THERAPEUTIC
STRATEGIES
DRUG DISCOVERY
TODAY
Resting
state
EEG
rhythms
as
network
disease
markers
for
drug
discovery
in
Alzheimer’s
disease
Claudio
Babiloni
1
,
2
,
*
,
Francesco
Infarinato
2
,
Antonio
I.
Triggiani
3
,
Roberta
Lizio
2
,
Claudio
Del
Percio
2
,
Nicola
Marzano
4
,
Jill
C.
Richardson
5
1
DepartmentofPhysiologyandPharmacology,UniversityofRome‘LaSapienza’,Rome,Italy
2
DepartmentofNeuroscience,IRCCSSANRAFFAELEPISANA,Rome,Italy
3
DepartmentofClinicalandExperimentalMedicine,UniversityofFoggia,Foggia,Italy
4
IRCCS‘SDN’,Naples,Italy
5
GlaxoSmithKlineR&DChinaGroup,GunnelsWoodRoad,Stevenage,Herts,UK
Alzheimer’sdisease(AD)inducesawidespread patho-logical extracellular accumulation of beta-amyloid (Ab)peptidesthataffectscorticalnetworks underpin-ning cognitive functions. This is related to abnormal functionalandeffectivebrainconnectivityas revealed bygraphmarkersofresting-stateeyes-closed electro-encephalographic (EEG) rhythms. Here we revised EEG studies in mild cognitive impairment and AD subjects showing that these markers are promising networkdiseaseendpointsforbasicresearch and AD drugdiscovery.
Sectioneditors:
JillRichardson– NeurosciencesTherapeuticArea, GlaxoSmithKlineR&D,Stevenage, UK
JackieHunter– Biotechnologyand BiologicalSciences ResearchCouncil, PolarisHouse,Swindon,UK
Introduction
Alzheimer’sdisease(AD)isthemostfrequent
neurodegener-ativedisorderandcauseofdementiaalongaging.Itis
char-acterized by a pathological accumulation of beta-amyloid
(Ab)andhyperphosphorylatedtaupeptidesthataffect
corti-calneuronalnetworksrelatedtocognitivefunctions[1].For
thisreason,anetworkdiseaseperspectivehasthepotentialto
provide novel markersof thepathology at preclinical and
prodromal AD stages [2], as well as instrumental targets
for drug discovery. In this line, resting-state eyes-closed
electroencephalographic(EEG)rhythmsprovideuseful
infor-mationonfunctionalandeffectivebrainconnectivityacross
long-rangeneuralnetworks[3,4].
The hypothesis of this review is that global topological
features of functional and effective brain connectivity as
revealed by the functional coupling of resting-state
eyes-closed EEG rhythms could build a platform of promising
network diseaseendpoints for basic researchand AD drug
discovery.
FunctionalandeffectivebrainconnectivityinMCIand ADsubjectsasrevealedbylinearandnon-linear couplingoftherestingstateEEGrhythms
Synchronousactivityofoscillatingnetworksarisesfromthe
interactionbetweentheintrinsicexcitabilityofneuronsand
theirinterconnectivity[5] andcorrelateswiththesubjects’
behavior andwith thestimulating conditions ofthe
envi-ronment[6].Inmoredetail,evensingleneuronsthemselves
can oscillate at multiple frequencies [7], a fortiori cortical
regionsdonotexhibitpure oscillationsbutacombination
Editors-in-Chief
RaymondBaker–formerlyUniversityofSouthampton,UKandMerckSharp&Dohme,UK EliotOhlstein–AltheRxPharmaceuticals,USA
Recent
advances
in
the
treatment
of
Alzheimers
ofdelta(<4Hz),theta (4–7Hz),alpha(8–12Hz),beta (13–
30),andgamma (>30Hz)rhythms,coexisting inthesame
areaorinteractingamongdifferentstructures,andgenerally
attributedtonetworkoperationsincortico-corticoand
cor-ticothalamic systems [8]. To this theoretical model, joint
EEG/functionalmagneticresonanceimaging(fMRI)analysis
showedthatlarge-scalefunctional-anatomicalnetworks–as
revealedbyauto-correlationofbloodoxygenation
level-de-pendent(BOLD)activity–oscillateinmultiple
electrophysi-ological frequency bands [9]. Neurodegenerative disorders
impairthesynapticintegrityunderlyingcooperationamong
neural networks; several studies demonstrated the role of
soluble Ab as a principal causative agent of synaptic and
cognitiveimpairmentinAD.Robustevidencedemonstrates
thatsolubleAbelicits a toxic signalingcascade by the
a7-nicotinicacetylcholinereceptor(a7nAChR),leadingto
syn-apticimpairments,intraneuronalAb42aggregates,and
cor-related cognitive deficits[10]. This compromisedsignaling
activateskinasesERK2andJNK1,leadingtotheformationof
neurofibrillary tangles (NFTs) namely aggregates of
hyper-phosphorylatedtauprotein.Cognitiveimpairmentand
mag-nitude ofsynapticdeficitin theADbrainare morehighly
correlated with soluble Ab than with the abundance of
amyloidplaques,reflectingthefactthatsolubleAbinflicts
synapticimpairment[11].
InthefieldoffMRIresearch,thereisawholelineofstudy
thatfocusesondifferentialbrainconnectivityinADpatients
whencomparedtohealthyelders.Inthisline,BOLDhasbeen
reportedto beapromising markerfortheidentificationof
disrupted functional connectivity patterns in AD patients
[12,13]. Moreover, it has been shownthat in ADpatients,
acetylcholinesterase inhibitor donepezil affects functional
connectivity of dorsolateral prefrontal cortex [14] and of
parahippocampal,temporal,parietalandprefrontalcortices
associatedtomedialcholinergicpathwaynetwork[15].
Fur-thermore,strongerrecoveryinthenetworkconnectivitywas
correlatedwithcognitiveimprovementasmeasuredby
Mini-MentalStateExamination(MMSE)andADAssessment
Scale-cognitivesubscalescores(ADASCog)[15].
In the framework of EEG techniques, functional brain
connectivity is reflected by statistical functional coupling
of EEG oscillatory activity in various frequency bands [3].
Moreover, effective brain connectivity is reflectedby how
EEGrhythmsrecordedatoneelectrodeaffectsthoserecorded
at another remote electrode, due to a causal hierarchical
interactionbetweenthetwocorrespondingcortical
genera-tors[3].Inthisframework,thefunctionalbrainconnectivity
is typically indexed by EEG spectral coherence (linear) or
synchronizationlikelihood(linear–nonlinear),whilethe
ef-fectivebrainconnectivityisindexedbymarkersderivedby
informationtheoryandgrangercausality[16].Theseindexes
docapturelinearandnonlinearrelationships amongbrain
regions[17].Thelinearindexesmodelthephaserelationship
betweentheEEGrhythmsrecordedatelectrodepairs,
where-asthenonlinearindexesdisclosecomplexrelationships
be-tweentheseEEGrhythms[17].
EEGspectralcoherence quantifiesthetemporal
synchro-nizationoftwoEEGtimeseriesbetweenpairsofelectrodesin
thefrequency domain,andcanbetypicallyderivedbyfast
Fouriertransform–FFT–[18].Atgrouplevel,spectral
coher-ence ofthe restingstate eyes-closed EEG rhythms differed
amongtheNold,MCI,andADsubjects[19–22].Themajority
ofpreviousEEGstudiesreportedaprominentdecreaseofthe
spectralcoherenceatalpharhythmsintheADthaninthe
Noldsubjects[19,20].Thiseffectwasfoundtobeassociated
with ApoE genetic risk of AD [19]. Other studies showed
decrease or increase of EEG coherence at delta and theta
rhythms in ADpatients [21]. A recent study averagedEEG
spectral coherence across all electrode pairs showing the
pathological increase of the ‘total coherence’ in the AD
patients in relation to cholinergic lesions [22]. Although
severalstudieshaveshownthatEEG coherencemightbea
diagnosticbiomarkersensitivetodiseaseseverity,itispoorly
knowntherelationshipbetweensuchEEGmarkersand
spe-cific abnormalities induced by Ab, partially recovered by
activecompoundsagainstAD(i.e.donepezil).Onlyonestudy
describednegligible differencesbetweenthepre-and
post-treatmentwithdonepezil5mgdailyfor1monthon
inter-hemispheric EEG coherence across homologous electrode
pairs[23].
‘Synchronization likelihood’ (SL) is an index capturing
both linearand non-linear dimensions of EEG functional
coupling. It measures the dynamical interdependencies
between EEG signals recorded at electrode pairs [24]. SL
relies on the concept of ‘state’ of one dynamical system
expressed intermsofthelevel ofneuralsynchronization,
asindexedbyserialmeasurementsoftheEEGvoltage[24].
Compared to Nold subjects, MCI and AD patients were
characterized by lower global SLacross all electrodepairs
athigh-frequencyalphaandbetabands[25].Furthermore,
therewasafronto-parietalreductionofSL(deltatoalpha)
greaterinMCIthaninADpatients[26].Moreover,patients
with vascular dementia and mild AD showed poor SL at
bothfronto-parietal(deltatoalpha)and
inter-hemispheri-cal (delta to beta) electrode pairs [27]. And the
fronto-parietal reduction of SL at alpha rhythms was greater in
mildADthaninVaDpatients[27].SLwasalsosensitiveto
anotherformofdementiaduetoneurodegeneration.
Com-paredtoParkinson’sdisease(PD)patients,PDpatientswith
dementia(PDD)werecharacterizedbylowervaluesofSLin
thefollowingelectrodepairsandfrequencybands:
fronto-temporalalpharhythms;inter-hemisphericaltemporal
del-ta, theta and alpha rhythms; and centro-parietal gamma
rhythms [28].Incontrast, parieto-occipitalSL at
high-fre-quencyalphaandbetabandswashigherin thePDDthan
OthernonlinearindexesoftheEEGfunctionalcoupling
derive frominformation theory, which is based upon the
conceptofentropydefinedastheuncertaintyassociatedwith
arandomvariable.PreviousEEGstudiesusingmutual
infor-mation oftheinformation theory havedisclosed a loss of
functionalconnectivityinADpatientsindifferentfrequency
bandsoftheresting-state eyes-closedEEG rhythms, witha
special engagement ofthe alphafrequencies [29].
Further-more,thelocalcross-mutual information (CMI) quantified
the information transmitted from one the EEG rhythms
recordedat oneelectrode over another; CMIwas lowerin
theADpatientsthanthatinnormalcontrols,especiallyover
the EEG rhythms recorded in frontal andantero-temporal
regions[30].Furthermore,therewasaprominentdecreasein
informationtransmissionbetweendistantEEGelectrodesin
therighthemisphereandbetweencorresponding
inter-hemi-sphericelectrodes [30]. Inaddition, theauto-mutual
infor-mation(AMI)estimated howmuch onaveragethevoltage
value oftheEEGrhythms canbepredicted fromvalues of
those at preceding points. It decreased significantly more
slowlywithdelaythroughoutthescalpinADthaninNold
subjects[30].
Fig. 1(top)illustratestheprincipal modelsoffunctional
brain connectivity as indexed by the linear or nonlinear
functional coupling of the resting-state eyes-closed EEG
rhythms. The linear models reflect thephase relations
be-tween the EEG rhythms recorded at electrode pairs. The
nonlinearmodelsdenoteacomplexrelationbetweenthese
EEGrhythms.
BothlinearandnonlinearindexesoftheEEGfunctional
brain connectivity have an important limitation: they do
reflectneitherthecausalaspectsoftherelationshipsamong
brain regionsnor the directionof theinformation among
these regions. One can overcome this limitation by the
estimation of the directional information flow with the
EEGcoupling. Fruitful approachesrelyontheinformation
theoryandGrangercausality[31].
Concerning theinformationtheory, transferentropy
in-dexes directed (time-asymmetric) information transfer
be-tween joint processes (i.e. the EEG rhythms recorded at
twoelectrodes) [32]. It isrobusteven withunknown
non-linear interactions [33]. On the other hand, the Granger
causalityrefersto thenotionthat,ifthepredictionofone
timeseriescould beimproved byincorporatingthe
knowl-edgeofpastvaluesofasecondone,thenthelatterissaidto
haveacausal influenceontheformer [34].A verypopular
procedurederivedfromtheGrangercausalityisthesocalled
directedtransferfunction(DTF),whichhasbeenproventobe
reliable for the modeling of directional information flux
within linear EEG functional coupling on the basis of a
multivariate autoregressive model (DTF) [31]. Concerning
itsclinicalapplications,ithasbeenreportedareductionof
the parietal-to-frontal directional information flow within
theEEGfunctionalcoupling inamnesic MCIandmild AD
subjectscomparedtoNoldsubjects[35].Thisfindingsuggests
acommonpathophysiologicalbackgroundlinking,on
aver-age, thegroups ofMCI andAD subjects. Noteworthy, the
fronto-parietalfunctionalcouplingisrelativelypreservedin
theamnesicMCIsubjectsinwhomthecognitivedeclineis
mainly explained by the extent of white-matter vascular
disease [36]. This finding supports the additive model of
thecognitiveimpairmentposingthatMCIstatusresultsfrom
Phase relation Complex relation time (s) time (s) [µV] [µV] Frontal Frontal Parietal Linear
EEG signals functional connectivity Funtional connectivity Non linear Parietal 3 0 3 0
Drug Discovery Today: Therapeutic Strategies
Figure1. Sketchofmainmodelsoffunctionalbrainconnectivityindexedbylinearandnonlinearfunctionalcouplingofresting-stateeyes-closed electroencephalographic(EEG)rhythms.
thecombinationofcerebrovascular andneurodegenerative
lesions [36]. In addition, other EEG studies used Granger
causality andstochastic eventsynchrony as models ofthe
directionalinformationflux[37,38].Resultsshowedalossof
EEG synchrony between electrode pairs in MCI and AD
patients with respect to age-matched control subjects
[37,38]. Furthermore, these markers provided a successful
leave-one-outindividualclassificationrateof83%and88%.
Fig.2plotsanexamplemodelofthedirectional
informa-tionflowbetweenelectrodepairsfromtheresting-state
eyes-closedEEGrhythms.
Itshouldbehoweverremarkedthatcomputational
mea-suresoffunctionalconnectivityarenotfreefrom
disadvan-tages. They do not directly estimate metastability and
typicallyuselinearmathematicalmodelsofthebrainsignals,
evenifbrainconnectivitymodelcannotberestricttoonly
linear phenomena[39]. Furthermore, interpretation ofthe
resultsmightbebiasedbytheuseofperiodsofnon-stationary
EEGsignals[40].
ThenetworkdiseaseinMCIandADsubjectsas revealedbygraph theory
HowtointegratethementionedEEGmarkersoffunctional
andeffectivebrainconnectivity(i.e.spectralcoherence,SL,
DTF,etc.)intoatopologicalmodelofthenetworkdisease?A
promisingtheoreticalframeworkisofferedbygraphtheory
[41]. In this theory, the graphs are simplified
representa-tionsofnetworksdenotedbyensemblesofnodes(vertices)
andconnections(edges).UsingEEGandMEGmarkersasan
input, graph theory studies showed that healthy subjects
werecharacterizedbyanefficientandrobustnetworkcalled
‘small-world’,withhighclusteringamong nearnodesand
relatively few ‘hubs’ connecting far nodes [41,42]. In AD
brains,‘small-world’networkpropertieswerereplacedbya
more ‘random’ overall network structure[41–43].Indeed,
AD patients were characterized by the mean clustering
coefficient decreased at the lower-frequency (EEG) alpha
andbetabands,andbythecharacteristicpath length(i.e.
globalconnectivity)decreasedatthelower-frequencyalpha
andgammabands[43].AparallelMEGstudyshowedthatin
theADpatientsthispathologicalchangewasbroughtabout
by a preferential decrease of connections between high
degreenodes (‘hubs’),ratherthan a non-specificdecrease
ofconnectionstrength[41].InanotherMEGstudy,network
analysiswasusedtoinvestigatetheroleoffunctional
sub-networks (modules) in the brainwith regard to cognitive
failureinAD[44].Itwasshownthattheparietalcortexwas
the most highlyconnected network area in both control
subjects andAD patients, but it was characterized by the
strongest intra-modular clustering losses in AD patients.
Furthermore, weakening of inter-modular and
long-dis-tance connectivity was even more outspoken, and more
strongly related to cognitive impairment [44,45]. These
results support the idea that the loss of communication
and relative less efficient information exchange among
differentfunctionalbrainregionsreflectsanabnormal
syn-aptic plasticity, neural loss, and cognitive decline in AD
[43,44].Noteworthy,lossofsmall-worldstructureinADwas
also demonstrated in recent MRI studies applying graph
theory[46,47].
Fig.3plotsamodeloftheaboveresultsofgraphtheory.
Normalcontrolsubjectsarecharacterizedbya‘small-world’
networkstructureofthefunctionalcouplingofthe
resting-stateEEGrhythms(left).Withrespecttothenormalcontrol
subjects, theAD patients manifest thedeviation of
‘small-world’networkpropertiestowardsamore‘random’ overall
networkstructure(right).
[µV]
0 3
Causal relation
0
time (s) 3
Directed Transfer Function (DTF) Effective connectivity
Parietal Frontal
Drug Discovery Today: Therapeutic Strategies
Figure2. Estimationofthedirectionalinformationflowbetween electrodepairsfortheresting-stateeyes-closedEEGrhythms.
Normal subject
Cluster Path AD patient ‘Small-world’ network
Drug Discovery Today: Therapeutic Strategies
Figure3. Sketchofthenodesandtopologicconnectionsof functionalbrainconnectivityinhealthysubjectsandinAlzheimer’s disease(AD)patientsasrevealedbyelectroencephalographic(EEG) ormagnetoencephalographic(MEG)rhythms.Thetopologyofthese nodesandconnectionsshowthedeviationoftheADpatients(i.e. reductionofthelocalclusteringandlong-rangeconnectivity)fromthe ‘small-world’networkpropertiesobservedinthecontrolsubjects.
Conclusions
Inconclusion,thedeepcomprehensionofnormalEEG
dy-namicsof functional andeffectiveconnectivity in healthy
withrespectofnon-healthysubjectsandthe
pharmacologi-calwayofrestorationofanormalpatternisausefulapproach
forbasic,applied,anddrugdiscoveryresearch.EEGmarkers
offunctionalandeffectiveconnectivitycanbeusedasinputs
forananalysistoproducecandidatenetworkdiseasemarkers
tobeusedasdependentvariablesforbasicresearch,aswellas
surrogateendpointsfordrugdiscoveryofnewsymptomatic
and disease-modifying compounds against the AD. Future
studies should demonstrate how EEG functional coupling
couldallowidentifying noveltargetsofAD
neurodegenera-tion;howitreflectssynapticplasticityandneuronalnetwork
activity;andhowitwillimpactnetworkbasedstrategiesfor
drug discovery. These issues will have to be addressed by
simultaneousrecordingofEEGandspikepotentialsat
differ-entscalesinthebrainofrodentmodelsofAD(e.g.TASTPM),
togetherwithdosingAbin thebrainandpharmacological
manipulationswithAbloweringdrugs.
Conflictofinterest
Theauthorshavenoconflictofinteresttodeclare.
Acknowledgments
WeareindebtedwithProf.PaoloM.RossiniandDr.Fabrizio
Vecchiofortheirinvaluablecriticalcontributiontothe
dis-cussion of thetopic of this review. Theresearch was
sup-portedinpartbytheSanRaffaeleS.p.A.Theactivityofsome
co-Authorsleading tothepresentreview wasdevelopedin
the framework of the PRIN2010-2011 project ‘CONNAGE’
andoftheGRIDCOREprojectofItalianMinistryofHealth.
References
1. PievaniM,deHaanW,WuT,SeeleyWW,FrisoniGB.Functionalnetwork
disruptioninthedegenerativedementias.LancetNeurol2011;10:829–43.
2. PalopJJ,MuckeL.Amyloid-beta-inducedneuronaldysfunctionin
Alzheimer’sdisease:fromsynapsestowardneuralnetworks.NatNeurosci
2010;13:812–8.
3.BabiloniC,FrisoniGB,PievaniM,VecchioF,LizioR,ButtiglioneM,etal.
HippocampalvolumeandcorticalsourcesofEEGalpharhythmsinmild
cognitiveimpairmentandAlzheimerdisease.Neuroimage2009;44:123–35.
4. D’AmelioM,RossiniPM.Brainexcitabilityandconnectivityofneuronal
assembliesinAlzheimer’sdisease:fromanimalmodelstohumanfindings.
ProgNeurobiol2012;99:42–60.
5. Buzsa´kiD,DraguhnA.Neuronaloscillationsincorticalnetworks.Science
2004;304:1926–9.
6. HasselmoME,Bodelo´nC,WybleBP.Aproposedfunctionforhippocampal
thetarhythm:separatephasesofencodingandretrievalenhancereversal
ofpriorlearning.NeuralComput2002;14:793–817.
7. Llina´sRR.Theintrinsicelectrophysiologicalpropertiesofmammalian
neurons:insightsintocentralnervoussystemfunction.Science
1988;242:1654–64.
8. SteriadeM.Groupingofbrainrhythmsincorticothalamicsystems.
Neuroscience2006;137:1087–106.
9. MantiniD,PerrucciMG,DelGrattaC,RomaniGL,CorbettaM.
Electrophysiologicalsignaturesofrestingstatenetworksinthehuman
brain.ProcNatlAcadSciUSA2007;104:13170–75.
10.DziewczapolskiG,GlogowskiCM,MasliahE,HeinemannSF.Deletionof
thealpha7nicotinicacetylcholinereceptorgeneimprovescognitive
deficitsandsynapticpathologyinamousemodelofAlzheimer’sdisease.J
Neurosci2009;29:8805–15.
11.Na¨slundJ,HaroutunianV,MohsR,DavisKL,DaviesP,GreengardP,etal.
Correlationbetweenelevatedlevelsofamyloidbeta-peptideinthebrain
andcognitivedecline.JAMA2000;283:1571–7.
12.BaiF,ZhangZ,WatsonDR,YuH,ShiY,YuanY,etal.Abnormalfunctional
connectivityofhippocampusduringepisodicmemoryretrievalprocessing
networkinamnesticmildcognitiveimpairment.BiolPsychiatry
2009;65:951–8.
13.HahnK,MyersN,PrigarinS,RodenackerK,KurzA,Fo¨rstlH,etal.
Selectivelyandprogressivelydisruptedstructuralconnectivityof
functionalbrainnetworksinAlzheimer’sdisease–revealedbyanovel
frameworktoanalyzeedgedistributionsofnetworksdetectingdisruptions
withstrongstatisticalevidence.Neuroimage2013;1:96–109.
14.ZaidelL,AllenG,CullumCM,BriggsRW,HynanLS,WeinerMF,etal.
Donepezileffectsonhippocampalandprefrontalfunctionalconnectivity
inAlzheimer’sdisease:preliminaryreport.JAlzheimersDis2012;31(Suppl
3):S221–6.
15.LiX,LiTQ,AndreasenN,WibergMK,WestmanE,WahlundLO.Ratioof
Ab42/P-tau181pinCSFisassociatedwithaberrantdefaultmodenetwork
inAD.SciRep2013;3:1339.
16.RossiniPM,RossiS,BabiloniC,PolichJ.Clinicalneurophysiologyofaging
brain:fromnormalagingtoneurodegeneration.ProgNeurobiol
2007;83:375–400.
17.StamCJ,HillebrandA,WangH,VanMieghemP.Emergenceofmodular
structureinalarge-scalebrainnetworkwithinteractionsbetween
dynamicsandconnectivity.FrontComputNeurosci2010;4:133.
18.PfurtschellerG,AndrewC.Event-relatedchangesofbandpowerand
coherence:methodologyandinterpretation.JClinNeurophysiol
1999;16:512–9.
19.JelicV,JulinP,ShigetaM,NordbergA,LannfeltL,WinbladB,etal.
ApolipoproteinEe4alleledecreasesfunctionalconnectivityinAlzheimer’s
diseaseasmeasuredbyEEGcoherence.JNeurolNeurosurgPsychiatry
1997;63:59–65.
20.JelicV,JohanssonSE,AlmkvistO,ShigetaM,JulinP,NordbergA,etal.
Quantitativeelectroencephalographyinmildcognitiveimpairment:
longitudinalchangesandpossiblepredictionofAlzheimer’sdisease.
NeurobiolAging2000;21:533–40.
21.BrunovskyM,MatousekM,EdmanA,CervenaK,KrajcaV.Objective
assessmentofthedegreeofdementiabymeansofEEG.
Neuropsychobiology2003;48:19–26.
22.BabiloniC,FrisoniGB,VecchioF,PievaniM,GeroldiC,DeCarliC,etal.
GlobalfunctionalcouplingofrestingEEGrhythmsisrelatedto
white-matterlesionsalongthecholinergictractsinsubjectswithamnesicmild
cognitiveimpairment.JAlzheimersDis2010;19:859–71.
23.ReevesRR,StruveFA,PatrickG.Theeffectsofdonepezilonquantitative
EEGinpatientswithAlzheimer’sdisease.ClinElectroencephalogr
2002;33:93–6.
24.StamCJ,vanDijkBW.Synchronizationlikelihood:anun-biasedmeasure
ofgeneralizedsynchronizationinmultivariatedatasets.PhysicaD
2002;163:236–51.
25.PijnenburgYA,vdMadeY,vanCappellenvanWalsumAM,KnolDL,
ScheltensP,etal.EEGsynchronizationlikelihoodinmildcognitive
impairmentandAlzheimer’sdiseaseduringaworkingmemorytask.Clin
Neurophysiol2004;115:1332–9.
26.BabiloniC,FerriR,BinettiG,CassarinoA,FornoGD,ErcolaniM,etal.
Fronto-parietalcouplingofbrainrhythmsinmildcognitiveimpairment:a
multicentricEEGstudy.BrainResBull2006;69:63–73.
27.BabiloniC,FerriR,MorettiDV,StrambiA,BinettiG,DalFornoG,etal.
Abnormalfronto-parietalcouplingofbrainrhythmsinmildAlzheimer’s
disease:amulticentricEEGstudy.EurJNeurosci2004;19:2583–90.
28.BosboomJL,StoffersD,WoltersECh,StamCJ,BerendseHW.MEGresting
statefunctionalconnectivityinParkinson’sdiseaserelateddementia.J
NeuralTransm2009;116:193–202.
29.JeongJ.EEGdynamicsinpatientswithAlzheimer’sdisease.Clin
30.JeongJ,ChaeJH,KimSY,HanSH.NonlineardynamicanalysisoftheEEG
inpatientswithAlzheimer’sdiseaseandvasculardementia.JClin
Neurophysiol2001;18:58–67.
31.KaminskiMJ,BlinowskaKJ.Anewmethodofthedescriptionofthe
informationflowinthebrainstructures.BiolCyber1991;65:203–10.
32.SchreiberT.Measuringinformationtransfer.PhysRevLett2000;85:
461–4.
33.VicenteR,WibralM,LindnerM,PipaG.Transferentropy–amodel-free
measureofeffectiveconnectivityfortheneurosciences.JComput
Neurosci2011;30:45–67.
34.GrangerCWJ.Investigatingcausalrelationsbyeconometricmodelsand
cross-spectralmethods.Econometrica1969;37:424–38.
35.BabiloniC,FrisoniG,VecchioF,LizioR,PievaniM,GeroldiC,etal.Global
functionalcouplingofrestingEEGrhythmsisabnormalinmildcognitive
impairmentandAlzheimer’sDisease.AmulticenterEEGstudy.J
Psychophysiol2009;23:224–34.
36.BabiloniC,FrisoniGB,PievaniM,VecchioF,InfarinatoF,GeroldiC,etal.
Whitemattervascularlesionsarerelatedtoparietal-to-frontalcouplingof
EEGrhythmsinmildcognitiveimpairment.HumBrainMapp
2008;29:1355–67.
37.DauwelsJ,VialatteF,LatchoumaneC,JeongJ,CichockiA.EEGsynchrony
analysisforearlydiagnosisofAlzheimer’sdisease:astudywithseveral
synchronymeasuresandEEGdatasets.ConfProcIEEEEngMedBiolSoc
2009;2009:2224–7.
38.DauwelsJ,VialatteF,MushaT,CichockiA.Acomparativestudyof
synchronymeasuresfortheearlydiagnosisofAlzheimer’sdiseasebasedon
EEG.Neuroimage2010;49:668–93.
39.LandaP,GribkovD,KaplanA.Oscillatoryprocessesinbiologicalsystems.
In:MalikSK,ChandrashekaranMK,PradhanN,editors.Nonlinear
PhenomenainBiologicalandPhysicalSciences.IndianNatl.Sci.Acad.;
2000.p.123–52.
40.FingelkurtsAA,FingelkurtsAA,Ka¨hko¨nenS.Newperspectivesin
pharmaco-electroencephalography.ProgNeuropsychopharmacolBiol
Psychiatry2005;29:193–9.
41.StamCJ,deHaanW,DaffertshoferA,JonesBF,ManshandenI,van
CappellenvanWalsumAM,etal.Graphtheoreticalanalysisof
magnetoencephalographicfunctionalconnectivityinAlzheimer’sdisease.
Brain2009;132:213–24.
42.StamCJ,JonesBF,NolteG,BreakspearM,ScheltensPh.Small-world
networksandfunctionalconnectivityinAlzheimer’sdisease.CerebCortex
2007;17:92–9.
43.deHaanW,PijnenburgYA,StrijersRL,vanderMadeY,vanderFlierWM,
ScheltensP,etal.Functionalneuralnetworkanalysisinfrontotemporal
dementiaandAlzheimer’sdiseaseusingEEGandgraphtheory.BMC
Neurosci2009;10:101.
44.deHaanW,vanderFlierWM,KoeneT,SmitsLL,ScheltensP,StamCJ.
Disruptedmodularbraindynamicsreflectcognitivedysfunctionin
Alzheimer’sdisease.Neuroimage2012;59:3085–93.
45.Sanz-ArigitaEJ,SchoonheimMM,DamoiseauxJS,RomboutsSA,MarisE,
BarkhofF,etal.Lossof‘small-world’networksinAlzheimer’sdisease:
graphanalysisofFMRIresting-statefunctionalconnectivity.PLoSOne
2010;5:e13788.
46.HeY,ChenZ,EvansA.Structuralinsightsintoaberranttopological
patternsoflarge-scalecorticalnetworksinAlzheimer’sdisease.JNeurosci
2008;28:4756–66.
47.SupekarK,MenonV,RubinD,MusenM,GreiciusMD.Networkanalysisof
intrinsicfunctionalbrainconnectivityinAlzheimer’sdisease.PLoS
