C.Babilonietal./NeuroImage44(2009)123–135127
LORETAbelongstothefamilyoflinearinversealgorithmslikeminimumnormsolution,weightedminimumnormsolutionorweightedresolutionoptimization(Pascual-Marquietal.,1999;Phillipsetal.,2002;YaoandHe,2001),andhasbeensuccessfullyusedinrecentEEGstudiesonpathologicalbrainagingusingthesameexperimentalsetup(electrodemontage,samplefrequency,etc.)ofthepresentstudy(Dierksetal.,2000;Babilonietal.,2004,2006a,2006b,2006c,2006d,2006e).
LORETAcomputes3Dlinearsolutions(LORETAsolutions)fortheEEGinverseproblemwithina3-shellsphericalheadmodelincludingscalp,skull,andbraincompartments.Thebraincompartmentisrestrictedtothecorticalgraymatter/hippocampusofaheadmodelco-registeredtotheTalairachprobabilitybrainatlasanddigitizedattheBrainImagingCenteroftheMontrealNeurologicalInstitute(TalairachandTournoux,1988).Thiscompartmentincludes2394voxels(7mmresolution),eachvoxelcontaininganequivalentcurrentdipole.Ofnote,EEGelectrodepositionswerenotco-registeredtoindividualbrainsourcemodels;unfortunately,theof cialLORETApackagedidnotincludesoftwaretodosoandwecouldnotobtainthedigitalizationoftheelectrodepositionfromourclinicalunits.
LORETAsolutionsconsistedofvoxelz-currentdensityvaluesabletopredictEEGspectralpowerdensityatscalpelectrodes,beingareference-freemethodofEEGanalysis,inthatoneobtainsthesameLORETAsourcedistributionforEEGdatareferencedtoanyreferenceelectrodeincludingcommonaverage.AnormalizationofthedatawasobtainedbynormalizingtheLORETAcurrentdensityateachvoxelwiththepowerdensityaveragedacrossallfrequencies(0.5–45Hz)andacrossall2394voxelsofthebrainvolume.Afterthenormal-ization,thesolutionslosttheoriginalphysicaldimensionandwererepresentedbyanarbitraryunitscale.Thisprocedurereducedinter-subjectsvariabilityandwasusedinpreviousEEGstudies(Babilonietal.,2004,2006a,2006b,2006c,2006d,2006e).Thegeneralprocedure ttedtheLORETAsolutionsinaGaussiandistributionandreducedinter-subjectvariability(Leuchteretal.,1993;Nuwer,1988).OthermethodsofnormalizationusingtheprincipalcomponentanalysisareeffectiveforestimatingthesubjectiveglobalfactorscaleoftheEEGdata(Hernándezetal.,1994).ThesemethodsarenotavailableintheLORETApackage,sotheywerenotusedinthisstudy.
SolutionsoftheEEGinverseproblemareunder-determinedandillconditionedwhenthenumberofspatialsamples(electrodes)islowerthanthatoftheunknownsamples(currentdensityateachvoxel).Inordertoproperlyaddressthisproblem,thecorticalLORETAsolutionspredictingscalpEEGspectralpowerdensitywereregularizedtoestimatedistributedratherthanpunctualEEGsourcepatterns(Pascual-MarquiandMichel,1994;Pascual-Marquietal.,1999;Pascual-Marquietal.,2002).Inlinewiththelowspatialresolutionoftheadoptedtechnique,weusedourMATLABsoftwaretocollapsethevoxelsofLORETAsolutionsatfrontal,central,parietal,occipital,temporalandlimbicregionsofthebrainmodelcodedintoTalairachspace.TheBrodmannareaslistedinTable2formedeachoftheseregionsofinterest(ROIs).
Finally,themainadvantageoftheregionalanalysisofLORETAsolutions–usinganexplicitsourcemodelco-registeredintoTalairachspace–wasthatourmodelingcoulddisentanglerhythmsofcontiguouscorticalareas(namelythosefromtheoccipital

source
Fig.1.Grandaverageoflowresolutionbrainelectromagnetictomography(LORETA)solutions(i.e.normalizedrelativecurrentdensityatthecorticalvoxels)modelingthedistributedelectroencephalographic(EEG)corticalsourcesfordelta,theta,alpha1,alpha2,beta1,beta2bandsinnormalelderly(Nold),mildcognitiveimpairment(MCI)andAlzheimer'sdisease(AD).Theleftsideofthemaps(topview)correspondstothelefthemisphere.Legend:LORETA,lowresolutionbrainelectromagnetictomography.Colorscale:allpowerdensityestimateswerescaledbasedontheaveragedmaximumvalue(i.e.alpha1powervalueofoccipitalregioninNold).Themaximalvalueofpowerdensityisreportedundereachcolumn.
single hippocampal generator or that cortical theta...sources of the observed changes in EEG theta and...(1999). Frontal midline theta rhythms re?ect ...
single hippocampal generator or that cortical theta...sources of the observed changes in EEG theta and...(1999). Frontal midline theta rhythms re?ect ...
热门文档