Embed Size (px)
Citation preview
4 Alphaandbetamodulationduringweightshifting
InpatientswithParkinson’sdisease(PD),congruentaugmentedvisualfeedback(VF)canimprovefunctionalmotorperformance.Conversely,patientswithPDappearmoresusceptible to unreliable, i.e. incongruent, VF, than healthy subjects. This might beaccompaniedbychangesinactivityoftheperceptual-motorneuralnetwork.Infact,changesinbetaactivitymayrelatetotheuncertaintyinfeedforwardestimationbyaninternalmodel.Wehypothesizedthatinvalidatingthecurrentsensorimotormappingthrough incongruentVFwillbe reflectedby similar changes in corticalactivity.Wetestedthisbyanalyzingevent-relatedelectroencephalographicalphaandbetaactivityinaposturalweight-shiftingtask.Twenty-fourpatientsandfifteenhealthy,age-andgender-matched controls performed rhythmic swaying movements. Feedback ofsubjects’ center-of-pressure was presented either in real time (congruent VF) ordelayed by 250 or 500ms (incongruent VF), or was entirely absent.We estimatedsource activity at bilateral primary visual cortex (V1), primarymotor cortex (M1),primarysomatosensorycortex,andsuperiortemporalsulcus/gyrus,anddeterminedthecorrespondingmotor-relatedspectralpowerandpowermodulationinalphaandbeta frequency bands.No significantdifferences in cortical activity between the PDpatientsandhealthycontrolswerepresentwhencongruentVFwasprovided.UnderincongruentVF,however,thePDgroupshowedsignificantlyhigherbetamodulationinM1,andhigheralphamodulation inV1thanhealthycontrols.Event-relatedbetamodulationinthemotornetworkandalphamodulationinvisualareasdiscriminatedbetweengroups.ThissupportstheideathatPDisaccompaniedbyalteredmodulationsof alpha/beta activity during perceptual-motor tasks, andmay explain the greaterrelianceoncongruentVFinthispatientgroup.
Adaptedfrom:Van den Heuvel, van Wegen, Beek , Kwakkel., & Daffertshofer. Incongruent augmented visualfeedback during a postural task enhances cortical alpha and beta modulation in patients withParkinson’sdisease.Submitted.
Chapter4
52
Introduction
Balanceisnotexclusivelycontrolledatasubcortical levelbutrequiresnon-trivialcontributionsof thecerebral cortex (Jacobs&Horak,2007).Dual-taskparadigmsrevealedthatbalanceperformancedeteriorateswhenanadditionalcognitiveloadisadded (Maki & McIlroy, 2007; Woollacott & Shumway-Cook, 2002). In general,cognitive strategies can facilitatedifferentmotor tasks, including those related tobalance performance (Morris, Iansek, & Kirkwood, 2009). Accordingly, cognitivestrategies are included in the guidelines for physiotherapy in patients withParkinson’s disease (PD) (Bloem et al., 2010). Such strategies may range frommemorizing explicit step-by-step instructions on how to best initiate and/ormaintain a movement (Morris et al., 2009) to cueing, in which one tries tosynchronizeone’smovementswitha(rhythmic)externalstimulus(Lim,vanWegen,Goede,etal.,2005;vanWegenetal.,2014).Nexttotheuseofcognitivestrategies,motor function can also improve when movement-related feedback is provided(Huangetal.,2006).Augmentedvisual feedback(VF)helpsbothhealthysubjectsandpatientswithPDtobettercoordinatetheirmovementswithanexternaltarget(van den Heuvel, Daffertshofer, Beek, Kwakkel, & vanWegen, 2016). These andrelated technologies are hence increasingly adopted within therapeutic settings(Dockxetal.,2016).
Theimprovementsingaitandgait-relatedactivitiesinpatientswithPDduetocueingandVFraisethequestionwhetherneuralpathwaysand/orprocessesmightbe involved in facilitating these improvements (Jahanshahiet al., 1995).Over thepasttwodecadesithasbecomeclearthatmotorsymptomsinPDareassociatedwithabnormallyhighlevelsofneuralactivityinthebasalgangliainafrequencyrangeof13-30Hz,otherwiseknownasthebetaband(Brown,2007).Suchoscillatoryactivityinthecentralnervoussystemarisesfromthesynchronizedfiringpatternsoflocalpopulations of neurons, and is typically studied in termsof (changes in) spectralpower.Intriguingly,inpatientswithPD,exaggeratedbetapower(inbasalganglia)correlateswiththedegreeofmotorimpairment(Brown,2007;Jenkinson&Brown,2011;Neumannetal.,2016)andittendsbacktonormalwithdopaminereplacementtherapy (Hammond, Bergman, & Brown, 2007; Kühn et al., 2009; Kühn, Kupsch,Schneider,&Brown,2006;Rayetal.,2008;vanWijketal.,2016).Withrespecttothepositiveeffectsofcueingonmotorbehavior,itisofparticularinterestthatsalientcuespresentedpriortomovementareabletosuppressbetaactivity(Oswal,Litvak,
Alphaandbetamodulationduringweightshifting
53
Sauleau,&Brown,2012).Inthecortex,rhythmicstimuluspresentationinafinger-tappingtaskwasseentoproduceanincreaseinbetapower(teWoerd,Oostenveld,Bloem,deLange,&Praamstra,2015).Itremainsunclear,however,whethertheroleofbetapowerisfunctionalorwhethertheactivityismerelyanepiphenomenon;butthatbeingsaid,betapowerdoesprovideaninterestingwindowonthetask-relatedchangesinneuralactivityduringmotoractivity.
Atpresent,therearenostudiesthathaveaddressedtheinfluenceofVFonbetabandactivity.Thereare tworeasonswhysuchan influence isexpected: first, theimprovementsthatbothVFandcueingproduceatthebehaviorallevelarelikelytoarisefromthesameorsimilarunderlyingmechanisms.Second,ithasbeenshownthat, at the cortical level, movement error is negatively correlated to beta bandactivity(Tan, Jenkinson,&Brown,2014).VFprovidesthesubjectwithanexplicitmovementerrorandmightthereforeaffectthebetabandactivityinsimilarways.Importantly, in the current study, we focus on whole-bodymovement as cueingstrategieshavethusfarbeenimplementedin,andshowngreatestbenefitsfor,grossmotortasks.Specifically,subjectshereperformedarhythmicswayingtaskwithandwithoutVF.
WhileVFmaybeausefulavenue forpromotingmotorbehaviorand learning,thereisalsoevidencethatsubjectswithPDoverlyrelyonvisualinformation,(seeforinstance,Azulay,Mesure,Amblard,&Pouget,2002;Bronstein,Hood,Gresty,&Panagi, 1990; De Nunzio, Nardone, & Schieppati, 2007). This makes themparticularly vulnerable to incongruent visual information, in which the mappingbetween executed movement and the visual consequences of that movement iscorrupted.InapreviousstudywefoundthatpatientswithPDarelessabletoadapttosituationswithincongruentVFthanhealthycontrols(vandenHeuveletal.,2016).Betaactivitymayindeedhavearoleinassessingthereliabilityofsensoryfeedback.Tan,Wade, andBrown (2016) showed thatpost-movementbeta synchronizationover the sensorimotor cortex negatively correlated with the uncertainty infeedforwardestimations.Inthepresentstudy,wethereforealsocontrastedmotor-relatedactivationinthecortexbetweenconditionsofincongruentandcongruentVFcomparingagroupofpatientswithPDwithagroupofhealthycontrols.InlinewiththeaforementionedfindingswehypothesizedbetaactivitytoalterinthepresenceofcongruentVFdependentongroup(hypothesis1).Moreimportantly,weexpectedmarkedgroup-specificdifferencesincorticalactivityinthepresenceofincongruentVF(hypothesis2).
Chapter4
54
Methods
DesignWe performed a cross-sectional study of neurophysiological responses during aposturalswaytaskinagroupofpatientswithPDandagroupofage-andgender-matchedhealthycontrols.Datawerederivedfrombaselineassessmentsperformedinarandomizedclinicaltrial(RCT),registrationnumberISRCTN47046299(vandenHeuvel et al., 2013). The protocol had been approved by the Medical EthicsCommittee of VU UniversityMedical Centre (VUmc) Amsterdam. All participantssigned informed consent. The posturographic analyses have been reported inChapter3.(cf.vandenHeuveletal.,2016)
ParticipantsSubjectswithPDwererecruitedfromtheDepartmentofRehabilitationMedicineofVUmc.Inclusioncriteriawere(i)adiagnosisof idiopathicPDaccordingtotheUKBrainBankcriteria(Gibb&Lees,1988),mildtomoderatestage(i.e.Hoehn&YahrstagesIIandIII),(ii)abletoparticipateineitherofthetrainingprograms(cf.Chapter5andChapter6),and(iii)writtenandverbalinformedconsent.Exclusioncriteriawere:(i)presenceof(other)neurological,orthopedic,orcardiopulmonaryproblemsthat could impairparticipation, (ii)MiniMental StateExamination (MMSE) scorebelow24points,(iii)arecentchangeindopaminergicmedication,and(iv)cognitive,visual,and/or languageproblems impedingparticipation.Patientsunderwent theassessmentintheON-phaseoflevodopamedication,approximately1.5hoursafterintakeofthelastmedicationdosage.ControlsubjectswererecruitedfromthesocialenvironmentoftheparticipatingPDpatients.
EEG-acquisitionTheexperimentalsetupisshowninFig.3.1.Allsubjectsworea64-electrodeEEGheadcap(TMSi,Enschede,TheNetherlands),mountedinaccordancewiththe10-20standardelectrodeplacement.ElectrodegelwasappliedbetweenscalpandAg/AgClelectrodesinordertominimizeimpedance.EEGsignalswererecordedagainstthecommon average using a 64-channel amplifier (Refa, TMSI, Enschede, TheNetherlands)andsampledat2048Hz.
Alphaandbetamodulationduringweightshifting
55
PerformancemeasureandproceduresThe postural assessment consisted of a lateral weight-shifting task in whichparticipantsstooduprightandshiftedtheircenterofmassinarhythmicfashion(seeFig.3.1.).Thetaskwasperformedwhilestandingona forceplate(Kistler9281B,Ostfildern, Germany, 600×400 mm, sampled at 1 kHz) and facing a computermonitor (15²-LCD at eye-height about 80 cm away). Trials consisted of 100 s ofvoluntary rhythmic swaying in the frontal plane (i.e. from side to side) underdifferentfeedbackconditions.Beforeandaftereachtrialthesubjectstoodquietlyfor20s.Duringatrial,atargetcirclemovedfromsidetosideatafrequencyof0.5Hz.Thetaskwastotrackthemotionofthistargetbyswayingthebodyinthefrontalplane. Visual feedback consisted of the participant’s center-of-pressure (COP)motion,whichwasindicatedonthescreenbyaredcircle.COPfeedbackwaslimitedtomotionsalongthemediolateral(ML)axisandsmoothedusinganonlinelow-passfilter(25Hzcut-off).Feedbackwaspresentedeitherinrealtime,ordelayedby250or500ms,furtherreferredtoasVFrt,VF250,andVF500,respectively.Inthecontrolcondition, only the target signal was visible (i.e. feedbackwas absent, VFno) andparticipantswereaskedtomatchthemotionofthetargetbyswayingcomfortably.After familiarization, every condition was repeated three times, with repetitionspresentedinrandomizedblocks.
DataanalysisFulldetailswithrespecttotheposturographicdataanalysiscanbefoundinvandenHeuvel et al. (2016). In brief, along the ML axis the difference between thenormalized COP and target time series served to define a tracking error (Error).Trackingstabilitywasquantifiedasthecircularvariance(Varc)oftherelativeHilbertphase between COP and target motion (Mardia & Jupp, 2000). The normalizedamplitude(Anorm)wasdeterminedastheCOP’saveragepeakexcursionduringeachtrial.
PreprocessingEEGdataprocessingwasperformedinMatlab(TheMathworks,Natick,MA,versionR2016b) using the fieldtrip toolbox (www.fieldtriptoolbox.org, Oostenveld, Fries,Maris,&Schoffelen,2010).Datawerefirstresampledto1kHzforcompatibilitywiththe co-registered force plate. Data were notch filtered at k×50 Hz, (k = 1…5,
Chapter4
56
bandwidth±½Hz)toremovepowerlinehumandsubsequentlyband-passfilteredwith a second-order bi-directional Butterworth band-pass filter (1.5-250 Hz) toreducemovementartifactsandhigh-frequencynoise.Badchannelsweredetectedonthebasisoftoolargeortoosmallmeansorstandarddeviations,andifnecessary,interpolatedviathesurroundingchannelsusingsphericsplines.
Foreverytrial,dataweresubjectedtoanindependentcomponent(IC)analysis(fastICA;Hyvärinen(1999)).ICswereconsideredartifactsif(i)medianfrequency<1Hz(~movementartifact),(ii)medianfrequency>60Hz(~EMGactivity),(iii)iftopography was dominated by the prefrontal channels (~ EOG activity, i.e. eyemovements). After omitting these ICs in the mixing matrix, the sensor EEG wasreconstructedforfurtheranalysis.
SourcereconstructionWedetermined spatial filters in terms of linearly constrainedminimumvariance(LCMV) beamformers (Hillebrand & Barnes, 2003; Van Veen, van Drongelen,Yuchtman,&Suzuki,1997).AnMRItemplatefromthefieldtriptoolboxwasused,segmentedusingaboundaryelementmethod.The leadfieldwascomputedusingstandardconductancesofthedifferenttissues(gray/whitematter,skull,scalp).TheMRIwasparcellatedusingtheautomatedanatomicallabeling(AAL)atlascontaining90 regions-of-interest (ROIs). In linewith the twohypotheses listedbelowunderStatisticalanalysis,werestrictedtheanalysistobilateralM1,somatosensorycortex(S1),V1,andSTS/STG.IntheAALatlastheseregionscorrespondtotheprecentralgyrus, postcentral gyrus, the calcarine fissure, and superior temporal gyrus,respectively(Schmahmannetal.,1999;Tzourio-Mazoyeretal.,2002).
Afterbroad-bandfiltering(1-80Hz),spatialfilterswerecomputedforeachROIusingEEG-covarianceestimatesinanevent-relateddesign.Eventsweredefinedasthemomentsofmaximumrightswayexcursionandepochsweredefinedat±400msaroundevents.Inthecaseofquietstance,eventsareconstructedbyrandomlytakingnon-overlappingepochsatsetintervals.AfteraveragingtheobtainedbeamformerweightswithinaROI,EEGsignalswereprojectedtoobtainsourcetimeseries foreveryROIandfurtherassessedfortheirspectralcontents.Sincevisual inspectiondidnot suggest signs of lateralizationwe combinedpower values of homologousROIs.
Alphaandbetamodulationduringweightshifting
57
SpectralanalysisSpectrogramswereestimatedbymeansofashort-timeFouriertransformusinga1-sslidingHammingwindow(seeFigureS4.1forillustration).Tostabilizenormality,spectrogramswerelog-transformed.Themeanpowerofthequietstancecondition(baselinecondition)wassubtractedfornormalizationpersubject.Nexttothebetafrequencyband(15-30Hz)wealsoconsideredalphaactvity(8-14Hz);thegammafrequencyband(30-80Hz)wasanalyzedaswellbutpoorsignal-to-noiseratioandlargeinter-subjectvariabilityhamperedconsistencyintheresultsofthisanalysis;wenotethatweabstainedfromanalyzingthetabandactivity(4-8Hz)inviewoftheinfluenceofresidualmovementartifacts.Ineveryfrequencybandthespectrogramswereaveragedyieldingthefollowingalphaandbetapoweroutcomes:(i)themeanpowerovertimeand(ii)thestandarddeviationovertime,withthelatterservingasmeasureofpowermodulation.RecallthathomologousROIswerecombined.
StatisticalanalysisAll statistical assessments were realized using IBM SPSS Statistics 24. OutcomesweretestedfordeparturesfromnormalityusingtheShapiro-Wilktestandinspectedforoutliers.Differencesinspectralpowerandpowermodulationweretestedusingmultiplemixed-designANOVAs,oneforeachfrequencybandandtypeofoutcome(meanpower,powermodulation).ToexaminegroupdifferencesinEEGactivitywithandwithoutfeedback,weusedthebetween-subjectsfactorgroupandthewithin-subjects factor feedback, which only contained the conditions VFno and VFrt(hypothesis1,group´feedback).
Since differences in motor performance may confound the analysis of theinfluence of incongruent VF, data were normalized with respect to VFno. Weexamined group differences in EEG activity during conditions of delayed (i.e.incongruent) feedback (hypothesis 2, group ´ delay), using the between-subjectsfactor group and the within-subjects factor delay, containing the conditions VFrt,VF250,andVF500.Huynh-FeldtcorrectionsarereportedifMauchly’stestofsphericitywassignificant.ForeveryANOVA,groupdifferencesforoutcomeswithasignificantinteractioneffectwereanalyzedbycomputingsimplemaineffects,whereasgroupdifferencesforoutcomeswithoutasignificantinteractioneffectwereanalyzedusingthemain,between-subjects,effectsreportedbytheANOVA.AsequentialBonferroni-type procedure was applied to each ANOVA to control the false discovery rate,
Chapter4
58
yieldingdistinctsignificancecriteriaofvaryingmagnitude(Benjamini&Hochberg,1995).
Results
ParticipantsTable4.1 shows theparticipants’ characteristics.DataofonehealthyandonePDsubjectdeviatedfromthemeanbymorethanfivetimesthestandarddeviationandwerethereforeexcludedfromtheanalyses.Atotalof24patients(8F/16M,67.6±
Table4.1Subjectcharacteristics
PD Healthycontrols
center dispersion center dispersion
Sex[m/f] 16/8 8/7 Age[yrs] 67.63 (±8.74) 66.93 (±6.83)Fallers[y/n] 10/14 3/12 FRT[cm] 27.49 (±7.46) 31.72 (±5.42)BBS[0-56] 53.00 (50.00-55.00) 56.00 (55.00-56.00)singlelegstance[s] preferred 41.56 (8.25-60.00) 60.00 (27.45-60.00)non-preferred 17.73 (5.89-46.07) 38.35 (18.05-60.00)
10mwalktest walkspeed[m/s] 1.17 (±0.29) 1.34 (±0.22)steplength[m] 0.67 (±0.10) 0.70 (±0.10)
FES[0-30] 4.00 (1.75-8.25) 0.00 (0.00-0.50)HADS anxiety[0-21] 4.00 (2.00-8.00) 2.00 (0.00-3.00)depression[0-21] 3.50 (2.00-5.00) 1.00 (1.00-3.50)
MFI general[4-20] 12.00 (9.00-15.00) 5.00 (4.50-6.00)physical[4-20] 11.50 (9.00-15.00) 6.00 (5.00-7.00)
Diseaseduration[y] 8.58 (±6.92) -UPDRS 45.00 (31.75-53.75) -HY[1.5/2/2.5/3/3.5] 0/9/12/3/0/ -PDQ39 23.40 (14.10-35.90) -L-dopaequivalence[mg/day] 751.1 (±412.0)
Alphaandbetamodulationduringweightshifting
59
8.7 years) and 15 healthy age-matched controls (7F/8M, 66.9 ± 6.8 years) wereenteredinthefinalanalyses.
PosturographyHypothesis1.Significantgroup×VFinteractioneffectsforswayperformancewereabsent.PatientswithPDhadsignificanthigherError(F(1,37)=8.887,p=0.003)andhigher Var (F(1,37) = 8.897, p = 0.001), see Fig. 4.1, than healthy controls. Theamplitude Anorm did not differ significantly between groups (F(1,37) = 0.973,p=0.417).
Hypothesis2.Theinteractioneffectsofgroup×delayweresignificantatthecorrectedsignificance level α = 0.0333 for normalized Error and Var(F(1.577,61.250) = 13.115, p < 0.001, and F(1.769,68.988) = 22.051, p < 0.001,respectively),butnotfornormalizedAnorm(F(1.319,51.443)=2.783,p=0.142).Bothnormalized Error and Varwere significantly higher in healthy controls than inpatientswithPD:forErrorintheconditionsVF250(F(1,37)=8.149,p=0.003)andVF500(F(1,37)=31.940,p<0.001),andforVarinVF500(F(1,37)=26.001,p<0.001);seeFig.4.2.RecallthatforthistestthethreeoutcomeswerenormalizedtoVFnoandthatthenon-normalizederrorwassignificantlyhigherinPD(alsoseeDiscussion).Normalized Anorm did not differ significantly between groups (F(1,37) = 0.648,p=0.289).
EEG
Hypothesis 1. No significant interaction effects were present, and neither meanpowernorpowermodulationdisplayedasignificantmaineffectofgroupinanyofthefourROIs;seeFig.S4.2.
Hypothesis 2. None of the normalized outcome measures displayed a significantgroup × delay interaction effect (corrected significance level α = 0.003125).Normalized alpha modulation in S1 showed a pronounced trend towards asignificant interaction of group and delay (F(2,74) = 5.280,p= 0.007). Between-subjecteffectsweretestedagainstacorrectedsignificancelevelofα=0.00625.BothnormalizedalphamodulationinV1(F(1,37)=9.480,p=0.004andnormalizedbetamodulationinM1(F(1,37)=9.293,p=0.004)werefoundtobesignificantlyhigher
Chapter4
60
inPDthaninhealthycontrols(Fig.4.3,lowerpanels).Fornormalizedmeanpower,noneoftheeffectsweresignificantinanyoftheROIs(Fig.4.3,upperpanels).
Discussion
WeinvestigatedcorticalsynchronizationanddesynchronizationaccompanyingtheperformanceofaposturalswayingtaskinthepresenceofVF.Task-relatedspectralpowerandmodulationsthereofwereexaminedundervariousfeedbackconditions.WhencongruentVFwasprovided,nosignificantdifferencesinspectralpowerandpower modulation between the patient group and healthy controls were found.However, as hypothesized,we did find a differential response in cortical activityunderincongruentVFconditions.
Fig.4.1 Posturographic outcomes summarizing the effect of congruent VF (black: VFno,white: VFrt).
Errorbarsindicatethestandarddeviation.NotethatforVar,untransformeddataareshowntoimprovelegibility;statisticalanalysiswasperformedonFisher-transformeddata.
Fig.4.2 Posturographic outcomes showing the effect of the feedback conditionsVFrt (white), VF250(lightgrey),andVF500(darkgrey).DatahavebeennormalizedwithrespecttoVFnotocontrolfordifferencesinmotorperformancebetweenthegroups.
PD C0
0.5
1
1.5Anorm
PD C0
0.2
0.4
0.6
0.8 Error
VFnoVFrt
PD C0
0.1
0.2
0.3
0.4 Var
PD C0
200
400
600
norm
alize
d Erro
r [%
]
Error
PD C0
2000
4000
6000
8000
norm
alize
d Var
[%]
Var
PD C0
50
100
150
norm
alize
d Ano
rm [%
]
AnormVFrtVF250VF500
Alphaandbetamodulationduringweightshifting
61
Fig.4.3 Normalizedmeanpower(A)andnormalizedpowermodulation(B)forVFrt,VF250,andVF500
normalizedwithrespecttoVFno.Theoutcomeofeitherfeedbackconditionwasdividedbythat of VFno and converted to percentages. Results are shown for the alpha and betafrequencybandsforthePDgroupandthehealthy,age-matchedcontrols.Thefourregions-of-interestrelatetoprimarymotorcortex(M1),primarysensorycortex(S1),visualcortex(V1),andauditoryareas(STS/STG).Activitywascollapsedoverleftandrighthemispheres.Asterisks indicate significant effects of group after corrections. Error bars indicate thestandarddeviations.
PD C
50
100
150
norm
alize
dme
an al
pha p
ower
[%]
Precentral (M1)
PD C
50
100
150
Postcentral (S1)
PD C
50
100
150
200
Calcarine (V1)
PD C
50
100
150
Temporal (STS/STG)
PD C406080
100120140160
norm
alize
dme
an be
ta po
wer [
%]
PD C406080
100120140160
PD C
50
100
150
200
PD C
50
100
150
PD C
95
100
105
no
rma
lize
d
alp
ha
po
we
r m
od
ula
tion
[%
]
Precentral (M1)
PD C
95
100
105
Postcentral (S1)
PD C
95
100
105
Calcarine (V1)
*
PD C
95
100
105
Temporal (STS/STG)
PD C
95
100
105
no
rma
lize
d
be
ta p
ow
er
mo
du
latio
n [
%]
*
PD C
95
100
105
PD C
95
100
105
PD C
95
100
105
VFrtVF250VF500
PD C
95
100
105
norm
alize
dalp
ha po
wer m
odula
tion [
%]
Precentral (M1)
PD C
95
100
105Postcentral (S1)
PD C
95
100
105Calcarine (V1)
*
PD C
95
100
105Temporal (STS/STG)
PD C
95
100
105
norm
alize
dbe
ta po
wer m
odula
tion [
%]
*
PD C
95
100
105
PD C
95
100
105
PD C
95
100
105
B
A
Chapter4
62
ThefirsthypothesiswasbasedonthenotionthatchangesinmotorfunctioninPD are associated with changes in cortical activation (van Wijk, Beek, &Daffertshofer, 2012). In PD, the availability of external cues and/or task-relatedfeedback can improve motor function (Lim et al., 2010; Nieuwboer et al., 2007;Rubinsteinetal.,2002).InlinewithOswaletal.(2012)wehypothesizedbetapowertodecreaseundertheinfluenceofVF.PosturographicresultsdidrevealanimpairedtaskperformanceinpatientswithPD(Fig.4.2),butthisdisparitywasnotreflectedinanyoftheEEGmeasures,presumablybecauseclearpatternsofmovement-relateddesynchronization and synchronization did not emerge. Typically, right beforemovement onset the amplitude of beta oscillations decrease and they remainsuppresseduntil apost-movement rebound, that canexceed thebaseline activitylevel(Pfurtscheller&LopesdaSilva,1999;forareview,seevanWijketal.,2012).However, these findings mostly come from studies on upper extremities (fingermovements).Far fewer studieshave investigatedcortical activityduringposturaland gait-related tasks, (see for instance, Bruijn et al., 2015; Gwin et al., 2011).Nevertheless,takenatfacevalue,thepresentresultssuggestthatforthisrhythmicswayingtaskwithcongruentVFbetaactivitydidnotdifferforsubjectswithPDandhealthycontrols.Byextension,thismaysuggestthatperipheral,PD-relatedeffectslikeincreasedrigidityyieldedthediminishedtaskperformance.5
ThesecondhypothesiswasbasedontheassumptionthatpatientswithPDaremoredependentonVF(Azulayetal.,2002;Bronsteinetal.,1990;DeNunzioetal.,2007). If so, their taskperformancewouldsuffermore from incongruentVF thanthatofhealthycontrolsandshowalteredbetamodulationinM1.WhereascongruentVF may enable subjects with PD to establish some form of voluntary control tooverruledisturbedautomaticreflexes,suchastrategywouldnotworkinconditionswith incongruent VF. We therefore expected that both sway performance andcortical ativity would reflect the increased task complexity. The results revealedsignificantlygreaternormalizedbetamodulation in themotor cortex forpatientswithPD.WithincongruentVFthePDgroupshowedanincreaseinM1betaactivationrelativetoVFno(Fig.4.3).Ontheotherhand,healthycontrolsshowedadecreaseinthe levelofM1activation relative toVFno. It thusappears that in thepresenceof
5OneofthepremisesofthisstudywasthatVFhelpsimproveperformance.Althoughatfirstglanceitseemedthat taskperformancedidnot appear to improveunder conditionsofVFrt, this is not entirely true: detailedanalysis revealed learning effects that took place over the course of the assessment (van den Heuvel,Daffertshofer,Beek,Kwakkel,&vanWegen,2016).Itwouldthushavebeenratherinterestingtoinvestigatetheseeffects in the EEG data as well, but the number of events needed for a reliable analysis of spectral powerunfortunatelydidnotallowforsuchananalysis.
Alphaandbetamodulationduringweightshifting
63
incongruent VF, the modulation of beta activity was suppressed in the healthycontrols,butnotinthepatientswithPD.Thiswasexpected,asbetasynchronizationhasbeenfoundtoberelatedtomovementerror(Tanetal.,2014)andhereincreasedwithincongruentVF(seeFig.4.2).Inconditionsofunfamiliarorunreliablefeedback,(pre-)existingmotorprogramsbecomeinvalidated.Currentsensoryfeedbackneedstobeintegratedinordertoupdatetheinternalmodel(Shadmehretal.,2010).Post-movementbeta synchronizationhasbeen shown tonegatively correlatewith theuncertaintyinfeedforwardestimationsderivedfromtheinternalmodel(Tanetal.,2016). Beta powermodulationmight thus reflect the relative usability of the VFand/ordrivemotoradaptation.Theseresultsmightthusbeinterpretedasindicatingthat therelativeuncertainty in the feedforwardestimations inhealthycontrols ishigherinconditionswithVFthanwithoutVF.VFinstilledrelativelowerconfidenceinthemotorexecution, therebyreflectingtheneedforadaptivebehavior.ThePDgroup, in contrast, showedgreater confidence in the existing internalmodel, andthusarelativelylowerneedforupdatingthefeedforwardcontrol.Thisfitswiththebehavioralobservations:forhealthycontrolsVarishigherforVFrtthanforVFnoandincreaseswithVF250andVF500;seeFigs.4.1&4.2,respectively,indicatingmuchmorevariable tracking patterns that are reflective of active adaptive behavior. For PDpatients,thiseffectonVarwasnotnearlythatcompelling,suggestingthattheymuchlesstriedtoadapttheirmotorbehavior.Itmayreflectadifficultyto‘uncouple’themovementfromthetargetmotion.Itmayalsoreflectageneraldisabilitytogenerateaproperresponse,aswasexploredbyotherstudiesonmotorperseverationinPD(see for instance, Stoffers, Berendse, Deijen, &Wolters, 2001;Mohammadi et al.,2015).
ThePDgroupshowedgreaternormalizedalphamodulationinV1;seeFig.4.3.Again, this suggests thatwith incongruent VF, themodulation of V1-activitywasreduced in healthy controls, but not in the patients with PD. Modulation ofsynchronizedalphaactivityinthisregionhasbeenascribedtoprocessingofvisualinformationaswellascognitiveandattentionalmechanisms(Pfurtscheller,Neuper,&Mohl,1994).
LimitationsNexttothealreadyacknowledgedproblemofmovementartifactsinEEGtherearealso more design-related issues that should be noted. The alignment of neural
Chapter4
64
activitytookplacewithrespecttothemaximumamplitudeoftheCOPofsuccessiveswayingmovements. It is debatable towhat extent this parameter – andnot, forexample, thepointofmaximalCOPvelocity– is thebest choice towhich toaligncortical activity. It should also be kept in mind that the task studied here is acontinuoustask;afterall,VFwasprovidedcontinuously,givingtheparticipantanongoingmeasureof the error. If betapower is indeeda functionof the extent towhichamotorplanrequiresupdating (Brittain&Brown,2014;Tanetal.,2014),neural dsynchronization/synchronization could thus have been taken placeunsystematically,atvariouspoints in theswaycircle,ultimatelygettingobscuredthrough the averaging over events. Another potential drawback of the presentdesignisthefactthatlearningtookplacethroughouttheexperiment(seevandenHeuvel et al., 2016). If performance was not (yet) at a steady state level for allsubjects,itisalsounlikelytobethecaseforEEGactivationpatterns.
Conclusion
The goal of the present study was to investigate the effects of VF on corticalactivation during a postural sway task. When congruent VF was provided, nosignificant differences in synchronized oscillatory activity were present betweenpatients with PD and healthy controls. In contrast, when incongruent VF wasprovided, increased event-related alpha and beta modulation across the motornetworkwere found in thePDgroup.This supports previous findings on alteredmovement-relatedmodulations of alpha/beta activity in patientswith PD, and isconsistentwithdatashowinggreaterrelianceoncongruentVFinPDpatients.Wecan confirm the notion of betamodulation as a cortical controller of (rhythmic)motor performance (Brittain & Brown, 2014; Houweling, Beek, & Daffertshofer,2010;vanWijketal.,2012).
Alphaandbetamodulationduringweightshifting
65
Supplementarymaterial
Fig.S4.1M
eanspectrogramsforhealthysubjectsforthefourregions(incolumns)andfourconditions(inrows).Spectralpow
erisshow
n±400msaroundtheevent(i.e.thetimeofmaximalrightCOPdiscursiontotheright-handside).Theanalysiswasrestrictedto
alpha(8-14Hz)andbetabands(15-30Hz).
Prec
entr
al
none-2
000
200
10203040 Frequency [Hz]
Calc
arin
e
-200
020
010203040
Post
cent
ral
-200
020
010203040
Tem
pora
l
-200
020
010203040
realtime
-200
020
010203040 Frequency [Hz]
-200
020
010203040
-200
020
010203040
-200
020
010203040
250 ms delayed
-200
020
010203040 Frequency [Hz]
-200
020
010203040
-200
020
010203040
-200
020
010203040
500 ms delayed
-200
020
0Ti
me [m
s]
10203040 Frequency [Hz]
-200
020
0Ti
me [m
s]
10203040
-200
020
0Ti
me [m
s]
10203040
-200
020
0Ti
me [m
s]
102030400.7
4
0.76
0.78
0.80.82
0.84
0.86
0.88
0.9
Normalized power
Chapter4
66
Fig.S4.2 ResultsforcongruentVF(VFnoandVFrt).A:Logmeanpower.B:logmeanpowermodulation.Shownarethealphaandbetafrequencybands,forthePDgroupandthehealthycontrols.The four regions-of-interest relate to M1, S1, V1, and STS/STG; activity of which wascollapsedoverleftandrighthemispheres.Errorbarsindicatestandarddeviations.
PD C
-0.05
-0.1
-0.15
-0.2
0
log m
ean a
lpha p
ower
Precentral (M1)
PD C-0.15
-0.1
-0.05
0 Postcentral (S1)
PD C
-0.05
-0.1
-0.15
-0.2
-0.25
0 Calcarine (V1)
PD C
-0.05
-0.1
-0.15
-0.2
0 Temporal (STS/STG)
PD C
-0.05
-0.1
-0.15
-0.2
0
log m
ean b
eta po
wer
PD C-0.15
-0.1
-0.05
0
PD C
-0.05
-0.1
-0.15
-0.2
0
PD C
-0.05
-0.1
-0.15
-0.2
0
PD C0
0.01
0.02
0.03
0.04
0.05
alp
ha p
ow
er
modula
tion
Precentral (M1)
PD C0
0.01
0.02
0.03
0.04
0.05Postcentral (S1)
PD C0
0.01
0.02
0.03
0.04
0.05Calcarine (V1)
PD C0
0.01
0.02
0.03
0.04Temporal (STS/STG)
PD C0
0.005
0.01
0.015
0.02
0.025
0.03
beta
pow
er
modula
tion
PD C0
0.005
0.01
0.015
0.02
0.025
0.03
PD C0
0.005
0.01
0.015
0.02
0.025
0.03
PD C0
0.005
0.01
0.015
0.02
0.025
0.03VFno
VFrt
PD C0
0.01
0.02
0.03
0.04
0.05
alpha
powe
r mod
ulatio
n
Precentral (M1)
PD C0
0.01
0.02
0.03
0.04
0.05 Postcentral (S1)
PD C0
0.01
0.02
0.03
0.04
0.05 Calcarine (V1)
PD C0
0.01
0.02
0.03
0.04 Temporal (STS/STG)
PD C0
0.005
0.01
0.015
0.02
0.025
0.03
beta
powe
r mod
ulatio
n
PD C0
0.005
0.01
0.015
0.02
0.025
0.03
PD C0
0.005
0.01
0.015
0.02
0.025
0.03
PD C0
0.005
0.01
0.015
0.02
0.025
0.03
B
A
