Network structure of the human musculoskeletal system ... · 08.01.2018 · The human body is a complex system consisting of many subsystems and regulatory pathways. The musculoskeletal

Networkstructureofthehumanmusculoskeletalsystem

shapesneuralinteractionsonmultipletimescales

JenniferN.Kerkmana,AndreasDaffertshofera,LeonardoL.Gollob,Michael

Breakspearb,c,TjeerdW.Boonstrab,d,1

aDepartmentofHumanMovementSciences,FacultyofBehaviouralandMovement

Sciences,VrijeUniversiteitAmsterdam,AmsterdamMovementSciences&InstituteforBrainandBehaviorAmsterdam,Amsterdam,theNetherlands

bQIMRBerghoferMedicalResearchInstitute,Brisbane,AustraliacMetroNorthMentalHealthService,Brisbane,QLD,Australia

dBlackDogInstitute,UniversityofNewSouthWales,Sydney,Australia

1Correspondingauthor

TjeerdBoonstra,BlackDogInstitute,HospitalRd,SydneyNSW2031,Australia.

Email:[email protected]

not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (which wasthis version posted January 8, 2018. ; https://doi.org/10.1101/181818doi: bioRxiv preprint

https://doi.org/10.1101/181818

2

Abstract

Humanmotorcontrolrequiresthecoordinationofmuscleactivityunderthe

anatomicalconstraintsimposedbythemusculoskeletalsystem.Interactions

withinthecentralnervoussystemarefundamentaltomotorcoordination,butthe

principlesgoverningfunctionalintegrationremainpoorlyunderstood.Weused

networkanalysistoinvestigatetherelationshipbetweenanatomicaland

functionalconnectivityamongst36muscles.Anatomicalnetworksweredefinedby

thephysicalconnectionsbetweenmusclesandfunctionalnetworkswerebasedon

intermuscularcoherenceassessedduringposturaltasks.Wefoundamodular

structureoffunctionalnetworksthatwasstronglyshapedbytheanatomical

constraintsofthemusculoskeletalsystem.Musclenetworksexhibitedamultilayer

architecturewithfunctionalinteractionsatdistincttimescales.Changesinpostural

taskswereassociatedwithafrequency-dependentreconfigurationofthecoupling

betweenfunctionalmodules.Combined,thesefindingssuggestthesespectral

modesmaysignifyacoordinativestructureforflexiblyorganisingmuscleactivity

duringposturalcontrol.Morebroadly,ourmulti-levelnetworkapproachtothe

motorsystemoffersauniquewindowintothecoordinatedneuralcircuitrythat

generatessynergisticinputtomusclesindifferentbehaviours.


https://doi.org/10.1101/181818

3

Introduction

Thehumanbodyisacomplexsystemconsistingofmanysubsystemsand

regulatorypathways.Themusculoskeletalsystemgivesthebodystructureand

createstheabilitytomove.Itismadeupofmorethan200skeletalbones,

connectivetissueandover600skeletalmuscles[1].Musclesareattachedtobones

throughtendinoustissueandcangeneratemovementaroundajointwhenthey

contract.Thecentralnervoussystemcontrolsthesemovementsthroughthespinal

motorneurons,whichserveasthefinalcommonpathwaytothemuscles[2].

Whiletheanatomicalandphysiologicalcomponentsofthemusculoskeletalsystem

arewellcharacterized[3,4],theorganisationalprinciplesofneuralcontrolremain

poorlyunderstood.Hereweelucidatetheinterplaybetweentheanatomical

structureofthemusculoskeletalsystemandthefunctionalorganisationof

distributedneuralcircuitryfromwhichmotorbehavioursemerge.

Thetraditionalideathatthecortexcontrolsmusclesinaone-to-onefashionhas

beenchallengedbyseverallinesofevidence[5,6].Forexample,itiswidely

recognizedthatthemanydegrees-of-freedom(DOFs)ofthemusculoskeletal

systemprohibitasimpleone-to-onecorrespondencebetweenamotortaskanda

particularmotorsolution;rathermusclesarecoupledandcontrolledin

conjunction[7].Acouplingbetweenmuscles–whethermechanicalorneural–

reducesthenumberofeffectiveDOFsandhencethenumberofpotential

movementpatterns.Thiscouplingtherebyreducesthecomplexityofmotor

control[8,9].

Thereiscontinuingdebateaboutthenatureofthecouplingbetweenmuscles.The

mechanicalcouplinginthemusculoskeletalsystemconstrainsthemovement

patternsthatcanbegenerated[10,11].Forexample,thebiomechanicsofthelimb

constrainrelativechangesinmusculotendonlengthtoalowdimensional

subspace,resultingincorrelatedafferentinputstospinalmotorneurons[12].The

couplingbetweenmusclescouldalsoresultfromredundanciesintheneural

circuitrythatdrivesspinalmotorneurons[13].Electrophysiologicalstudiesreveal

thatacombinationofonlyafewcoherentmuscleactivationpatterns–ormuscle

synergies–cangenerateawidevarietyofnaturalmovements[14,15].Someof


https://doi.org/10.1101/181818

4

thesepatternsarealreadypresentfrombirthanddonotchangeduring

development,whereasotherpatternsarelearned[16].Thissupportsthenotion

thattheneuromuscularsystemhasamodularorganisationthatsimplifiesthe

controlproblem[17,18].Spinalcircuitryconsistingofanetworkofpremotor

interneuronsandmotorneuronsmaygeneratebasicmovementpatternsby

mediatingthesynergisticdrivetomultiplemuscles[19,20].Stimulationof

interneuronalregionsinthespinalcordshowsthatthesemicrocircuitsare

organisedintodiscretemodules,eachgeneratingaspecificpatternofmuscular

forces[21,22].Thesespinalnetworksmayencodecoordinatedmotoroutput

programs[23],whichcanbeusedtotranslatedescendingcommandsformulti-

jointmovementsintotheappropriatecoordinatedmusclesynergiesthatunderpin

thosemovements[4].

Networktheorycanprovideanalternativeperspectiveonthemodular

organisationofthemusculoskeletalsystem.Oneofthemostrelevantfeaturesof

complexnetworksarecommunityormodularstructures,whichrefertodensely

connectedgroupsofnodeswithonlysparserconnectionsbetweenthesegroups

[24,25].Theinvestigationofcommunitystructureshasbeenwidelyusedin

differentdomainssuchasmetabolic[26]andbrainnetworks[27].Ithasrecently

beenappliedtoinvestigatethestructureandfunctionofthemusculoskeletal

system:Theanatomicalnetworkcanbeconstructedbymappingtheoriginand

insertionofmuscles[28,29].Previously,wehaveshownhowfunctionalmuscle

networkscanbeconstructedbyassessingintermuscularcoherencefromsurface

electromyography(EMG)recordedfromdifferentmuscles[30].Thesefunctional

networksrevealfunctionalconnectivitybetweengroupsofmusclesatmultiple

frequencybands.CoherencebetweenEMGsindicatescorrelatedorcommoninputs

tospinalmotorneuronsthataregeneratedbysharedstructuralconnectionsor

synchronisationwithinthemotorsystem[13,31-33].Functionalconnectivity

patternshenceallowtoassessstructuralpathwaysinthemotorsystemusingnon-

invasiverecordings[34].

Hereweinvestigatetheorganisationalprinciplesgoverninghumanmotorcontrol

bycomparingthecommunitystructureofanatomicalandfunctionalnetworks.We


https://doi.org/10.1101/181818

5

usemultiplexmodularityanalysis[35]toassessthecommunitystructureof

functionalmusclenetworksacrossfrequenciesandposturaltasks.As

biomechanicalpropertiesofthemusculoskeletalsystemconstrainthemovement

patternsthatcanbegenerated,weexpectasimilarcommunitystructurefor

anatomicalandfunctionalmusclesnetworks.Deviationsincommunitystructure

indicateadditionalconstraintsimposedbythecentralnervoussystem.Wealso

comparefunctionalconnectivitybetweenmodulesduringdifferenttasks

conditionstoinvestigatechangesinfunctionalorganisationduringbehaviour.

Whileaveragefunctionalconnectivityisconstrainedbyanatomicalconstraints,we

expectthatfunctionalmusclenetworksreconfiguretoenabletask-dependent

coordinationpatternsbetweenmuscles.Suchtaskmodulationswouldindicate

thatfunctionalinteractionsbetweenmusclesarenothard-wired,butareinstead

governedbydynamicconnectivityinthecentralnervoussystemthatisshapedby

theanatomicaltopologyofthemusculoskeletalsystem.

Results

Anatomicalmusclenetwork

Weassessedtherelationshipbetweenanatomicalandfunctionalconnectivityof

keymusclesinvolvedintheposturalcontroltasks(36musclesdistributed

throughoutthebody).Weinvestigatedamuscle-centricnetworkinwhichthe

nodesrepresentthemusclesandtheedgesofthenetworkareanatomical

connectionsorfunctionalrelationsbetweenmuscles.Anatomicalmusclenetworks

weredefinedbymappingthephysicalconnectionsbetweenmusclesandbones

[28],whichformabipartitenetwork[29].Theanatomicalnetworkconstituteda

denselyconnected,symmetricalnetwork(Fig.1;networkdensityis0.27).

Modularityanalysisrevealedfivemodulesthatdividedtheanatomicalmuscle

networkintothemainbodysegments(rightarm,leftarm,torso,rightlegandleft

leg)withamodularityof0.38.


https://doi.org/10.1101/181818

6

Figure1.Communitystructureoftheanatomicalmusclenetwork.A)Topological

representationoftheanatomicalnetwork.Thenodesofthenetworkrepresentthemusclesandedgesrepresentanatomicalconnectionsbetweenmusclesthatareattachedtothesamebonesorcartilages.Thefivemodulesarecolour-coded.B)Spatialrepresentationofanatomicalmusclenetworkdisplayedonthehumanbody[36].Thesize

ofeachnoderepresentsthenumberofothernodesitisconnectedto.

Functionalmusclenetwork

Functionalmusclenetworksweredefinedbymappingcorrelatedinputsbetween

muscles.Tomapfunctionalnetworks,wemeasuredsurfaceEMGfromthesame36

muscleswhilehealthyparticipantsperformeddifferentposturaltasks.Afull

factorialdesignwasusedinwhichwevariedposturalcontrol(normalstanding,

instabilityinanterior-posteriorormedial-lateraldirection)andpointing

behaviour(nopointing,pointingwiththedominanthandorwithbothhands;see

Methodsfordetails).Weassessedfunctionalconnectivitybymeansof

intermuscularcoherencebetweenallmusclecombinationsandusednon-negative

matrixfactorisation(NNMF)todecomposethesecoherencespectraintofrequency

componentsandcorrespondingedgeweights.Thisyieldedasetofweighted

networkswiththeircorrespondingspectralfingerprints(frequencycomponents).


https://doi.org/10.1101/181818

7

Weobservedfourseparatefrequencycomponents(component1:0-3Hz,

component2:3-11Hz,component3:11-21Hz,component4:21-60Hz;Fig.2A),

whichserveasseparatelayersofamultiplexnetworkandexplainedmostofthe

varianceofthecoherencespectra(R2=0.90).Weightswerethresholdedtoobtain

aminimallyconnectedbinarynetworkacrosslayersandtokeepthenumberof

edgesconstantacrosslayers(relativethresholdof0.035).Usingmultiplex

modularityanalysis,weobtainedafixedcommunitystructureacrossallfour

frequenciesandnineconditions,whichrevealedsixmodules:rightupperarm

(rUA),bilateralforearms(FA),torso(T),rightupperleg(rUL),leftupperleg(lUL)

andbilaterallowerlegs(LL)(Fig.2B).Figure2Cdepictshowthesemodulesare

distributedacrossthebody.Distinctnetworktopologieswereobservedacross

layerswithamorewidelyconnectednetworkatlowerfrequenciesandmore

partitionednetworkathigherfrequencies:networkdensitywas0.10,0.09,0.08,

and0.06forcomponents1to4,respectively(Fig.2D).

Figure2.Communitystructureofmultiplexfunctionalmusclenetworks.A)ThefrequencyspectraofthefourcomponentsobtainedusingNNMF.B)Multiplexcommunitystructureoffunctionalmusclenetworkacrosslayersandconditions.Thedominanthand

ofallparticipantsisdisplayedontherightsideofthehumanbody.C)Spatialrepresentationoftheaveragemusclenetworkdisplayedonthehumanbody[36].Thesizeofthenodesrepresentsthenumberofothernodesitisconnectedtoandthewidthoftheedgestheaverageweight.D)Thebinarymusclenetworksforeachlayer.


https://doi.org/10.1101/181818

8

Comparisonbetweenanatomicalandfunctionalnetworks

Thecommunitystructuresoftheanatomicalandfunctionalmusclenetworkswere

verysimilar(Randindex=0.80,adjustedRandindex=0.36,P<0.001).Amarked

differencebetweenanatomicalandfunctionalnetworksisthebilateral

connectionsbetweenhomologousforearmandlowerlegmusclesinthefunctional

networks,whichwereabsentintheanatomicalnetwork.Thisisreflectedinthe

communitystructureofthefunctionalnetworks,wherebilaterallowerlegmuscles

andbilateralforearmmusclesweregroupedinmodules.

Figure3.Relationshipbetweenfunctionalconnectivityandanatomicaldistance.A)Adjacencyanddistancematrixoftheanatomicalmusclenetwork.Maximumanatomicaldistance(pathlength)is4.B)Percentageoffunctionaledgesofthresholdednetworksacrossexperimentalconditionsasafunctionofanatomicaldistance.C)Distributionof

edgeweightsoffunctionalnetworksasafunctionofanatomicaldistanceforeachlayer.Weightswereaveragedacrossexperimentalconditions.Edgesconnectingmuscleswithinthesamemodulearecolour-coded(rUA:rightupperarm,FA:bilateralforearms,T:torso,rUL:rightupperleg,lUL:leftupperleg,andLL:bilaterallowerlegs)andgreydots

representedgesbetweenmodules.


https://doi.org/10.1101/181818

9

Thecomparisonbetweenanatomicaldistance(pathlength)andfunctional

connectivityrevealedthatanatomicallynearbynodesaremorelikelytoreceive

commoninput(Fig.3A).Wefirstexaminedthepercentageofallpossibleedges,i.e.

thenumberofedgesabovethreshold,whichdecreasedasafunctionofanatomical

distance:11.3%,0.9%,0.6%and0.0%foranatomicaldistance1to4,respectively.

Thisdeclinewithdistancewasevenmorepronouncedforthehigherfrequency

components(Fig.3B).Next,weexaminedthedistributionoffunctionalweightsas

afunctionofanatomicaldistance.Thehighestweightswereobservedforedges

connectingmuscleswithinthesamemodule.Theedgeswithinmostmoduleshad

ananatomicaldistanceof1.Onlyafewedgeshadananatomicaldistanceof2or3

andalloftheseedgeswerecontainedintheFAandLLmodules.Inparticular,

edgesconnectingbilaterallowerlegmuscles(LL)showedrelativelargeweightsat

ananatomicaldistanceof3(Fig.3C).

Task-dependentmodulations

Wenextsoughttostudytheinfluenceoftaskonthisstructure-function

relationship.Thiswasachievedbyemployingclusteredgraphstocompare

functionalmusclenetworksacrosstaskconditions.Thefunctionalmodules

identifiedusingtheprecedingmultiplexmodularityanalysisformthenodesof

theseclusteredgraphs.Figure4Ashowstheclusteredgraphsinthenine

experimentalconditionsandforthefourlayers.Theclusteredgraphswerevery

sparse,asmoduleshavedensewithinmoduleconnectionsbutsparseconnections

betweennodesinothermodules.Mostedgeswereobservedbetweenlegmuscles

modules(LL,rULandlUL)atthelowestfrequencycomponents(0-3and3-11Hz),

inparticularwhenposturalstabilitywaschallengedbyinstabilityinanterior-

posteriorormedial-lateraldirection.Edgesbetweenthearmmusclemodules(rUA

andFA)andthetorso(T)weremainlyobservedatthehigherfrequency

components(11-21and21-60Hz)duringpointing(unimanualandbimanual).


https://doi.org/10.1101/181818

10

Figure4.Clusteredgraphsoffunctionalmusclenetworksacrossconditions.A)Theclusteredgraphsinthenineexperimentalconditions(columns)andforeverylayer

(rows).Thenodesarethemodulesidentifiedusingmultiplexmodularityanalysis.Nodesizerepresentsthenetworkdensitywithinandthewidthoftheedgestheconnectiondensitybetweenmodules.B)Spatialrepresentationofthefunctionalmodulesonthehumanbody:rightupperarm(rUA),bilateralforearms(FA),torso(T),rightupperleg(rUL),leftupperleg(lUL)andbilaterallowerlegs(LL).Weusedtoolboxesforgeometry

processingtogeneratethecolouredmeshes[37]anddisplayitonthehumanbody[36].C)Significantdifferencesinclusteredgraphsbetweenthestabilityconditions.Twocontrastswereassessed:normalstability–anterior-posteriorinstabilityandnormalstability–medial-lateralinstability.Apermutationtestwasusedwithasignificancethreshold

correctedformultiplecomparisons(P<0.0033).Significantdifferencesarecolour-coded:Reddepictsanincreaseandblueadecreaseintheaverageweights.D)Significantdifferencesinclusteredgraphsbetweenthepointingconditions.Twocontrastswereassessed:nopointing–unimanualpointingandnopointing–bimanualpointing.


https://doi.org/10.1101/181818

11

Theeffectsofthestabilitytaskswerelargelyconfinedtothelegmodules(Fig.4C).Increasedconnectivitywasobservedduringposturalinstability(anterior-posteriorandmedial-lateral)comparedtonormalstandingwithinmostfrequencycomponents.Atthelowestfrequencycomponent(0-3Hz),connectivityincreasedwithinandbetweenmostlegmodules(pcorrected<0.01).Onlysmalldifferenceswereobservedat3-11Hz:increasedconnectivitybetweenthetorso(T)andlowerleg(LL)modulesduringanterior-posteriorinstability(+25%,range[-9,46%],pcorrected=0.01)anddecreasedconnectivitywithinthetorsomoduleduringmedial-lateralinstability(-21%,range[-50,0.3%],pcorrected=0.01).Connectivityincreasedagainatthehighestfrequencycomponents(11-21and21-60Hz)withinandbetweenthetorsoandlegmodules(rUL,lUL,andLL,pcorrected<0.02).

Thepointingtasksshowedadifferentpatterncomparedtotheposturaltasks,but

theeffectsofunimanualandbimanualpointingwereverysimilar(Fig.4D).During

pointing,connectivitydecreasedwithinthetorso(T)moduleatthelowest

frequencycomponents(0-3Hz,-61%,range[-90,-1%],pcorrected<0.005;3-11Hz,-

59%,range[-86,2%],pcorrected<0.02)andbetweenthetorsoandtherightupper

arm(rUA)moduleonlyatthelowestfrequencycomponent(0-3Hz,-67%,range[-

93,-9%],pcorrected<0.005).Incontrast,asignificantincreaseinconnectivitywithin

therUAmodulewasobservedduringunimanualpointcomparedtonopointingat

thehighestfrequencycomponents(11-21Hz,+64%,range[-4,95%],pcorrected=

0.005;21-60Hz,+66%,range[-12,93%],pcorrected=0.015).Inaddition,therewas

increasedconnectivitybetweenthetorsoandtheforearm(FA)modules(+41%,

range[-8,82%],pcorrected<0.01)andbetweenrUAandFA(+44%,range[0,82%],

pcorrected<0.005)duringpointing(unimanualandbimanual)comparedtono

pointingatfrequencycomponent3(11-21Hz).

Discussion

Weusedanetworkapproachtostudythestructure-functionrelationshipofthe

humanmusculoskeletalsystem.Severalprinciplesofthefunctionalrelationship

betweenmuscleswereuncovered:(i)Functionalconnectivitypatternsbetween

musclesarestronglyshapedbytheanatomicalconstraintsofthemusculoskeletal

system,withfunctionalconnectivitystrongestwithinanatomicalmodulesand

decreasedasafunctionofanatomicaldistance;(ii)Bilateralconnectivitybetween


https://doi.org/10.1101/181818

12

thehomologousupperandlowerextremitiesisakeycharacteristicofthe

functionalmusclenetworks;(iii)Thefunctionalrelationshipsaretask-dependent

withposturaltasksdifferentiallyimpactinguponfunctionalconnectivityat

differentfrequencyranges.Theuseofamultiplexapproachallowstheintegration

offunctionalmusclenetworksacrossfrequenciesandprovidesaunifyingwindow

intothedistributedcircuitryofthehumancentralnervoussystemthatcontrols

movementsbyinnervatingthespinalmotorneurons.

Identifyingrelationshipsbetweenanatomicalandfunctionalmusclenetworksis

crucialforunderstandinghowmovementiscoordinated.Previousstudieseither

investigatedhowbiomechanicalpropertiesofthemusculoskeletalsystem

constrainthemovementpatternsthatcanbegenerated[11,12],orhowmuscle

activationpatternscanbeexplainedbyacombinationofonlyafewcoherent

muscleactivationpatterns[14,15].Ourcombinedanalysesofanatomicaland

functionalmusclenetworksrevealastrongrelationshipbetweentheanatomical

connectionsinthemusculoskeletalsystemsandcorrelatedinputstospinalmotor

neurons.Thisbuildsonpreviousresearchshowingthatcommoninputisstrongest

tospinalmotorneuronsthatinnervatemusclespairsthatareanatomicallyand

functionallycloselyrelated[13,31].Thesimilaritybetweenstructuraland

functionalnetworkshasbeenasignatureofthestudyofbrainnetworks[38]and

thetopologyofbrainnetworksdependsonthebrain'sspatialembedding[39,40].

Thepresentfindingssuggestthattheprinciplesgoverningembodiedstructural

andfunctionalnetworksalsoappliestotheneuralcircuitrythatcontrols

movementsandmayhencereflectageneralprincipleofthecentralnervous

system.

Thesimilaritiesbetweenanatomicalandfunctionalconnectivitymayindicatethat

theanatomicalstructureconstrainsthefunctionalinteractionsbetweenmuscles.

Theanatomicalconnectionsbetweenmusclesremainlargelyunchangedoverthe

lifespan[41]anditismorelikelythatthefast-changingfunctionalnetworksare

constrainedbythemuchslowerchanginganatomicalnetworksthanviceversa.

Theseconstraintsmaybeimposedthroughafferentactivity.Themusculoskeletal

propertiesofthehumanbodyrestricttheposturaldynamics[12]andthese


https://doi.org/10.1101/181818

13

mechanicalcouplingsresultincorrelatedproprioceptivefeedbacktospinalmotor

neurons.Theinfluenceofbiomechanicsonfunctionalmusclenetworksisexpected

tobemostpronouncedatthelowerfrequencycomponents:Musclesactasalow

passfilterofneuronalinputs[43]andkinematicsofthemusculoskeletalsystem

unfoldonaslowtimescale.Thisgeneratescorrelatedactivityatlowfrequencies

thatarefedbacktospinalmotorneuronsviasensoryafferents.Thespatial

distributionofcommoninputwouldarguablymirrorthetopologyofthe

musculoskeletalsystem.

Anatomicalconstraintsmayalsobeimposedduringneuraldevelopment.During

earlydevelopment,changesinthetopographicaldistributionofaxonterminalsof

descendingprojectsaredependentonpatternsofmotoractivityandanatomical

connectivitybetweenmuscles[44].Likewise,largechangesinfunctionalcoupling

isobservedininfantsbetween9and25weeks,whichmayreflectasensitive

periodwherefunctionalconnectionsbetweencorticospinaltractfibresandspinal

motorneuronsundergoactivity-dependentreorganization[45].Theanatomyof

themusculoskeletalsystemwilllimitthemotoractivitypatternsthatcanbe

performed.

Anatomicalandfunctionalconnectivitybetweenmusclesmayalsobothbe

influencedbyexternalfactors.Forexample,theconnectivitypatternsof

descendingpathwaysisinpartgeneticallydetermined[46].Asomatotopic

organisationisobservedacrosstheneuralmotorsystemandthecommunity

structureoftheanatomicalmusclenetworkmirrorstheorganisationofprimary

motorcortexcontrolmodules[29].Likewise,thespatialorganisationofmotor

neuronsofthespinalcordisalsorelatedtotheanatomicalorganisationofmuscles

[47].Thetopographicorganisationofspinalmotorneuronsissimilaracross

species[48]andmayhencebearesultofevolutionaryconservation[49].

Musculoskeletalanatomyandneuronalpathwaysarehencebothsubjecttosome

sortofgeneticcontrol.

Functionalconnectivitywasnotentirelydeterminedbyanatomyandweobserved

severalkeydifferencesbetweenanatomicalandfunctionalmusclenetworks.


https://doi.org/10.1101/181818

14

Bilateralmodulesconsistingoftheupperandlowerextremitieswereakey

characteristicofthefunctionalmusclenetworkthatwasabsentintheanatomical

network.Thetwobilateralforearmmuscles(FDSandED)showedcoherent

activityat3-11Hz,consistentwithpreviousstudiesshowingbimanualcouplingat

~10Hzbetweenhomologoushandandforearmmuscles[50,51].Theobserved

bimanualcouplingat3-11Hzmaybegeneratedbytheolivocerebellarsystem,

whichisknowntoproduceoscillationsinthisfrequencyrangeandforits

involvementintheformationoffunctionalmusclecollectives[51,52].The

bilateralforearmmuscleswereonlyweaklycoupledtoothermuscles(Fig.2),

whichmayreflecttherelativelyhighproportionofdirectcorticospinalprojections

–andthusarelativelowproportionofdivergingprojections–tomotorneurons

innervatinghandandforearmmuscles[53].

Incontrast,thebilateralmoduleoflowerlegmusclesrevealedstrongcouplingat

multiplefrequencybands,consistentwithpreviousanalysesonfunctionalmuscle

networks[30],andshowedthestrongestlong-rangeconnectionsobservedinthe

presentstudy(Fig.3C).Bilateralconnectivitybetweenhomologousmusclesduring

balancingcouldbegeneratedbythevestibulospinaltract,whichisknowntobe

involvedinposturalstabilityandinnervatethespinalgreymatterbilaterally[31].

Bilateralconnectivityhasbeenobservedatalllevelsofthecorticospinalaxis[54]

andisparamountforfunctionalbrainnetworks,particularlybetweenhomologous

left-rightcorticalregions[55,56].Thepresentfindingssuggestthatbilateral

couplingisalsoadefiningfeatureoffunctionalmusclenetworks.

Functionalconnectivitydisplayeddistincttask-dependentmodulationsthatwere

linkedtothetaskthesubjectsperformed:functionalconnectivitywasincreased

withinandbetweenthelegmodulesduringposturalinstability,andincreased

withinandbetweenarmandupperbodymodulesinthepointingconditions.

Functionalconnectivitybetweenmusclesisthustaskdependent[31,50],which

maysuggestthepresenceofmultifunctionalcircuitsinwhichagivenanatomical

connectivitypatterncangeneratedifferentfunctionalactivitypatternsunder

variousconditions[42].Suchadistributedcircuitrycreatesthesubstrateto

supportmanybehavioursthataredrivenbytheconcertedactionsofalarge


https://doi.org/10.1101/181818

15

distributednetworkasopposedtosimple,dedicatedpathways.Theunderlying

networkconnectivityhenceconstrainsthepossiblepatternsofpopulationactivity

toalow-dimensionalmanifoldspannedbyafewindependentpatterns–neural

modes–thatprovidethebasicbuildingblocksofneuraldynamicsandmotor

control[57].Again,thisfindssimilaritieswithrecentinvestigationsofthe

functionalprinciplesofcognitivenetworksinthebrain[58,59].

Task-dependentchangesoccurredatdifferentfrequencies,whichindicatethe

functioningofamultiplexnetworkorganisation,wherebythefourfrequency

componentsreflectdifferenttypesofinteractionsbetweenmuscles.Fourdistinct

frequencycomponents(0-3,3-11,11-21,and21-60Hz)wereextractedusing

NNMF.Thesefrequencybandscloselymatchthosefoundpreviously[30],

demonstratingtherobustnessofthisfinding.Aninterestingpossibilityisthat

thesefrequencycomponentsreflectthespectralfingerprintsofdifferentpathways

thatprojectontothespinalmotorneurons.Ithasbeensuggestedthatthese

differentfrequenciesmayhavespecificrolesincodingmotorsignals[60].

Functionalconnectivityatthelowestfrequencycomponentsmayresultfrom

afferentpathways,whilefunctionalconnectivityathigherfrequenciesmayreflect

correlatedinputfromdescendingpathways.Forexample,functionalconnectivity

inthebetaband(15-30Hz)mostlikelyreflectscorticospinalprojections[13,50].

Thehighestfrequencycomponentsobservedinthisstudy(21-60Hz)showedthe

mostlocalconnectivitypatterns.Theselocalconnectivitypatternsmayreflect

propriospinalpathways[4,23].Thesefunctionalconnectivitypatternsmaybe

usedtouncoverthecontributionofstructuralpathwaysintheformationof

coordinatedactivitypatternsinthemotorsystem[34].

Insummary,ournetworkanalysisrevealedwidespreadfunctionalconnectivity

betweenmuscles,indicativeofcorrelatedinputstospinalmotorneuronsat

multiplefrequencies.Correlatedinputsindicatedivergentprojectionsorlateral

connectionsintheneuralpathwaysthatinnervatespinalmotorneurons.These

findingsareconsistentwithamany-to-manyratherthananone-to-onemapping

betweenbrainandmuscle[5],inwhichcomplexmovementsarisethrough

relativelysubtlechangesintheco-activationofdifferentdistributedfunctional


https://doi.org/10.1101/181818

16

modes.Wepresentanovelapproachthatalignsmovementneurosciencewith

currentresearchonbrainandphysiologicalnetworksbyshowinghowthecentral

nervoussysteminteractswiththemusculoskeletalsystemofthehumanbody.

Fromasystemsbiologyperspective,thebrainandspinalcordareinterwovenwith

thebody:theyare‘embodied’[10].Usingnetworkanalysisasacommon

framework,futureworkcouldintegratefunctionalinteractionsbetweensupra-

spinal,spinalandperipheralregionssimultaneously.Suchanintegrated

frameworkalignswiththebroaderapproachofnetworkmedicine,whichmay

providenewinsightsandinterventionsforneurologicaldisorders[61-63].

Methods

Dataacquisition

Fourteenhealthyparticipants(sevenmalesandsevenfemales,meanage25±8

years,tenrightandfourlefthanded)withoutanyneurological,motordisorderor

diabetesmellitusandwithaBMIbelow25wereincludedinthisstudy.The

experimentswereapprovedbytheEthicsCommitteeHumanMovementSciences

oftheVrijeUniversiteitAmsterdam(referenceECB2014-78)andperformedinfull

compliancewiththeDeclarationofHelsinki.Allparticipantswerewrittenand

verballyinformedabouttheprocedureandsignedaninformedconsentpriorto

participation.

Participantswereinstructedtoperformninedifferentposturaltasks.Afull

factorialdesignwasusedinwhichposturalstability(normalstanding,instability

inanterior-posteriordirectionandinstabilityinmedial-lateraldirection)and

pointingbehaviour(nopointing,pointingwithdominanthand,pointingwithboth

hands)werevaried.Posturalstabilitywasmanipulatedusingabalanceboardwith

onedegreeoffreedom,whichallowsmovementeitherintheanterior-posterioror

medial-lateraldirection.Inthepointingtask,participantsheldalaserpointerwith

theirdominanthand(unimanual)orwithbothhands(bimanual)andpointediton

awhitetarget(25cm2)locatedatadistanceof2.25meter,paralleltothe

transversalaxisofthebodyattheheightoftheacromionoftheparticipant.The

experimenthenceconsistedofnine(3´3)experimentalconditions.Thedurationof

atrialwas30secondsandeachconditionwasrepeatedsixtimes.


https://doi.org/10.1101/181818

17

BipolarsurfaceEMGwasrecordedfrom36musclesdistributedacrossthebody,

i.e.18bilateralmuscles(Table1).Thesemusclesaretheprimarymusclesinvolved

inthestabilityandpointingtasksthatcanbeproperlymeasuredwithsurface

EMG.EMGwasacquiredusingthree16-channelPortisystems(TMSi,Enschede,

TheNetherlands),onlinehigh-passfilteredat5Hzandsampledat2kHz.

Table1.Listofmuscles

Muscle Abbreviationtibialisanterior TAgastrocnemiusmedialis GMsoleus SOLrectusfemoris RFbicepsfemoris BFvastuslateralis VLadductorlongus ALexternaloblique EOpectoralismajor PMAsternocleidomastoideus SMAlongissimus LOlatissimusdorsi LDtrapezius TZdeltoid Dbicepsbrachii BBtricepsbrachii TRBextensordigitorum EDflexordigitorumsuperficialis FDS

Anatomicalmusclenetwork

Theanatomicalmusclenetworkwasdefinedbymappingthephysicalconnections

betweenmuscles[28].Thenodesrepresentthe36muscles(18leftand18right)

andtheedgesofthenetworkrepresentthetendinousattachmentsofmusclesonto

bonesandcartilages.Thestructuralconnectionsweredefinedbasedontheorigin

andinsertionofthemuscles[3].Bonesthatshownooralmostnomotioninthe

jointbetweenthemwereconsideredasonerigidbonystructure,i.e.thepelvis,

skeletonthoracisorossacranii.Theconnectionsbetweenmusclesandbones

listedinTable2denoteabipartitenetwork𝐶withmusclesasonegroupand

bonesasthesecondgroup.Wethencreatedamuscle-centricnetworkastheone-

modeprojectionsof𝐶:𝐵 = 𝐶𝐶% [29].Thisgaveaweightedadjacencymatrix

wheretheweightsreflectthenumberofattachmentsbywhichtwomusclesare


https://doi.org/10.1101/181818

18

connected.Weconvertedthistoabinarynetworkbysettingallnone-zeroweights

to1.

Table2.OriginandinsertionofmusclesMuscle Origin1 Origin2 Origin3 Insertion1 Insertion2TA tibia oscuneiformemediale ossa

metatarsiGM femur calcaneus SOL fibula tibia calcaneus RF oscoxae1,* osilium1,* tibia BF femur osischii1,* fibula tibiaVL femur tibia AL ospubis1,* femur EO costae2,* lineaalba osilium1,*PMA clavicula costae2,* sternum2,* humerus SMA clavicula sternum2,* ostemporale3,* LO ligamentum

sacrospinale1,*vertebra* costae2,* vertebra*

LD costae2,* fasciathoracolumbalis*

vertebra* humerus

TZ ligamentumnuchae* osoccipitale3,* vertebra* clavicula scapulaD clavicula scapula humerus BB scapula radius TRB humerus scapula ulna ED humerus ossadigitorum FDS humerus radius ulna ossadigitorum 1Partofthepelvis;2Partoftheskeletonthoracis;3Partoftheossacranii;*Connectivestructureonthemidlineofthebodyconnectingbilateralmuscles

Functionalmusclenetwork

EMGwasfilteredofflinewithaband-passfilter(1-400Hz)beforeindependent

componentanalysiswasusedforelectrocardiographyremoval[64].Foreach

participant,oneortwoindependentcomponentswereremoved.EMGdatawas

thenhigh-passfiltered(20Hz)andEMGenvelopeswereextractedbymeansofthe

Hilbertamplitude[33].

WefollowedtheproceduredescribedinBoonstraetal.[30]toextractfunctional

musclenetworksfromsurfaceEMG.First,complex-valuedcoherencywas

estimatedandaveragedovertrialswithineachconditionforeachparticipant.The

absolutevalueofcoherencywassquaredtoobtainmagnitude-squaredcoherence.

Intermuscularcoherencewasassessedbetweenall630musclepairs.Next,non-

negativematrixfactorisation(NNMF)[65]wasusedtodecomposethese

coherencespectraacrossallmusclecombinations,conditionsandparticipantsinto

fourdistinctfrequencycomponentsandthecorrespondingweights.Thisyieldeda


https://doi.org/10.1101/181818

19

setofweightsforeachfrequencycomponent,whichdefinedanundirected

weightednetworkforeachconditionandparticipant.

Thesefunctionalnetworkswereconvertedtobinarynetworkstofacilitate

comparisontotheanatomicalnetwork.Weightswerethresholdedtoobtaina

minimallyconnectednetworkacrossconditionsandfrequencycomponents.This

thresholdingprocedureyieldsasingle,uniquethresholdvalue,whichcorresponds

tothepercolationthreshold[66].Thisresultedinsparsenetworksinwhicheach

nodewasconnectedtoatleastoneothernodebyanedgeatoneofthelayersof

themultiplexnetwork.

Communitystructure

ThestandardLouvainalgorithmwasusedtoextractthemodulesfromthe

anatomicalnetworks[67].AstheLouvainalgorithmisstochastic,weused

consensusclusteringtoobtainastablepartitionacross1000iterations[68].

Multiplexmodularityanalysis[35]wasusedtoidentifythemodulesoffunctional

musclenetworkacrosstheconditionsandfrequencycomponents.WeusedMolTi,

astandalonegraphicalsoftware,todetectcommunitiesfrommultiplexnetworks

byoptimisingthemultiplex-modularitywiththeadaptedLouvainalgorithm

(https://github.com/gilles-didier/MolTi).Moduleswereextractedacrossthe36

(9´4)binarynetworks,asmulti-graphapproachescanimproveclustering

accuracy[69].WeusedtheRandindexandtheadjustedRandindextocompare

themodulesoftheanatomicalandfunctionalmusclenetworks[25].

Comparisonoffunctionalnetworksacrossconditions

Tofacilitatethecomparisonoffunctionalnetworksacrosstaskconditions,we

coarse-grainedthenetworks[70].Weusedthesetoffunctionalmodulesestimated

acrossconditionsandfrequencycomponentsasaframeofreferencetocoarse-

grainthe36binarynetworksandthencomparedthestrengthoftheinter-and

intra-moduleconnectionsacrossnetworksusingthesemoduleboundaries.Inthe

clusterednetworksthenodesrepresentthemodules(groupsofmuscles,identified

above)andtheedgesrepresenttheconnectionsbetweenmodules.Thenon-

diagonalelementsoftheresultingweightedadjacencymatrixrepresentthe


https://doi.org/10.1101/181818

20

averageedgeweightsbetweentwomodulesanddiagonalelementstheaverage

edgeweightswithinamodule.

Tocomparetheclusterednetworksacrossconditions,weusedsimplecontrasts

betweentaskconditionsandquantifieddifferencesinthenumbersofconnections

betweenandwithinmodules.Wetestedfourcontrasts:(i)unimanualand(ii)

bimanualpointingcomparedtonopointingand(iii)anterior-posteriorand(iv)

medial-lateralinstabilitycomparedtonormalstanding.Totestthestatistical

significanceofthesecontrasts,weperformedpairedpermutationtestsseparately

oneachofthematrixelements[70].Theclusterednetworkshadamuch-reduced

dimensionalitycomparedtotheoriginalfunctionalmusclenetworks(21insteadof

630edges).Family-wiseerrorcontrolwasmaintainedusingBonferronicorrection

tocorrectformultiplecomparisons(4´21=84comparisons).

Dataavailabilitystatement

Thedatathatsupportthefindingsofthisstudyareavailablefromthe

correspondingauthoruponreasonablerequest.

Acknowledgment

WethankDanieleMarinazzo,SimonFarmer,SjoerdBruijn,HenkSchutte,and

NadiaDominicifortheirinsightfulcomments.Thisworkwassupportedbythe

NetherlandsOrganizationforScientificResearch(NWO45110-030and

016.156.346),theARCCentreofExcellenceinIntegrativeBrainFunction,andthe

AustralianNationalHealthandMedicalResearchCouncil(APP1037196,

APP1110975).

Authorcontributions

J.K.,A.D.andT.B.conceivedthestudy;allauthorsdesignedtheexperiments;J.K.

acquiredtheEMGdata;J.K.andT.B.performedtheevaluationoftheanatomical

network;J.K.,L.G.andT.B.performedtheevaluationofthefunctionalnetwork;T.B.

performedthestatisticalanalyses;J.K.andT.B.wrotethemanuscript;andallthe

authorseditedthemanuscript.


https://doi.org/10.1101/181818

21

References

1. BerkovitzB,HeimerL,HenkelC,MaT,McNealJ,MooreN.TerminologiaAnatomica:InternationalAnatomicalTerminology.NewYork:ThiemeStuttgart;1998.

2. SherringtonC.Theintegrativeactionofthenervoussystem:CUPArchive;1910.3. MartiniFH,TimmonsMJ,P.MM.HumanAnatomy(3rdEdition):PrenticeHallCollege

Div;1999.4. Pierrot-DeseillignyE,BurkeD.Thecircuitryofthehumanspinalcord:itsrolein

motorcontrolandmovementdisorders:CambridgeUniversityPress;2005.5. GrazianoM.Theorganizationofbehavioralrepertoireinmotorcortex.AnnuRev

Neurosci.2006;29:105-34.6. MiriA,WarrinerCL,SeelyJS,ElsayedGF,CunninghamJP,ChurchlandMM,etal.

Behaviorallyselectiveengagementofshort-latencyeffectorpathwaysbymotorcortex.Neuron.2017;95:683-96e11.

7. BernsteinN.Thecoordinationandregulationofmovements.London:Pergamon;1967.

8. TurveyMT.Coordination.AmPsychol.1990;45:938-53.9. KelsoJS.Dynamicpatterns:Theself-organizationofbrainandbehavior:MITpress;

1997.10. ChielHJ,BeerRD.Thebrainhasabody:adaptivebehavioremergesfrominteractions

ofnervoussystem,bodyandenvironment.TrendsNeurosci.1997;20(12):553-7.11. KuoAD,ZajacFE.Humanstandingposture:multi-jointmovementstrategiesbasedon

biomechanicalconstraints.ProgrBrainRes.1993;97:349-58.12. KutchJJ,Valero-CuevasFJ.Challengesandnewapproachestoprovingtheexistenceof

musclesynergiesofneuralorigin.PLoSComputBiol.2012;8:e1002434.13. FarmerSF,BremnerFD,HallidayDM,RosenbergJR,StephensJA.Thefrequency

contentofcommonsynapticinputstomotoneuronesstudiedduringvoluntaryisometriccontractioninman.JPhysiol.1993;470:127-55.

14. d'AvellaA,SaltielP,BizziE.Combinationsofmusclesynergiesintheconstructionofanaturalmotorbehavior.NatNeurosci.2003;6:300-8.

15. TingLH,MacphersonJM.Alimitedsetofmusclesynergiesforforcecontrolduringaposturaltask.JNeurophysiol.2005;93:609-13.

16. DominiciN,IvanenkoYP,CappelliniG,d’AvellaA,MondìV,CiccheseM,etal.Locomotorprimitivesinnewbornbabiesandtheirdevelopment.Science.2011;334(6058):997-9.

17. GhahramaniZ,WolpertDM.Modulardecompositioninvisuomotorlearning.Nature.1997;386(6623):392.

18. Mussa-IvaldiFA.Modularfeaturesofmotorcontrolandlearning.CurrOpinNeurobiol.1999;9(6):713-7.

19. GiszterSF.Motorprimitives—newdataandfuturequestions.CurrOpinNeurobiol.2015;33:156-65.

20. TakeiT,ConfaisJ,TomatsuS,OyaT,SekiK.Neuralbasisforhandmusclesynergiesintheprimatespinalcord.ProcNatAcadSci.2017;114(32):8643-8.

21. TreschMC,SaltielP,BizziE.Theconstructionofmovementbythespinalcord.NatNeurosci.1999;2(2):162–7.

22. HartCB,GiszterSF.Aneuralbasisformotorprimitivesinthespinalcord.JNeurosci.2010;30(4):1322-36.

23. LevineAJ,HinckleyCA,HildeKL,DriscollSP,PoonTH,MontgomeryJM,etal.Identificationofacellularnodeformotorcontrolpathways.NatNeurosci.2014;17(4):586.

24. NewmanME.Modularityandcommunitystructureinnetworks.ProcNatAcadSci.2006;103(23):8577-82.

25. FortunatoS.Communitydetectioningraphs.PhysRep.2010;486(3):75-174.


https://doi.org/10.1101/181818

22

26. RavaszE,SomeraAL,MongruDA,OltvaiZN,BarabásiA-L.Hierarchicalorganizationofmodularityinmetabolicnetworks.Science.2002;297(5586):1551-5.

27. BullmoreE,SpornsO.Complexbrainnetworks:graphtheoreticalanalysisofstructuralandfunctionalsystems.NatRevNeurosci.2009;10(3):186-98.

28. Esteve-AltavaB,DiogoR,SmithC,BoughnerJC,Rasskin-GutmanD.Anatomicalnetworksrevealthemusculoskeletalmodularityofthehumanhead.SciRep.2015;5:8298.

29. MurphyAC,MuldoonSF,BakerD,LastowkaA,BennettB,YangM,etal.Structure,Function,andControloftheMusculoskeletalNetwork.arXiv:161206336.2016.

30. BoonstraTW,Danna-Dos-SantosA,XieHB,RoerdinkM,StinsJF,BreakspearM.Musclenetworks:ConnectivityanalysisofEMGactivityduringposturalcontrol.SciRep.2015;5:17830.

31. GibbsJ,HarrisonLM,StephensJA.Organizationofinputstomotoneuronpoolsinman.JPhysiol.1995;485(1):245-56.

32. FarinaD,NegroF,JiangN.Identificationofcommonsynapticinputstomotorneuronsfromtherectifiedelectromyogram.JPhysiol.2013;591(10):2403-18.

33. BoonstraTW,BreakspearM.Neuralmechanismsofintermuscularcoherence:implicationsfortherectificationofsurfaceelectromyography.JNeurophysiol.2012;107(3):796-807.

34. BoonstraTW,FarmerSF,BreakspearM.Usingcomputationalneurosciencetodefinecommoninputtospinalmotorneurons.FrontHumNeurosci.2016;10.

35. DidierG,BrunC,BaudotA.Identifyingcommunitiesfrommultiplexbiologicalnetworks.PeerJ.2015;3:e1525.

36. MakarovSN,NoetscherGM,NazarianA.TriangularSurfaceHumanBodyMeshesforComputationalPurposes.Low-frequencyElectromagneticModelingforElectricalandBiologicalSystemsUsingMATLAB.NewYork:Wiley;2015.p.89-130.

37. JacobsonA.MatlabtoolboxforGeometryProcessing.GitHubRepository.2017:https://github.com/alecjacobson/gptoolbox.

38. HoneyCJ,KötterR,BreakspearM,SpornsO.Networkstructureofcerebralcortexshapesfunctionalconnectivityonmultipletimescales.ProcNatAcadSci.2007;104(24):10240-5.

39. RobertsJA,PerryA,LordAR,RobertsG,MitchellPB,SmithRE,etal.Thecontributionofgeometrytothehumanconnectome.NeuroImage.2016;124:379-93.

40. HorvátS,GămănuțR,Ercsey-RavaszM,MagrouL,GămănuțB,VanEssenDC,etal.Spatialembeddingandwiringcostconstrainthefunctionallayoutofthecorticalnetworkofrodentsandprimates.PLoSBiol.2016;14(7):e1002512.

41. TielandM,TrouwborstI,ClarkBC.Skeletalmuscleperformanceandageing.JCachexiaSarcopeniaMuscle.2017.

42. BriggmanK,KristanJrW.Multifunctionalpattern-generatingcircuits.AnnuRevNeurosci.2008;31:271-94.

43. PartridgeLD.Signal-handlingcharacteristicsofload-movingskeletalmuscle.AmJPhysiol.1966;210(5):1178-91.

44. MartinJH.Thecorticospinalsystem:fromdevelopmenttomotorcontrol.Neuroscientist.2005;11(2):161-73.

45. Ritterband-RosenbaumA,HerskindA,LiX,Willerslev-OlsenM,OlsenMD,FarmerSF,etal.AcriticalperiodofcorticomuscularandEMG-EMGcoherencedetectioninhealthyinfantsaged9-25weeks.JPhysiol.2017;595(8):2699-713.

46. GoodmanCS,ShatzCJ.Developmentalmechanismsthatgenerateprecisepatternsofneuronalconnectivity.Cell.1993;72:77-98.

47. BácskaiT,FuY,SengulG,RusznákZ,PaxinosG,WatsonC.Musculotopicorganizationofthemotorneuronssupplyingforelimbandshouldergirdlemusclesinthemouse.BrainStructFunct.2013;218(1):221-38.

48. McHanwellS,BiscoeT.Thelocalizationofmotoneuronssupplyingthehindlimbmusclesofthemouse.PhilTransRoyalSoci.1981;293(1069):477-508.


https://doi.org/10.1101/181818

23

49. LacquanitiF,IvanenkoYP,D'avellaA,ZelikK,ZagoM.Evolutionaryanddevelopmentalmodules.FrontCompNeurosci.2013;7:61.

50. DeVriesIEJ,DaffertshoferA,StegemanDF,BoonstraTW.Functionalconnectivityinneuromuscularsystemunderlyingbimanualcoordination.JNeurophysiol.2016;inpress.

51. McAuleyJ,FarmerS,RothwellJ,MarsdenC.Common3and10Hzoscillationsmodulatehumaneyeandfingermovementswhiletheysimultaneouslytrackavisualtarget.JPhysiol.1999;515(3):905-17.

52. WelshJP,LlinásR.Someorganizingprinciplesforthecontrolofmovementbasedonolivocerebellarphysiology.ProgrBrainRes.1997;114:449-61.

53. LemonRN.Descendingpathwaysinmotorcontrol.AnnuRevNeurosci.2008;31:195-218.

54. CarsonR.Neuralpathwaysmediatingbilateralinteractionsbetweentheupperlimbs.BrainResRev.2005;49(3):641-62.

55. BiswalB,ZerrinYetkinF,HaughtonVM,HydeJS.FunctionalconnectivityinthemotorcortexofrestinghumanbrainusingechoplanarMRI.MagneticResMed.1995;34(4):537-41.

56. ShenK,MišićB,CipolliniBN,BezginG,BuschkuehlM,HutchisonRM,etal.Stablelong-rangeinterhemisphericcoordinationissupportedbydirectanatomicalprojections.ProcNatAcadSci.2015;112(20):6473-8.

57. GallegoJA,PerichMG,MillerLE,SollaSA.Neuralmanifoldsforthecontrolofmovement.Neuron.2017;94(5):978-84.

58. ColeMW,BassettDS,PowerJD,BraverTS,PetersenSE.Intrinsicandtask-evokednetworkarchitecturesofthehumanbrain.Neuron.2014;83(1):238-51.

59. vandenHeuvelMP,SpornsO.Networkhubsinthehumanbrain.TrendsCogSci.2013;17(12):683-96.

60. McAuleyJH,MarsdenCD.Physiologicalandpathologicaltremorsandrhythmiccentralmotorcontrol.Brain.2000;123(8):1545-67.

61. FreundP,FristonK,ThompsonAJ,StephanKE,AshburnerJ,BachDR,etal.Embodiedneurology:anintegrativeframeworkforneurologicaldisorders.Brain.2016;139(6):1855–61.

62. IvanovPC,LiuKK,BartschRP.Focusontheemergingnewfieldsofnetworkphysiologyandnetworkmedicine.NewJPhys.2016;18(10):100201.

63. BashanA,BartschRP,KantelhardtJW,HavlinS,IvanovP.Networkphysiologyrevealsrelationsbetweennetworktopologyandphysiologicalfunction.NatCommun.2012;3:702.

64. WilligenburgNW,DaffertshoferA,KingmaI,vanDieenJH.RemovingECGcontaminationfromEMGrecordings:acomparisonofICA-basedandotherfilteringprocedures.JElectromyogrKinesiol.2012;22(3):485-93.

65. LeeDD,SeungHS.Learningthepartsofobjectsbynon-negativematrixfactorization.Nature.1999;401(6755):788-91.

66. GallosLK,MakseHA,SigmanM.Asmallworldofweaktiesprovidesoptimalglobalintegrationofself-similarmodulesinfunctionalbrainnetworks.ProcNatlAcadSciUSA.2012;109(8):2825-30.

67. BlondelVD,GuillaumeJ-L,LambiotteR,LefebvreE.Fastunfoldingofcommunitiesinlargenetworks.JStatMech.2008;2008(10):P10008.

68. LancichinettiA,FortunatoS.Consensusclusteringincomplexnetworks.SciRep.2012;2:336.

69. PapalexakisEE,AkogluL,IenceD,editors.Domoreviewsofagraphhelp?communitydetectionandclusteringinmulti-graphs.Informationfusion(FUSION),201316thinternationalconferenceon;2013:IEEE.

70. KujalaR,GlereanE,PanRK,JääskeläinenIP,SamsM,SaramäkiJ.Graphcoarse‐grainingrevealsdifferencesinthemodule‐levelstructureoffunctionalbrainnetworks.EurJNeurosci.2016;44(9):2673-84.


https://doi.org/10.1101/181818

Network structure of the human musculoskeletal system ... · 08.01.2018 · The human body is a complex system consisting of many subsystems and regulatory pathways. The musculoskeletal

Documents