedtoassessthispressurebecause compressionandyieldingoccursimultaneouslyjustbefore pierc-ing.Therefore,thisprocedureallowedustoassessaquantitywhich isdifficulttoexplicitlyachieveinexperiments.
Finally, the interaction between peripheral nerves and microneedleswasroughlymodelledusingaCoulombiandynamic frictionmodel.Thiswasasimplificationofmorecomplex phenom-enathatarosebetweentheneedleshaftandthesurroundingtissue.
Nevertheless,thissimplemodelwasabletocatchthemain macro-scopicfeaturesofthisphenomenon,givingphysicallyreasonable valuesforthedynamicfrictioncoefficientbetweenbiologicaland conventionalmaterials,accordingto[29].Inaddition,astatistical analysisofdatawasprovidedtoinvestigatethecourseofthe dis-tributionofthedynamicfrictioncoefficients.First,aroughanalysis oftheabsolutedifferencesbetweenmeanandmedianvalueswas performedresultingin1.22×10−2(∼7.5%)forthefirstgroupand 1.74×10−2(∼10%)forthesecond.Althoughthesevaluesrevealed asimilarity,thiskindofanalysiswasnotabletoinvestigatethe globalcourseofthedistribution.Therefore,quantile–quantileplots werecreatedandtheShapiro–Wilksnormalitytestwasperformed. Theplotsshowedanalmostnormaldistribution,whilethetest pro-videdhighcorrelationindexes(W>0.9)andp-values>0.05forboth groups.Indeed,experimentaldatawereclosetothestraightline (normaldistribution)andthenullhypothesiswasacceptedinthe standardcaseof5%ofsignificancelevel.
5. Conclusionsandfutureworks
Insertion experiments together with a simple mathemati-cal framework wereprovided toinvestigate in vivo interaction between tungstenmicroneedles and ulnar porcine nerves. This approach allowedustoquantitatively assess themainphysical quantitiesfromwhichtheexperimentalcurvesdepend:the instan-taneousYoung’smodulus,compressionforce,theworkneededto piercethetissue,puncturingpressure,and thedynamic friction coefficientbetweentungstenmicroneedlesandlivingnerveswere quantified. Specifically,thedistribution of thedynamic friction coefficientwasfoundtobeapproximatelynormal.Thisstudycould provide usefulquantitiesfor modellingand automating [13,30] peripheralnerve insertion tasks. Aboveall,this framework can beextendedtoallcasesinwhichtissuedimplingissmallerthan theproductbetweencharacteristicrelaxationtimeand piercing velocity:forlongrelaxationtime coupledwithquitelow inser-tionvelocity,orshortrelaxationtimecoupledwithhighinsertion velocity.
Futureworkwillextendtheexperimentaltrialstootherliving nervesfeaturingdifferentkindsofneedlesanddiverse combina-tionsofrelaxationtimeandpiercingvelocity,toexplicitlyaccount forviscoelasticeffects.
Acknowledgements
TheauthorsthankDr.JacopoCarpanetoforhisvaluablehelp. ThisworkwassupportedinpartbytheEuropeanUnion(EU)within theTIMEProject(FP7-ICT–2007-224012,Transverse, Intrafascicu-larMultichannelElectrodesystemforinductionofsensationand treatmentofphantomlimbpaininamputees).
Conflictofinterest None.
References
[1]NathS,ChenZ,YueN,TrumporeS,PeschelR.Dosimetriceffectsofneedle diver-genceinprostateseedimplantusing[sup125]Iand[sup103]Pdradioactive seeds.MedPhys2000;27:1058–66.
[2]GuptaS,MadoffDC.Image-guidedpercutaneousneedlebiopsyincancer diag-nosisandstaging.TechVascIntervRadiol2007;10:88–101.
[3]GokselO,DehghanE,SalcudeanSE.MissedbreastcancersatUS-guidedcore needlebiopsy:howtoreducethem.MedEngPhys2009;31:1069–78. [4]AbolhassaniN,PatelR,MoallemM.Needleinsertionintosofttissue:asurvey.
MedEngPhys2007;29:413–31.
[5] GandeviaSC,HalesJP.Themethodologyandscopeofhuman microneurogra-phy.JNeurosciMeth1997;74:123–36.
[6]VallboAB,HagbarthKE,WallinBG.Microneurography:howthetechnique developedanditsroleintheinvestigationofthesympatheticnervoussystem. JApplPhysiol2004;96:1262–9.
[7]DiMaioSP,SalcudeanSE.Needlesteeringandmotionplanninginsofttissues. IEEETransBiomedEng2005;52:965–74.
[8]DiMaioSP,SalcudeanSE.Interactivesimulationofneedleinsertionmodels. IEEETransBiomedEng2005;52:1167–79.
[9] GokselO,DehghanE,SalcudeanSE.Modelingandsimulationofflexibleneedles. MedEngPhys2009;31:1069–78.
[10] YoshidaK,LewinskyI,NielsenM,HyllebergM.Implantationmechanicsof tungstenmicroneedlesintoperipheralnervetrunks.MedBiolEngComput 2007;45:413–20.
[11]JensenW,YoshidaK,HofmannUG.Invivoimplantmechanicsofsingle-shaft microelectrodesinperipheralnervoustissue.In:Proceedingsofthe3rd inter-nationalIEEE/EMBSconferenceonneuralengineering.2007.p.1–4. [12] JuM-S,LinC-CK,FanJ-L,ChenR-J.Transverseelasticityandbloodperfusionof
sciaticnervesunderinsitucircularcompression.JBiomech2006;39:97–102. [13] ReedKB,OkamuraAM,CowanNJ.Modelingandcontrolofneedleswith
tor-sionalfriction.IEEETransBiomedEng2009;56:2905–16.
[14] OkamuraAM,SimoneC,O’LearyMD.Forcemodelingforneedleinsertioninto softtissue.IEEETransBiomedEng2004;51:1707–16.
[15]MahvashM,DupontPE.Mechanicsofdynamicneedleinsertionintoabiological material.IEEETransBiomedEng2010;57:934–43.
[16]SneddonIN.Boussinesq’sproblemforarigidcone.ProcCambridgePhilosSoc 1948:492–507.
[17]SneddonIN.Therelationbetweenloadandpenetrationintheaxisymmetric Boussinesqproblemforapunchofarbitraryprofile.IntJEngSci1965;3:47–57. [18]PoonB,RittelD,RavichandranG.Ananalysisofnanoindentationinlinearly
elasticsolids.IntJSolidsStruct2008;45:6018–33.
[19]MillesiH,ZochG,ReihsnerR.Mechanicalpropertiesofperipheralnerves.Clin OrthopRelatRes1995:76–83.
[20]TimoshenkoS.Strengthofmaterials.D.VanNostrandCompanyInc;1941. [21]DavisSP,LandisBJ,AdamsZH,AllenMG,PrausnitzMR.Insertionof
micronee-dlesintoskin:measurementandpredictionofinsertionforceandneedle fractureforce.JBiomech2004;37:1155–63.
[22]JuMS,LinCCK,LinCW.Transverseelasticityofrabbitsciaticnervesunderin vitro.compression.JChinInstEng2004;7:965–71.
[23]Pailler-MatteiC, BecS,ZahouaniH.Invivo measurementsoftheelastic mechanicalpropertiesofhuman skinbyindentationtests.MedEngPhys 2008;30:599–606.
[24] TöyräsJ,Lyyra-LaitinenT,NiinimäkiM,LindgrenR,NieminenMT,KivirantaI, etal.EstimationoftheYoung’smodulusofarticularcartilageusingan arthro-scopicindentationinstrumentandultrasonicmeasurementoftissuethickness. JBiomech2001;34:251–6.
[25] Gottsauner-WolfF,GrabowskiJJ,ChaoEY,AnKN.Effectsoffreeze/thaw con-ditioningonthetensilepropertiesandfailuremodeofbone-muscle-bone units:abiomechanicalandhistologicalstudyindogs.JOrthopRes1995;13: 90–5.
[26] WellsSM,LangilleBL,AdamsonSL.Invivoandinvitromechanicalproperties ofthesheepthoracicaortaintheperinatalperiodandadulthood.AmJPhysiol 1998;274(5Pt2):H1749–60.
[27] StidhamDB,StagerDR,KammKE,GrangeRW.Stiffnessoftheinferioroblique neurofibrovascularbundle.InvestOphthalmolVisSci1997;38:1314–20. [28]SergiPN,CarrozzaMC,DarioP,MiceraS.Biomechanicalcharacterization
ofneedlepiercingintoperipheralnervoustissue.IEEETransBiomedEng 2006;53:2373–86.
[29]WuJZ,DongRG,SchopperAW.Analysisofeffectsoffrictiononthe defor-mationbehaviorofsofttissuesinunconfinedcompressiontests.JBiomech 2004;37:147–55.
[30]FamaeyN,VerbekenE,VinckierS,WillaertB,HerijgersP,VanderSlotenJ. Invivosofttissuedamageassessmentforapplicationsinsurgery.MedEng Phys2010;32:437–43.