2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 1 Development of Complex Curricula for Molecular Bionics and Infobionics Programs within a consortial* framework** Consortium leader PETER PAZMANY CATHOLIC UNIVERSITY Consortium members SEMMELWEIS UNIVERSITY, DIALOG CAMPUS PUBLISHER The Project has been realised with the support of the European Union and has been co-financed by the European Social Fund *** **Molekuláris bionika és Infobionika Szakok tananyagának komplex fejlesztése konzorciumi keretben ***A projekt az Európai Unió támogatásával, az Európai Szociális Alap társfinanszírozásával valósul meg. PETER PAZMANY CATHOLIC UNIVERSITY SEMMELWEIS UNIVERSITY sote_logo.jpg dk_fejlec.gif INFOBLOKK 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 2 Peter PazmanyCatholicUniversity Facultyof InformationTechnology ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND MUSCULAR SYSTEM Methodsforanalysingthebioelectricsignals www.itk.ppke.hu (Azideg-ésizom-rendszerelektrofiziológiaivizsgálómódszerei) (A bioelektromos jelek elemzõ módszerei) RICHÁRD FIÁTH and GYÖRGY KARMOS ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 3 www.itk.ppke.hu Contents • Aims of the lecture • Types of bioelectric signals • Data acquisition –from analog to the digital domain • Noise • Filters • Basic signal processing methods • Analysis of single-unit activity(Spike sorting) and multi-unit activity • Advanced signal processing methods • Neurometrics ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMELECTROMYOGRAPHY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 4 AIMS: Inthislecturethestudentwillbecomefamiliarwiththemethodsusedfortheanalysisofbioelectricsignals.Firstofallthebasicsofdataacquisitionwillbeintroduced,namelyhowtheanalogelectricsignalsareamplifiedandconvertedintoadigitalformtobecamestorableoncomputersanddigestablefortheprocessingalgorithms.Thisfollowsashortdescriptionofoneofthegreatestenemiesofsignalanalysis:noise.Filteringthesignalscanhelpustogetridofsomenoiseandartefactsfromtherecordings,andisusedalsotoretrievethesignalofinterestwithouttheundesiredfrequencycomponents.Basicmethodsforsignalprocessingwillbereviewedinthenextsection,likeaveraging,spectralanalysisorcorrelation.Afterthatthereadergetacquaintedwiththeprocedureofspikesortingandothermethodsinsingle-andmulti-unitanalysis.Intheadvancedmethodssectionwecangetaninsightintoasmallcollectionoftechniquesfrequentlyusedinneuroscience.LastlyaquantitaveEEGmethodwillbedemonstrated,whichmakesitpossibletodetectandquantifyabnormalbrainstates:Neurometrics. www.itk.ppke.hu ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 5 www.itk.ppke.hu TYPES OF BIOELECTRIC SIGNALS • Electrocardiogram (ECG) • Electromyogram (EMG) • Electroencephalogram(EEG) • Electroretinogram (ERG) • Local field potentials (LFP) • Multi-unit activity (MUA) • Single-unit activity (SUA) • Patch clamp recordings from ion channels • Optical recordings with voltage sensitive dyes BIOMAGNETIC SIGNAL • Magnetoencephalogram (MEG) ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 6 www.itk.ppke.hu DATA ACQUISITION Signalprocessinginneurosciencemeansapplyingvariousalgorithmstobioelectricsignalsthatcanbeone-dimensionaltimeseries(e.g.ECG),multi-dimensionaltimeseries(e.g.EEG,LFP)orseriesofimages(e.g.opticalrecordings).Thesetypeofsignalsareanaloginnature:theyarecontinuousbothintimeandamplitude.Bioelectricsignalsarepickedupusuallybyelectrodesandamplificated.Amplificationhastwosteps:firstthesignalisamplifiedbyanpreamplifierthatisfollowedbythemainamplifier.Afteramplificationthesignalisfilteredtoremoveundesiredfrequencycomponents.Thiscanbeaband-passfilteroranotchfiltertocutoutthenoiseofthepowerlines.Theanti-aliasingfilterisusedtoattenuatefrequenciesthataretoohightobedigitizedbytheanalog-to-digitalconverter(ADC).Thelaststepisconvertingthesignaltothedigitaldomain:asample-and-holdcircuitsamplestheanalogsignalandtheADCperformesthequantization.(Figure1.) ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 7 Figure 1. The process of data acquisition from bioelectric sources. data acquisition.png ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 8 www.itk.ppke.hu DATA ACQUISITION –A/D conversion Manytaskscanbeperformedwithanalogprocessing(e.g.filtering,rectifying),butdigitaltechnologyhastherealprocessingcapabilitywithitsrichvarietyofprocessingtechniques.Fordigitalprocessingofthesignal,ithastobeconvertedintoadiscreterepresentation.Discretizationoftimemeanssamplingthecontinuouswaveatagiveninterval.Therateatwhichthedigitalvaluesaresamplediscalledthesamplingrateorsamplingfrequency.Theoriginalsignalcanbeexactlyreproducedifthesamplingrateisatleasttwicethanthehighestfrequencyofthesignal(Nyquistsamplingtheorem).Theamplitudescaleismadediscretebyananalog-to-digitalconverter(ADC):thisprocessiscalledquantizationandisperformedbyroundingortruncatingameasuredreal-valuetoanintegerrepresentation.ThemostimportantpropertyofanADCistheamplituderesolutionwhichtellsusthenumberofdiscretevalues(levels)thattheADCcanproduceovertherangeofanalogvalues.ForexampleanADCwithan8-bitrangehas2^8=256levels.AnotherattributeoftheADCistheinputrangewhichismeasuredinVolts.(Figure2.) ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 9 www.itk.ppke.hu DATA ACQUISITION –A/D conversion –example Forexampleassumethatwehavea16-bitA/Dconverterwithaninputrangeof10V.Thepreamplifierintheheadstageamplifiestheanalogsignal10xwhichisfurtheramplifiedbythemainamplifierby1000.Thisresultsinatotalamplificationof10000x,sotherangefortheinputoftheacquisitionsystemis10V/10000=1mV.Theconverterhas2^16=65536levelswhichgivesaresolutionattheinputof1mV/65536=15nV.Thissystemhasaveryhighprecision,butusesalotofmemoryandstoragecapacity.ToconstructaproperADCsystemwehavetomakeatrade-offbetweentheresolution,rangeandstoragecapacitydependingontheneedsofourapplication. A/Dconversionhasseveralerrors:quantizationerror(orquantizationnoise),apertureerrorandnon-linearity. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 10 www.itk.ppke.hu A-D conversion.png Figure2.Schematicofanalog-to-digitalconversionwith3bitresolution–Aftertheamplificationoftheanalogsignalasample-and-holdcircuitsamplesitindefinedintervals.Afterthateverysamplewillbequantized:theamplitudevalueofoneoftheA/Dlevels(incaseof3bitresolutionthereare8levels)whichisclosesttotherealamplitudeofthesignalatthistimepointwillbeassignedtothesample.Finallytheresultsarestoredonastoragedevice. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 11 www.itk.ppke.hu NOISE Noiseisanundesiredperturbationsuperimposedonthesignalofinterest.Itcanoriginatefromseveralsources:wecandistinguishnaturalandhuman-madenoises.Naturalnoisesarerandomandcanbebiologicalorelectric(ormagnetic).Biologicalnoiseisforexampletheactivityrecordedfromneuronsfarfromtheextracellularelectrode,thesocalledbackgroundactivity,whichcaninterferewiththesignaloftheneuron,whichwewouldliketorecordandanalyse,andmaskitsactivity.(Figure3.) Artifactsareunwantedalterationsintherecordingsandcanberegardedassomenoise-likephenomen.Theyoriginatefromsourcesotherthantheelectrophysiologicalstructurebeingstudied.Itcanbeelectriclikethehumofthepowerlines,mechanical(e.g.cabelmovement)orbiologicalsuchastheelectricalactivityofthemuscleoreyesuperposedontheEEGrecording.Artifactscanbeassignedtohuman-madenoisesandmostofthemcanbeavoidedwithcarefulexperimentalsettings.(Figure4.,Figure5.) ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 12 www.itk.ppke.hu Figure3.A1–Thestrongbackgroundactivityofneuronsmaskthesmallamplitudesingleunitactivitesandmakescorrectspikesortingimpossible.A1isrepresentedinagreatertimescaleonA2. B1–Astrongsingleunitactivityoriginatingfromaneuronclosetotherecordingelectrodecaneasilysortedbecauseitsamplitudeismuchhigherthantheamplitudeofthebackgroundactivity.B2isB1displayedonagreatertimescale. background_noise.png ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 13 www.itk.ppke.hu artifacts.png Figure 4. Examples of artifacts related to EMG. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 14 www.itk.ppke.hu Figure5.EOG(electro-oculogram)artifactpresentonallofthechannelsofa2secondsegmentofanEEGrecording.TheamplitudeoftheartifactisthehighestattheEOGelectrodeofcourse,wherewerecordtheactivityoftheeyedirectly. Untitled.png ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 15 www.itk.ppke.hu NOISE Ontheotherhandnaturalelectricnoisesarealwayspresentandoriginatemainlyfromtheelectricequipmentoftherecordingchain(e.g.electrodes,amplifiers,A/Dconverters).Fortunatelyinmoderndevicesthelevelofthenoiseisminimal.Noiseisusuallycharacterizedbyitsprobabilitydensityfunction(PDF)orbyitspowerspectraldensity(PSD).NoisescanbeclassifiedbytheirPSD,differenttypeofnoisesarenamedafterdifferentcolors:forexamplethereexistwhite,pink,brownetc.noise.Thenoisecanbemeasuredalsoasanelectricpowerinwattsorasvoltageinvolts. There are well-known examples for random electric noises: • Johnson–Nyquistnoiseorthermalnoise:isgeneratedbytherandomthermalmotionofchargecarriers.Thepowerspectraldensityofthermalnoiseisnearlyequalthroughoutthefrequencyspectrumsoitiscalledawhitenoise.ThePDFoftheamplitudeisnearlyGaussian. • Anotherexamplesforelectricnoises:shotnoise,flickernoise(1/fnoise),burstnoise,avalanchenoise. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 16 www.itk.ppke.hu NOISE Quantizationnoise–Thediscretizationerror(duetotruncationorrounding)madeduringtheanalog-to-digitalconversionofthebioelectricsignal.ItsamplitudedependsontherangeandtheresolutionoftheA/Dconverter,butusuallyitisafewµVorless. TherelationoftheamplitudeofsomenoisesourcestotheamplitudeofbioelectricsignalsrecordedwithdifferentmethodsarepresentedonFigure6.Wecansee,thatfortunatelythemagnitudeofmostofthenoisesissosmall,thatitdoesnotaffectthesignalofinterest.Othertypesofnoisecanbepreventedwithpropertools,likeforexampletheuseofaFaradaycagetoshieldthebioelectricrecordingsfromelectromagneticradiaton.TheFaradaycageattenuatesorcompletlyblockstheelectromagneticwavesoutsideit.Figure7.showsanoisylocalfieldpotentialrecordedwithoutapplyingaFaradaycage. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 17 www.itk.ppke.hu noise.png Figure6.Theamplitudeofdifferentnoisesrelatedtotheamplitudeof bioelectricsignals. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 18 www.itk.ppke.hu Figure7.Eightchannelsofalocalfieldpotentialrecordingcontaminatedwith50Hznoiseoriginatingfromthepowerlinesandwithelectrodemovementartifacts.SlowDCshiftsarealsopresent. noisy_signal.png ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 19 www.itk.ppke.hu NOISE –SIGNAL-TO-NOISE RATIO Signal-to-noiseratio(SNRorS/N)isasignalqualitymeasureandshowshowmuchasignalhasbeencorruptedbynoise.Itisdefinedastheratioofsignalpower(meaningfulinformation)tothenoisepower(unwantedsignal)corruptingthesignal: wherePistheaveragepower.UndercertainconditionstheSNRcanbecalculatedasthesquareoftheamplituderatio.SNRisoftenexpressedusingthelogarithmicdecibelscale: SNR.png SNR2.png ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 20 www.itk.ppke.hu FILTERS Filteringinthefieldofbioelectricsignalsmeansremovingofcertainfrequencycomponentsfromthesignal,usuallynoise.Theattenuatedfrequenciesdependontheactualtask:forthesuppressingofthe50or60Hzofthepowerlinesanotchfilterisusedaroundthetargetfrequenciesorifwewanttoextractthesingleunitactivityfromawidebandrecordingwecanuseaband-passfilterfrom500Hzto5000Hz,thiswillattenuatethefrequencycomponentsbelow500Hzandabove5000Hz. Filterscanbestudiedeitherinthetimedomainorinthefrequenydomain.Usuallytheoperationofafilterisdescribedinthefrequencydomain:itremovestheunwantedfrequencycomponents(stopband)whileleavingtheotherintact(passband)withatransitionregion(theborderbetweenthestopandpassbands)ofzerowidth.Thiswouldbetheidealfilter(Figure8.).Inrealworldfiltersthegainsinthepassandstopbandarenotconstant,ripplesmaybepresentandthewidthofthetransitionareaisgreaterthanzero.(Figure9.) ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 21 www.itk.ppke.hu ideal_filter.png sample_filter.png Figure8.Frequencyresponsefunctionofanidealband-passfilter.(ontheright) Figure9.Simplifiedfrequencyresponsefunctionofareallow-passfilter.(ontheleft) ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 22 www.itk.ppke.hu FILTERS Filters can be classified upon many different bases: • analog (continuous-time) or digital (discrete-time ) • linear or nonlinear • Analog filters can be: active or passive • Types of digital filters: infinite impulse response (IIR) or finite impulse response (FIR) The best-known filter families: • Chebysev filter • Butterworth filter • Bessel filter • Elliptic filter Thesemodernlinearfiltersaredesignedinthewayofnetworksynthesismethodology.Allofthemcanbeimplementedonanalogandalsoondigitalfilters.Thedifferencebetweenthemisthattheyuseadifferentpolynomialfunctiontoapproximatetotheidealfilterresponse.(Figure10.) ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 23 www.itk.ppke.hu Figure10.Frequencyresponsefunctionsofthefourbasiclinearfiltersandtheidealfilter(dashedline).Eachfilterhasitsownadvantage. filter_families.png ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 24 www.itk.ppke.hu FILTERS Thefrequencyresponsecanbeclassifiedintoanumberofdifferentbandformsdescribingwhichfrequenciesthefilterpassesandwhichitattenuates: • Low-pass filter–The low frequencies are passed, the high frequencies are attenuated. • High-pass filter–The high frequencies are passed, the low frequencies are attenuated. • Band-pass filter–Onlyfrequencies in definedfrequency band are passed. • Band-stop filter or band-reject filter–Onlyfrequencies in adefinedfrequency band are attenuated • Notch filter–Rejects just one specific frequency: it isan extreme band-stop filter. Thecutofffrequencyisthefrequencybeyondwhichthefilterwillnotpasssignals. ThesimplifiedfrequencyresponsefunctionofthementionedfilterclassesaredisplayedonFigure11. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 25 www.itk.ppke.hu Figure11.Simplifiedfrequencyresponsefunctionsofdifferentfilterclasses. filters.png ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 26 www.itk.ppke.hu BASIC SIGNAL PROCESSING METHODS • Properties of bioelectric signals • Classifying the methods • Signal averaging • Frequency (spectral) analysis –Fourier Transformations • Covariance • Correlation • Coherence • Laplace-transform • Z-transform ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 27 www.itk.ppke.hu PROPERTIES OF BIOELECTRIC SIGNALS Amplitude–Meansthevaluesofthetime-varyingsignalalongthevertical-axis.IncaseofbioelectricsignalsitismeasuredinmV,µV,mA,µAorfT.(MEG) Duration–Thedurationmeansthelengthofthesignal(timevalue-horizontalaxis)oraparticalurtimeintervalofthesignal.Forexamplethedurationofanactionpotentialisaround1ms. Latency–Thelatencyisthetimedelaybetweentheonsetofastimulusandtheresponseittriggers.Thetermlatencyisusedforexamplebyevokedpotentialsorbycalculatingthenerveconductionspeeds. Phase–Incaseofperiodicsignalsthephaseofthesignalisthefractionofacompletecycleelapsedasmeasuredfromaspecifiedreferencepointandoftenexpressedasanangle. The mentioned properties are displayed on Figure12. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 28 www.itk.ppke.hu signal_pro.png Figure12. –Properties of an EMG signal (stimulating the elbow and recording the response) ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 29 www.itk.ppke.hu CLASSIFYING THE METHODS Signalanalysingmethodscanbegroupedinseveralways.Forexamplemethodscanbeperformedeitherinthetimedomainorinthefrequencydomain.Anotherclassificationwayisparametricversusnonparametric. • Methods (mainly) in the time domain: Amplitude analysis, Period analysis, Auto-and cross-correlation, Hjorth slope descriptors, Phase analysis, Autoregression, Mimetic analysis, Signal averaging, Covariance, Current source density analysis • Methods (mainly) in the frequency domain: Frequency analysis, Cross-spectrum, Coherence, Causality analysis, Inverse filtering, Kalman filtering, Complex demodulation • Nonparametric methods: Amplitude analysis, Period analysis, Auto-and cross-correlation, Complex demodulation, Coherence, Spectral analysis, Hjorth slope descriptors • Parametric methods: Inverse filtering, Kalman filtering, Mimetic analysis, Template matching ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 30 www.itk.ppke.hu SIGNAL AVERAGING Signalaveragingisatechniqueusedinthetimedomaintoincreasethestrengthofthesmallamplitudesignalsofinterestburiedinnoise.Withaveragingwecanincreasethesignal-to-noiseratio(S/N)byaproportiontothesquarerootofthenumberofmeasurements(N).Bestexamplesoftheapplicationsofsignalaveraginginelectrophysioliogyaretheevokedpotentials.(Figure13.)ForexampleletsassumethatwehaveanEEGrecordingwith40µVbackgroundactivityandanauditoryevokedpotentialwith10µVamplitudehiddeninthebackgroundactivityoftheEEG.SointhiscasetheS/Nis1.ToincreasetheS/Nto2/1(8xincrease)wehavetoaverage64responsesorepochs.(epochsareshortsegmentscutoutfromthecontinuousrecordingswithdefinedtimeintervals,incaseoftheevokedpotentialsthismeansusuallyaseveralhundredmsofthemeasurmentsbeforeandafterthestimuli)Theidealaveringhastofulfilseveralconditions:thesignalandnoiseareuncorrelated,thetimingofthesignalisknown,thesignalistimelockedtothestimulusandthenoiseisrandomwithzeromean. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 31 www.itk.ppke.hu average.png Figure13.Howtocreateanaverage.Auditoryevokedpotentialofasleepingcatto1/sclickstimuli.Thelocalfieldpotentialwasrecordedfromtheauditorycortexofthecatwitha24contactmultielectrode.Thesignalaveragingwasperformedononeselectedchannel. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 32 www.itk.ppke.hu FREQUENCY (SPECTRAL) ANALYSIS Severalmethodsforanalyzingthebioelectricsignalsworkinthetimedomain,becausethesetypeofsignalsaremainlytimeseries.Howeverthereisanotherimportantapproachtogetoutinformationfrombioelectricrecordings:analysisinthefrequencydomainorwithanotherwords,spectralanalysis. TheFourierseriestechniqueisusedtodecomposeperiodicfunctionsintotheircosineandsinecomponents.Forexampleatimedomainsignal,thesquarewave,canbedecomposedintofivesinewaves,eachwithadifferentfrequencyandamplitude.FromthecomplexFourierseriesatransformationcanbederived,whichtransformsdatafromthetimedomainintothefrequencydomain.ThisoperationiscalledFouriertransformandwithitshelpofit,wecanexaminethesignalsinthefrequencydomain.TheFouriertransformincontinuoustime(analogsignals)isreferredtoasthecontinuousFouriertransform(CFT).Indiscretetime(digitizedsignals)itiscalleddiscreteFouriertransform(DFT).Anefficientalgorithmusedbycomputerprograms,thatcalculatestheDFTistheFastFourierTransform(FFT). ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 33 www.itk.ppke.hu FREQUENCY (SPECTRAL) ANALYSIS Fourier series: Complex Fourier series: Continuous Fourier transform: Discrete Fourier transform: fourier_series.png complex_FS.png ft.png dft.png fs_coeff2.png complex_fs_coeff.png fs_coeff1.png ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 34 www.itk.ppke.hu FREQUENCY (SPECTRAL) ANALYSIS Therawcomplex-valuedoutputoftheFouriertransformsisdifficulttointerpretdirectly.Herearethreeapproachesforinterpretation: Powerspectrum:ItiscomputedbymultiplyingtheFFToutputwithitscomplexconjugate.Itcanbenormalizedbydividingbythenumberofdatapoints. Amplitude spectrum: It is the square root of the power spectrum. Phasespectrum:ItisthearcustangentofthequotientofimaginaryandrealpartsoftheFFT. SpectralanalysisisoftenusedinEEGanalysistoevaluatetheclassicalEEGfrequencybands(delta,theta,alpha,beta,gamma).ThefrequencydomaincharacteristicsintheEEGdataarerelevantbecauseoftheclinicalsignificanceofthevariousrhythms.ThespectruminFigure14.showsaclearpresenceofthealpharhythm,oneofthemostobviouscomponentsintheEEGinawakesubjectswithbotheyesclosed.Figure15.showsanEMGrecordingwiththecalculatedfrequencyspectrum. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 35 www.itk.ppke.hu FREQUENCY (SPECTRAL) ANALYSIS Inphysiologicalsignals,interpretationofspectrarequirescautionbecausethesetimeseriesarerarelystationaryandusuallycontainbothnonperiodicandperiodiccomponents.Notallpeaksinthespectrumdirectlycorrespondtoactualphysiological,periodicprocessesinthesystemathand.Itmaycontainlowfrequencycomponentsduetoslownonperiodicactivity(e.g.,trends)orhighfrequencycomponentsmaybecontaminatedbyhigh-frequencynonperiodicprocesses(e.g.,suddenevents).Becausetheperiodicactivityinphysiologicalsignalsisusuallyfarfrompurelysinusoidal,spectralcomponents,calledharmonics,canalsoappearathigherfrequencies. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 36 www.itk.ppke.hu frequency_spectrum_eyes_closed.png Figure 14. Frequency spectrum of an EEG recording (Pz electrode). During the recording the eyes were closed, wich resulted in an increased alpha activity (red color). ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 37 www.itk.ppke.hu ip_spectrum.png Figure15.EMGrecordingofaninterferencepatternrecordedfromthebicepsmuscle.(upperimage)Thefrequencyspectrumoftheinterferencepatternisshownonthebottomimage. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 38 www.itk.ppke.hu COVARIANCE Covarianceisameasureofhowmuchtworandomvariablesorsetsofrandomvariables(forexamplethetimeseriesofbioelectricsignals)changetogether.Betweentworandomvariablesthecovarianceis: Cov(X,Y) = E[(X-E[X])(Y-E[Y])], whereE(X)andE(Y)aretheexpectedvaluesofXandY.IfXandYarerandomvectors(withdimensionmandn)thanwegetthemxncovariancematrixthefollowingway: Cov(X,Y) = E[(X-E[X])(Y-E[Y])’], whereM’isthetransposeofM.The(i,j)-thelementofthismatrixisequaltothecovarianceCov(Xi,Yj)betweenthei-thscalarcomponentofXandthej-thscalarcomponentofY.Randomvariableswithzerocoveriancearecalleduncorrelated.Thecovariancematrixisusedinseveralalgorithmsrelatedtobioelectricsignalanalysis.(e.g.calculatingthecovariancematrixoftheactivityrecordedfromtwodifferentlocationsofthebrain) ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 39 www.itk.ppke.hu CORRELATION ThecorrelationorPearsonproduct-momentcorrelationcoefficientisameasureofthedegreeoflinearrelationshipbetweentwovariablesXandYorsetsofrandomvariable(forexamplethetimeseriesofbioelectricsignals).Thecorrelationisdefinedas: wherecov(X,Y)isthecovariancebetweenXandY,and.Xand.Yarethestandarddeviations.ThevalueofthePearsoncorrelationis+1inthecaseofmaximalpositivecorrelationand-1inthecaseofmaximalnegativecorrelation(anticorrelation).Ifthecorrelationiszero,thanthetwovariablesareuncorrelated.Independentvariablesarealwaysuncorrelated,butiftwovariablesareuncorrelated(i.e.theyhaveacorrelationvaluezero)wecannotsayforsurethattheyareindependent!Thecloserthecoefficientistoeither-1or1,thestrongeristherelationshipbetweenthetwovariables. cov.png ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 40 www.itk.ppke.hu CORRELATION IfwehavetwotimeseriesofbioelectricsignalsXandYwithnsampleswrittenasxiandyiwherei=1,2,...,n,recordedfromtwodistinctlocations,thanthesamplecorrelationcoefficientiscalculatedthefollowingway: wherexandyarethesamplemeansofXandY,sxandsyarethesamplestandarddeviationsofXandY.Thismethodgivesjustonenumberasaresult,sotogainmoreinformationabouttherelationshipbetweentwosignalstheuseofthecross-correlationissuggested. cor.png ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 41 www.itk.ppke.hu CROSS-CORRELATION Cross-correlationisameasureofsimilarityoftwowaveformsasafunctionofatime-lagappliedtooneofthem.Forcontinuousfunctions,fandg,thecross-correlationisdefinedas: wheref*denotesthecomplexconjugateoff.Fordiscretefunctions,thecross-correlationisdefinedas: Tounderstandhowcross-correlationworks,letsassumetworealvaluedfunctionsfandg,thatdifferonlybyanunknownshiftalongthex-axis.Thecross-correlationfindshowmuchgmustbeshiftedalongthex-axistomakeitidenticaltof.Theformulaslidesthegfunctionalongthex-axis,calculatingtheintegraloftheirproductateachposition.Whenthefunctionsmatch,thevalueofismaximized.ForexamplesseeFigure16.-18.Theauto-correlationisthecross-correlationofasignalwithitself. cross-corr1.png cross-corr2.png ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 42 www.itk.ppke.hu crosscorr2.png Figure16.Twosinewaveswithidenticalfrequenciesshiftedby5timeunitsalongthex-axis.Thebottomgraphisthecross-correlationofthesinewavesshowingamaximumatT=5andrepeatingatmultiplesofthesinewavefrequency.Withthismethodwecanstudyforexamplehowmuchisthesynchronicitybetweentheactivitesofdifferentbrainregionsinthedeltafrequencyband. wikimedia_logo.bmp ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 43 www.itk.ppke.hu crosscorr1.png Figure17.Examplesoftwoidenticalpulsesshiftedby5timeunits.Bottomgraphisthecross-correlationofthepulsesshowingamaximumatT=5.Thesetwopulsescanrepresentforexampletwoactionpotentialsgeneratedbyneurons. wikimedia_logo.bmp ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 44 www.itk.ppke.hu crosscorr3.png Figure18.Examplesofawide-band,timeseriessignalsproducedwitharandomnumbergeneratorandasmoothingfilter.Thetwosignalsareshiftedtotherightby5timeunits.Bottomgraphisthecross-correlationofthesignalsshowingamaximumatT=5.Thesewide-bandsignalsaregoodrepresentationsofrealbioelectricsignals. wikimedia_logo.bmp ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 45 www.itk.ppke.hu COHERENCE Coherenceisanormalizedmeasureoflinearcorrelationasafunctionoffrequencyandisdefinedbetweentwosignalsx(t)andy(t)as: whereGxyisthecross-spectraldensity(orcross-spectrum)betweenxandy,andGxxandGyytheautospectraldensityofxandyrespectively.Thecross-spectrumistheDFTotthecross-correlation,whiletheautospectraldensitiesaretheDFTsoftheautocorrelations.Valuesofcoherencearebetweenzeroandone.HighcoherenceimpliesthatamplitudesatagivenfrequencyarecorrelatedforexampleacrossEEGsamples.OnFigure19.thecoherenceofthreeelectrodesreferencedtotheFzelectrodeisshown.The1minutelengthEEGdatawasrecordedwith29electrodesandtheeyeswereclosedduringtherecordingsession.Thisresultedintheappearingof8-13Hzalphaoscillations,mainlyontheoccipitalandpariatalrecordingsites. coherence.png ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 46 www.itk.ppke.hu cov_alpha.png Figure19.ThecoherencebetweenCz-Fz,T4-FzandOz-Fz.DuringtheEEGrecordingtheeyeswereclosed,soadominantalphaoscillationispresentonmostoftherecordingsites.Thegreatercoherencevaluesrelatedtoalphaactivitycanbeobservedonallofthethreecoherencediagrams.(redcolor) ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 47 www.itk.ppke.hu LAPLACE TRANSFORM TheLaplacetransformisusedintheanalysisofcontinuoustimesystemsandisrelatedtotheFouriertransform.Theideausingtransformationsisthatintherearepropertiesinthetransformeddomainthatmakesomeproblemseasier.TheLaplacetransformisdefinedas: Itcanbeusedtosolveordinarydifferentialequations(ODE),forexampleitcanbeappliedtothesimplifiedionchannel(membrane)model: wherewhereRandCareconstantscorrespondingtothemembraneresistanceandcapacitance,respectively. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 48 www.itk.ppke.hu Z TRANSFORM TheztransformistheequivalentoftheLaplacetransformfordiscretetimeanditisusefulforanalyzingdifferenceequations.Z-transformconvertsadiscretetime-domainsignalintoacomplexfrequency-domainrepresentation: Theztransformhasananothermeaningtoo:standarizationorauto-scaling.Ifwewanttomaketwosamplescomparable(e.g.twonormaldistributions)thanztransformconvertsvaluesofsamplesintoz-scores: whereziarethez-transformedsampleobservations,xiaretheoriginalvaluesofthesample,xisthesamplemeanandsisthestandarddeviationofthesample.Figure20.showsanexampleforstandardizationbyztransform. ztransform.png ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 49 www.itk.ppke.hu ztransform2.png Figure20.Ztransformoftwonormaldistributionswithdifferentmeansandstandarddeviations.Afterthetransformationthetwonewnormaldistributionsbecamecomparablewithzeromeanandstandarddeviationwithvalueone. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 50 www.itk.ppke.hu ANALYSIS OF SINGLE-UNIT ACTIVITY Meanstheanalysisofneuralactionpotentialorspikes.Tounderstandhowthenervoussystemfunctions,wehavetounderstandfirstthespikeactivityoftheindiviualneuronsandafterthathowthisactivityisrelatedtootherneuralcellsinthenetwork.Theneuronalspikeactivitycanberegardedasapointprocess.ThePoissonprocessisarandompointprocesswheretheevents(actionpotentials)areidenticallydistributedbuttheintervalsbetweentheeventsarenotindependent.ThefunctionofneuronscanberegardedasaPoissonprocess. Therearetwotypesofsingle-unitactivityrecording:wecanrecordspontaneouslyfromagivenlocationofthebrain(Figure21.)orwerecordtheevokedunitactivitybysensoryorelectricalstimulation. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 51 www.itk.ppke.hu Figure 21. Single-unit activity recorded with a tetrode sua.png ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 52 www.itk.ppke.hu SPIKE SORTING Duringextracellularrecordingstheelectrodeisinsertedsomewhereintotheextracellularmatrixofthebrainandrecordstheactivityofseveralneurons.Thenumberofcellsdependsonseveralfactors(electrodelocationinthebrain,sizeoftherecordingsites,typeoftheelectrodeetc.),butinaveragethe„eyeshot”ofanelectrodeisabout150µm.Sotheoreticallywecanrecordtheactionpotentials(AP)ofneuronsfromthevolumeofacylinderwitharadiusof150micrometers.Inthecaseofcorticaltargetsthismeansapproximatelythousandneuralcells,mainlypyramidalcells.Howevertheampitudeoftheelectricalpulsesofcellsontheedgeofthecylinderissosmall,thatitfadesintothebackgroundactivity.Becausetheaimofspikesortingistodeterminewichspike(AP)correspondstowichneuron,onlytheneuralactivityneartotheelectrodecanbeconsideredpracticallyusefulforspikesorting.InpracticeAPswithanamplitudeofmininum60µVarebigenoughforreliablespikesorting,thismeansrecordingtheactivityofabout100neuronsfromthevolumeofacylinderwith100µmindiameter.(Figure22.) ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 53 www.itk.ppke.hu Figure 22. The „eyeshot” of a tetrode eyeshot.png ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 54 www.itk.ppke.hu Eachneuronhasactionpotentialsofcharacteristicshapewichdependsonthedistanceandspatiallocalizationoftheelectroderelativetotheneuronanditsmorphology.Becausethecellsareatdifferentdistancesfromtheelectrodeandalsohaveadifferentmorphology,thefeaturesoftheextracellularactionpotentialcanbeusedtodifferentiatetheindividualcells.Theextracellularspikecanbeapproximatedwiththenegativefirstderivativeoftheintracellularactionpotentials(atleastitsinitialnegativephase,Figure23.).Inanidealcaseonlytwoneuronsarevisible,thesignalisfiltered,sonoslowactivityispresent,andtheshapeoftheactionpotentialdiffers(Figure24.)Spikesortingcanbeperformedonline,duringtherecordingswithanamplitudeorawindowdiscriminatororwithofflineanalysiswhichmeansprocessingthedatawithspikesortingalgorithmsconsistingofafewsteps. Spikesortingallowstheanalysisoftheactivityofafewclose-byneuronsfromtherecordingelectrode,sotheconnectivitypatternsoftheseneuronscanbestudiedwhichgivesusvaluableinformationsaboutthefunctioningofthebrain. WHY IS SPIKE SORTING POSSIBLE? ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY TÁMOP –4.1.2-08/2/A/KMR-2009-0006 55 www.itk.ppke.hu 0 -60 mv 0 -60 mv A B Intracellular recording Extracellular recording A B A B Figure 23. Difference between extracellularly and intracellularly recorded action potentials. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY TÁMOP –4.1.2-08/2/A/KMR-2009-0006 56 www.itk.ppke.hu Intracellular recording A B Extracellular recording ABAABBBA Figure 24. Spike sorting in ideal case ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 57 www.itk.ppke.hu THE ALGORITHM OF SPIKE SORTING 1. Filtering the recordings 2. Spike detection (voltage thresholding) 3. Feature extraction or feature selection 4. Cluster analysis 5. Cleaning the cell clusters Other issues: • Role of the interspike-interval (ISI) histogram • Analysing the cross-correlation histrograms • Tetrodes • Difficulties with spike sorting: Overlapping spikes, Bursting cells, Electrode drift, Rarely spiking neurons etc. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 58 www.itk.ppke.hu SPIKE SORTING –FILTERING Thisisthefirststepofspikesorting:applyingabandpassfiltertotherecordedwidebanddatatoremovetheunnecessarylowfrequencyactivity(likedeltaorgammaactivity,artifactsetc.)andtomaketheactionpotentialsvisible.Noncausalfilterswithabandpassbetween300and3000Hzareusedusually,becausethedurationofspikesisabout1mslong.Thelowercutofffrequencyis300Hz,whichmeansthatfrequenciesbelow300Hz(slowcomponentsoftherawdata)arefilteredout.Theuppercutofffrequencyofthefilter(3000Hz)istoremovenoisyelementsfromthespikeshapes. AnexampleoffilteringispresentedonFigure25.:fourchannelsofarawwidebanddataarefilteredofflinewithaFIRbandpassfilterbetween300and3000Hz.Afterthefilteringthespikesbecameclearlyvisibleanddistinguishable. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 59 www.itk.ppke.hu 1-5000Hz 300-3000Hz Wideband local field potential Frequency band of action potentials Filtering Figure 25. -Filtering out the action potentials from local field potentials ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 60 www.itk.ppke.hu SPIKE SORTING –SPIKE DETECTION Spikedetectionisperformedwithavoltage(oramplitude)threshold.Agoodchoiceonthethresholdlevelisveryimportant:ifthethresholdistohightheactionpotentialswithloweramplitudesaremissedorifweusealowthresholdfordetection,noisecomponentsofthesignalwillbemarkedasspikes(falsepositives).Thethresholdlevelcanbesetmanuallyorautomatically.AutomaticthresholdinglevelscanbecalculatedasamultipleofthestandarddeviationofthesignalorwiththemethoddevelopedbyQuianQuiroga: wherexisthebandpassfilteredsignalandisanestimateofthestandarddeviationofthebackgroundnoise.Afterthedetectionthedetectedspikes(events)arestored(usually64datapointsarestored,whichmeansa2-3mssegmentofthedataaroundthedetectedevents).Finallythespikeshapeshavetobealigned,inmostofthecasestheyarealignedtotheirmaximum.(Figure26.-27.) képlet.png 5d44df42e157df5a3186ce5793857ecc.png ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 61 www.itk.ppke.hu Extracellular recording Figure 26. Threshold detection and storing the events. Voltage threshold Event 1 2 3 4 5 6 7 8 ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 62 www.itk.ppke.hu Figure 27. Aligning to the peaks. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 63 www.itk.ppke.hu SPIKE SORTING –FEATURE EXTRACTION Theaimofthefeatureextractionisdimensionalityreduction:thedecreaseofthedimensiondeterminedbythenumberofdatapointstoalowerdimensionalspace(usuallytwoorthreedimension)ofafewfeatures.Thechoiceofthebestfeaturesisnotaneasytask.Thesimplestwayistotakethebasiccharacteristicsofspikes:peakamplitude,peak-to-peakamplitude,widthofthespike,oritsenergy.However,inmanycasesthesefeaturesdonotgivegoodresultsindifferentiatingthespikeshapes.Themostusedmethodforfeatureselectionistheprincipalcomponentanalysis(PCA).Thefirst2or3principalcomponentsareusedasparameters,whichcontainmorethan80%oftheenergyofthesignal.Otherpossibilitiestoselecttheparametersforseparatingtheclustersofactionpotentialsoriginatingfromdifferentneuronsaretheindependentcomponentanalysis(ICA)ortheuseofwavelets(discretewavelettransformation-DWT).Inanidealcasethetwodifferentspikeshapesareeasilydistinguishable,sotwofeaturesdependingonthecharacteristicsoftheactionpotentialsaregoodenoughforaproperspikesorting(Figure28.) ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 64 www.itk.ppke.hu Figure28.Featureextraction:twofeaturesdependingonthecharacteristicsofthespikearechoosenforspikesorting e.g. peak amplitude (x-axis) amplitude of the descending leg (y-axis) A B ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 65 www.itk.ppke.hu SPIKE SORTING –CLUSTERING THE DATA Afterwehaveselectedtheappropriatefeatures,wehavetogroupspikeswithsimilarfeaturesintodifferentclusters.Oneclustercorrespondstooneindividualneuron.Clusteringcanbeperformedmanuallybydrawingpoligonsorellipsesaroundtheclustersformedbytheselectedfeaturesinthe2-dimensionalspaceorbydelimitingtheclusterswithspheresinthe3Dspace.Thislow-levelsolutionofclusteringistimeconsumingandcanhaveerrors,forexampleiftheclustersoverlapthanthemanualselectionofclusterboundariesisverysubjective.Semi-automaticalgorithmsneedjustasmallinterventionfromtheusers,forexampletosetsomeinitialparameters.TheK-Means,thestandardExpectation-Maximization(E-M)ortheSuper-ParamagneticClusteringbelonghere.AutomaticalgorithmsliketheValleySeekingortheT-distributionE-Malgorithmareveryfast,butworkeffectiveonlyintheidealcases,whentheclustersaregoodseparated.Figure29.showstheidealcasewhentwoclearlydistinguishableclustersarepresent. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 66 www.itk.ppke.hu Figure 29. Displaying by the parameters -Classifying into groups (Clustering) x y A B A B Clustering amplitude of the descending leg (y-axis) e.g. peak amplitude (x-axis) ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 67 www.itk.ppke.hu Extracellular recording Voltage threshold Event 1 2 3 4 5 6 7 8 AB A A B B B A Figure 30. The end result of the spike sorting in the ideal case 1....A 2....B 3....A 4....A 5....B 6....B 7....B 8....A ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 68 www.itk.ppke.hu SPIKE SORTING –CLEANING THE CLUSTERS Inrealcases,thespikesortingisnotaneasytask,likeitisintheidealcasepresentedbefore,supervisionoftheclustersisadvisedaftertheclustering.Errorscanbeinducedforexampleatthestageofthresholddetectionorbymanualorautomaticclustering.Erroriswhenaspikeofaneuronisassignedtoananotherneuronbecausetwoclustersoverlapinthefeaturespaceandthechoicetowhichclusterthisspikebelongsisnotobvious.(Figure31.)Wrongassignmentsofthistypecanbepartlycorrectedbycheckingtheinterspikeintervalhistogramsoftheclustersorbyexaminingmorefeaturespaces,whereperhapsthistwoclustersdonotoverlap.Artifacts,thataremoreorlesssimilartotheactionpotentials,canalsoinfluencethesorting.Fortunatelyinmostofcasestheseartifactsareoutliersinthefeaturespace,butsomeclusteringalgorithmscanassignthese„falsespikes”tosomeclustersdespitethis.Sometimesautomaticalgorithmsseparateoneclusterofaneurontotwosmallerclusters,butbycheckingforsimilarclustersafterclusteringandbymergeingtheseparatedclustersafterthatwecancorrectthesetypeoferrors. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 69 www.itk.ppke.hu x y A B x y A B Figure31.Twodifferentsolutionsforclusteringthespikesoftheneurons,wherethetwoclustersoverlap.Incase1,oneofthespikes,thatisattheedgeofthetwoclustersbytheoverlappingpart,isassignedtwoclusterA.Incase2thesamespikeisassignedtwoclusterB. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 70 www.itk.ppke.hu SPIKE SORTING –OTHER ISSUES Interspikeinterval(ISI)histogram–TheISIhistogramisagoodhelptocheckwhetherthereareincorrectspikeassignmentsintheclusters.Itdisplaysthenumberofspikesfollowingagivenspikeincertaintimeintervals.Itiswell-known,thatafteradischargethereisa1mslongabsoluterefractoryperiodandarelativerefractoryperiodlastingfor2-3ms,whentheneuron„recovers”forthenextactionpotential.SoifapeakispresentintheISIhistogrambefore5ms,thereisabigchancethatwehavespikescorrespondingtootherneuronsintheexaminedcluster.(Figure32.A) Cross-correlogram–Withthistypeofhistogramtherelationshipbetweentheclusterscanbeexamined.ForexampleifaneuronBthatisfiringusuallyshortlyafterneuronA,thanontheircross-correlogramwecanobserveapeakatthetimewheretheprobabilityofthedischargeofneuronBisthehighest. Tetrodes–Tetrodesaremicroelectrodeswithfourcontactsclosetoeachother,sothedifferentcontacts„see”thedischargingneuronsfromdifferentdistances.Thiscanbeusedtoimprovethesortingqualities.(Figure33.) ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 71 www.itk.ppke.hu ISI.png Figure32.A–ISIhistogramofaclusterthatwasbadclustered:youcanseea peakbefore5mswith400events.Soabout15%ofthespikesdischargeinthe refractoryperiod.Thisisatoohighpercenttoacceptthisclusterasgood,butafew percentofthespikesareallowedintherefractoryperiod,becauseperfectspike sortinginrealcasesisalmostimpossible.TheISIofagoodclusterispresentedonB. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 72 www.itk.ppke.hu Figure 33. The side-view and front-view of a tetrode tetrode.png ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 73 www.itk.ppke.hu DIFFICULTIES WITH SPIKE SORTING Overlappingspikes–Whentwoneuronsdischargeatthesametimeorwithasmalldelay,theirspikeswilloverlap,resultinginaspikeshape,thatisthesumoftheactionpotentialsofthetwocells.(Figure34.)Dealingwithoverlappingspikesisaverychallengingandanimportanttask,severalsolutionsweredeveloped,butamajorbreakthroughisstillneeded. Burstingcells–Aburstingneuronisfiringtwoormorespikesoneafteranother,inafastsequence.(Figure35.)Theamplitudeofthespikeschangesthroughthebursting:usuallyitdecreases,butanincreaseisalsopossible.Severalalgorithmsseparatethespikesofburstingcellsindifferentclusters,becauseofthedifferenceintheamplitudes.TheISIhistogramandthecross-correlogramcanhelptoidentifyandmanageburstingneurons.ThereisapeakpresentontheISIhistogramat2or3ms,ortherecanbemorepeakswhenaburstconsistsofmorethantwospikes.Iftheburstingneuronwasseparatedindifferentclusters,thanthecross-correlogramofthetwoclustersshowsastrongrelationshipbetweenthem(peakat2or3ms). ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 74 www.itk.ppke.hu overlapping spikes.png Figure 34. Examples of overlapping spikes (A, B, C) and a „good” spike (D) ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 75 www.itk.ppke.hu Figure35.Exampleoftwoburstingneuronsrecordedfromthethalamusofananesthetizedrat.Bothoftheneuronsdischargefourtimeswithdecreasingactionpotentialamplitudes.TheshapeoftheAPsremainmoreorlesssimilarduringtheevolutionoftheburst. burst.png ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 76 www.itk.ppke.hu DIFFICULTIES WITH SPIKE SORTING Electrodedrift–Thebraincanslowlymovearoundtheelectrode(thebrainpulsatesforexamplewhenatoobigwindowisdrilledintotheskull),sointhiscasethedistancebetweentherecordingsiteandthedischargingneuronisnotconstant.Thisresultsinthechangeofamplitudeofthespikes.(Figure36.)Iftheelectrodedriftiscontinuous,thanitcanhappenthattheamplitudeofspikessuitableforsortingcandecreaseinsucharate,thatnomoreaccuratesortingispossibleorwecanalsocompletelylosethesignalofthenearbyneuron. Non-Gaussianclusters–MostofspikesortingalgorithmsassumeaGaussiandistributionoftheclusters,butusuallythisconditionisnotfulfilledingeneralbecauseofseveralreasons(overlappingspikes,burstingneurons,electrodedrift,multi-unitactivity,non-stationarybackgroundnoise,correlationbetweenspikesandlocalfieldpotentials),producingelongatednon-Gaussianclusters. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 77 www.itk.ppke.hu electrode drift.png Figure36.Exampleofelectrodedrift(a3minutelongsegmentofatetroderecordingfromthesomatosensorycortexofananesthetizedratisshownatthetop)Theinitialamplitudeoftheactionpotentialsofthecellareslowlydecreasingbecauseoftheelectrodemovementinthebrain. electrode drift_AP.png ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 78 www.itk.ppke.hu ADVANCED SIGNAL PROCESSING METHODS • Wavelet analysis • Principal component analysis (PCA) • Independent component analysis (ICA) • Current source density (CSD)analysis • Compressed EEG spectral array • EEG source localization methods • Quantitative EEG (Brain mapping) • Analysis of event-related potentials • Nonlinear techniques ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 79 www.itk.ppke.hu WAVELET ANALYSIS Awaveletisawave-likeoscillationwithanamplitudethatstartsoutatzero,increases,andthendecreasesbacktozero(Figure37.).Waveletanalysisisveryusefulforanalyzingphysiologicalsystemsbecause,asopposedtomostclassicalsignalanalysisapproaches,itprovidesthemeanstodetectandanalyzenonstationarityinsignals.Whenperformingspectralanalysisonasampledtimeseries,thespectrumrevealsfrequencycomponentsintheinputsignal.Becausethespectrumrepresentsthewholetimedomainepoch,itisuncertainwhereexactlyanyparticularfrequencycomponentislocatedintime.IncaseofFourier-basedspectralanalysisanychoiceoftheepochlengthisalwaysassociatedwithacompromisebetweentimeandfrequencyresolution,itisimpossibletochooseanepochlengththatwillaccommodatebothahightemporalandahighspectralresolution.Averyhightemporalresolution(smallepoch)isalwaysassociatedwithalowspectralresolutionandviceversa.Thewavelettransformationeliminatesthisproblem:ithasanaccuratetimeandfrequencyresolutionatthesametime.(Figure38-39.) ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 80 www.itk.ppke.hu wavelet_morlet.png Wavelet_-_Meyer.png Wavelet_-_Mex_Hat.png Figure 37. –Classical wavelets: Morlet, Meyer and the mexican hat. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 81 www.itk.ppke.hu Figure 38. –Spectra of single sweeps single sweep EEG single sweep TFR absolute wavelet power baseline baseline corrected single sweep TFR relative wavelet power Time Frequency 100 50 1Hz 0 4sec ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 82 www.itk.ppke.hu Figure 39. –Averaged spectra -1500 -1000 -500 0 500 1000 1500 10 20 30 40 50 60 70 80 90 100 20 25 30 35 40 45 50 55 60 65 70 -1500 -1000 -500 0 500 1000 1500 10 20 30 40 50 60 70 80 90 100 1 -10 -8 -6 -4 -2 0 2 4 6 8 10 Averaged absolute wavelet power Averaged relative wavelet power baseline µV 100 200 -100 -200 Averaged ERP µV 100 200 -100 -200 Averaged ERP Frequency (Hz) Frequency (Hz) Time (ms) Time (ms) ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 83 www.itk.ppke.hu PRINCIPAL COMPONENT ANALYSIS (PCA) PCAcanbeusedtoreducethenumberofvariables(dimensionalityreduction)ortodetectstructureintherelationshipsamongvariables.Itisanorthogonaltransformationwheretheendresultsareuncorrelated,orthogonalvariables,calledprincipalcomponents.Thefirstprincipalcomponenthasthehighestvarianceamongtheotherprincipalcomponents,thesecondprincipalcomponenthasthesecondhighestvariance,etc.Thefirstthreeprincipalcomponentscontainappromixately80%ofthevarianceofthesignal.PCAcanbeperformedbyeigenvaluedecompositionofadatacovariancematrixorbysingularvaluedecompositionofadatamatrix.PCAhasmanyapplicationsonthefieldofbioelectricsignals:itcanbeusedinEEGsourcelocalization,todetermineandremoveunwantedbackgroundactivitiesandartefactstomakedipolelocalizationsmoreaccurate.Itispracticalforremovingartefacts(e.g.blink)fromevokedpotentials.Incaseofspikesortingitisusedtoreducethedimensionofthemultidimensionalspikeshapesforfasterandeasierclusteringofactionpotentials.(fromaround64dimensionsto2or3dimensions) ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 84 www.itk.ppke.hu pca.png Figure40.Thefirstthreeprincipalcomponents(picturesontheright)ofthreedifferentspikeclusters(picturesontheleft,severalhundredoverlayedspikes).Thethicknessofthelinesoftheprincipalcomponentsrepresentthemagnitudeoftheeigenvalues.Thefirsteighteigenvaluesaredisplayedintheupperrightcornerofthepicturesontheright.Theprincipalcomponentsofthefirsttwoclusters,whichweresortedfromthesamerecording,areverysimilar,perhapsthespikesofthesetwoclustersoriginatefromoneneuron,thatburstswithchangingspikeamplitude.Thethirdexampleatthebottomwasrecordedfromadifferentbrainlocation. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 85 www.itk.ppke.hu INDEPENDENT COMPONENT ANALYSIS (ICA) ICAhassimilarapplicationsrelatedtobioelectricsignals,thanPCA.TheexamplesmentionedbythePCAarealsogoodexamplestotheICAmethod:itgivesgoodfeaturesforthereductionofthedimensionsinspikesortingandcanbeusedtodetectandremoveundesiredartefacts(likeocular,movement,ECGartefacts)andbackgroundactivitiesfromthesignalsforamoreaccuratesourceanalysis.HoweverICAisamoresophisticatedmethodcomperedtoPCA:itgeneratespatternsandloadingsusingastrictercriteriaforstatisticalindependence.Awell-knownprincipleisthatdifferentphysicalprocessesgeneratestatisticallyindependentsignals.TheEEGrecordedfromthescalpisthesummationofsignalsoriginatingfrommultiplesources.ICAcomputesindividualsignalsthatarestatisticallyindependent,andwhicharethereforelikelytohavebeengeneratedbydifferentphysiologicalprocesses.AcommonlyusedalgorithmforICAisagradient-descentmethodcalledINFOMAXandisbasedonmaximizationoftheentropy. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 86 www.itk.ppke.hu COMPRESSED EEG SPECTRAL ARRAY (CSA) Thecompressedspectralarrayshowstemporalchangesinthepowerspectrumoftherecordedsignals.Spectralanalysis(FFT)onabout1minutelongEEGsegments(epochs)areperformedateveryminute,andthesespectrumsareplottedbeforeeachotherformingasemi-three-dimensionalgraph,withthefrequencyonthex-axis,timeonthey-axisandpoweronthez-axis.(Figure43.)Thistechniqueisusedmainlyduringsurgerytocontinuouslymonitorthedeepnessoftheanesthesia.Otherfeaturesliketheheartrate,bloodpressure,somatosensoryevokedpotential,brainauditoryevokedpotentialetc.aresupervisedtogetherwiththeCSA. OnFigure44.aCSAofa15minutesleepsessionofacatisdisplayed.Wecanobservethedecreaseofthepoweroflowfrequenciesrelatedtoslow-wavesleep(SWS)atthetransitionfromSWStorapid-eyemovement(REM)sleep.ThecatwokeupforacoupleofsecondsandafterthatcontinuedsleepingshortlyreachingSWSstateagain. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 87 www.itk.ppke.hu Figure43.Theschemeofacompressedspectralarray. CSA2.png ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 88 www.itk.ppke.hu csa.png Figure44.Thecompressedspectralarrayofa15minutesleepsessionofacat.Spectralanalysiswasmadeatevery45seconds.TheEEGwasrecordedfromthesurfaceoftheauditorycortex. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 89 www.itk.ppke.hu EEG SOURCE LOCALIZATION METHODS ManytimesthereisaneedtolocalizetheintracranialsourcesofthescalpEEG,forexampleinthecaseofapatientwithfocalepilepsywewanttoknowtheexactlocationsoftheepilepticfociinthebrainorinscientificexperimentstheplacesoforiginoftheEEGsignalscanbeveryimportant.WecoulddothelocalizationwiththehelpofconventionalEEG,butseveralfactorslimitaccuratelocalizationincludingthelargeinterelectrodedistancesortheattenuationanddistortionofthesignalsbythevolume-conductingmedium,thescalp,theskull,theCSFandthebrainitself. Tobegindiscovertheareaofsourcelocalizationwehavetofamiliarizeuswiththeforwardproblem.Theforwardproblemiswhenwewanttopredicttheelectricpotentialormagneticfieldvectorthatwouldbeexternallymeasuredwiththeelectrodesfromtheactivityofsomesourceswereinsidethebrain.TolocalizetheEEGsourceswehavetosolvetheinverseproblem:estimatingthecurrentdensityoractivityvaluesofthesourcesthatgeneratedthemeasuredelectricpotentialormagneticfieldvector. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 90 www.itk.ppke.hu EEG SOURCE LOCALIZATION METHODS Thereareseveraldifficultiesthatmakesourcelocalizationachallengingtask:forexampleartifactsandnoisecontaminationoftheEEGsignalorthataninfinitenumberofsetsofintracranialsourcesmayproduceexactlythesamepotentialdistributionontheheadsurface.Sothegeneralinverseproblemhasnouniquesolution,wehavetomakeconstraintsonthenumber,typeorlocationofthesources.Thiscanbeachievedbytheuseofdifferentsourcemodels.Thesearecalledmodel-dependentmethods. MODEL-DEPENDENTMETHODS ThesearedipolelocalizationmethodsandassumethattheEEGorevokedpotentialisgeneratedbyoneormoreintracranialdipolesources(equivalentdipoles).Themethodscanbeclassifiedintolinearandnonlineardipolelocalizationmethods. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 91 www.itk.ppke.hu EEG SOURCE LOCALIZATION METHODS NONLINEAR DIPOLE LOCALIZATION METHODS Asinglesourcedipolecanbecharacterizedbysixparameters(3positionparametersandthedipolemomentvector).Withtheseparametersandamodelofthevolumeconductingmedium,thepotentialateachpointonthescalpsurfacecanbecalculated.Anonlinearleast-squaresminimizationalgorithmisusedtocalculatethe6parametersofthesourcedipole,thatfitstothemeasuredpotentials.Thismethodissensitivetonoise,artifacts,inaccuraciesintheheadandiscomputationallydemanding.Similarmodelsrelatedtothisarethefixeddipolemodelandtherotatingdipolemodel.Otherdipolescanbeaddedtothemodelsuptoatheoreticalmaximumthatdependsonthenumberofelectrodes. Oneexamplefortheclinicalapplicationistheequvivalentdipolemodelingofepilepticspikepotentials.Itcandeterminepropagationpatternsofepilepticspikes. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 92 www.itk.ppke.hu EEG SOURCE LOCALIZATION METHODS LINEAR DIPOLE LOCALIZATION METHODS Linearmethodsassumealargernumberofdipolesandthesourcescanbedistributedmorewidelywithinthecortex.Thelocationsandorientationsofthedipolesourcesareknown,onlythestrengthofeachdipolesourceneedtobedetermined.Itisafastercomparedtononlinearmethods.Wecandistinguishthreesubgroupswithinlinearmethods.IfthereareNdipoles,andMelectrodes,thanincaseof: 1. NM-Thereareinfinitenumberofsolutionstotheinverseproblemandtheone„minimumnorm”solutionisfound.Examplemethods:corticalimagingtechnique(CIT),Low-resolutionElectromagneticTomography(LORETA) ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 93 www.itk.ppke.hu EEG SOURCE LOCALIZATION METHODS IncaseoflinearsourcemethodsthetemporalresolutionofEEGcanbecombinedwiththespatialresolutionofimagingmethods,likeMRIorSPECTtoachievebetteraccuracy. MODEL-INDEPENDENT METHODS Model-independentmethodsdonotrequireanyassumptionsaboutthenumber,type,orconfigurationofthesourcesinthebrain. Topographicdisplaymethods(brainmapping):Algorithmsthatcaninterpolatepotentialstointermediatepointsbetweenthescalpelectrodepositions.Examplesofsuchalgorithms: • Nearest neighbor inverse distance-weighted • All-electrode inverse distance weigthed • Rectangular surface splines or rectangular three-dimensional splines • Spherical surface splines • Spherical harmonic expansion • Single dipole or multidipole source model ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 94 www.itk.ppke.hu EEG SOURCE LOCALIZATION METHODS Laplacianmethodsareanothergroupofmodel-independentmethods,theseusesimilaralgorithmsmentionedbytopographicdisplaymethods. Multivariate Statistical Methods PCAandICA,thatweredescribedinaprevioussection,maybeusedtodecomposeanepochofmultichannelEEGintomultiplelinearlyindependentcomponents.TheoriginalEEGcanbereconstructedasalinearcombinationofallcomponents.Thesecomponentsmaybeusedasastartingpointfordipoleanaysisorothersourcelocalizationtechniques. SOURCE LOCALIZATION IN MEG TheaccuracyoflocalizationofintracranialsourcesincaseofMEGisnotlimitedbythesmearingeffectsofthevolumeconductingmediumonelectricalpotentials,becauseallthetissuesbetweensourcesandthemagneticfielddetectorsaretransparenttomagneticfields.Soasimplehomogenousspheremodelofthevolumeconductorisusuallysufficienttoobtainaccuratesourcedipolelocalization. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 95 www.itk.ppke.hu EEG SOURCE LOCALIZATION METHODS VOLUME CONDUCTOR MODELS Theonlyelectricalpropertyofthevolumeconductorthatisusuallymodeledistheelectricalconductivity.Modelsoftheheadasavolumeconductormaybeclassifiedashomogeneousorinhomogenous.Mostcommonmodelsarebasedonsphericalsurfaces.Thesimplestmodelisthehomogenousspheremodel:theheadisaperfectspherewithuniformelecticalconductivitythroughout.Thethreeshellsphericalmodelconsistsofthebrain,theskullandthescalp,allofthewithdifferentresistivities.(Figure45.)Thefourspheremodelhasanadditionalpart:thelayerofcerebrospinalfluid.Amorerealisticmodelcanbeachievedwiththefiniteelementmethod:importantheadregionsaredecomposedintomultiplesmalladjacentelements,approximately1to2cminlengthandthickness.Theactualgeometryofthesclap,skullandbrainsurfacesmaybeobtainedfromMRIdata.ThisvolumeconductormodelisusedforexampleinLORETA. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 96 www.itk.ppke.hu three_sphere_model.png Figure 45. The three shell spherical head model. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 97 www.itk.ppke.hu QUANTITATIVE EEG (BRAIN MAPPING, QEEG) QEEGreferstoacomplexanalysisofbrainwavefrequencybandwidthsthatmakeuptherawEEG.TheEEGdataacquiredareusedtocreatetopographiccolor-codedmapsthatshowelectricalactivityofthebrain.Itprovidescomplexanalysisofseveralbrainwavecharacteristicslikecoherence,power,dominantfrequency,symmetry,phase,andamplitude.ThefirstbrainmappingsystemtheBEAM(Brainelectricalactivitymapping)wasdevelopedbyDuffyinthe1970s.OnFigure46.thedistributionofthepowerofdifferentfrequencybandsonthescalparedisplayed. However,wehavetobecarefulwithbrainmaps,becauseincaseEEGrecordings,recordedwithasmallernumberofelectrodes(forexample19electrodes)thebrainmapsaregeneratedwithalotofinterpolationused,becausetheelectrodesrecordjustfromasmallareaandtheinterelectrodedistancesarerelativelybig.Interpolationsgivejustassumptionswhathappensinbrainareasnotcoveredwithelectrodes,soitisrecommendedtorecordthebrainactivitywithasmanyelectrodesaspossible. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 98 www.itk.ppke.hu brainmap.png Figure46.Brainmapsofthedifferentfrequencybandpowers.InparenthesestherearetherelativepowersofthefrequencybandsrelatedtoEEG. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 99 www.itk.ppke.hu Figure47.Brainmapsconstructedonthebasisofanonlinearmethod(synchronizationlikelihood–introducedlater).Thedistributionofsynchronizationbetweenthe33electrodestoareferenceelectrodeinthealpha1band(8-11Hz)inthecaseofeyesclosed(displayedontheleft)andinthecaseofeyesopened(ontherightside). a1_topo_cs_.emf a1_topo_ny_.emf ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 100 www.itk.ppke.hu ANALYSIS OF EVENT-RELATED POTENTIALS Event-relatedpotentials(ERP)orevokedpotentials(EP)areextractedfromthecontinuousEEGdata(Figure48.)usingsignalaveraging.First,smallsegments(calledepochsorsweeps)arecutoutfromtheEEG.Theseepochsareusuallyseveralhundredmillisecondslongandcontainthestartofthetask/stimulusandthebrainactivitythatprocessestheactualtask/stimulus.(Figure49.)TheamplitudeoftheERPsaresmallcomparedtothebackgroundactivity,soafteracarefulartifactrejectionprocess,wheretheunusablesweepsareremoved,theremainingonesareaveragedtoincreasethesignal-to-noiseratio.Asaresultthesmall-amplitudeERPcomponentsthatweretime-lockedtotheonsetofthetast/stimulusarerevealedandcanbefurtheranalized.(Figure50.)ForexampledifferencesofEPcomponentsbetweendifferentelectrodesitescanbeexaminedorthetheERPsoriginatingfromdifferentsubjectsorsubjectgroupscanbealsocompered.(Figure51.) ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 101 www.itk.ppke.hu EEG_sample.png Figure 48. A 4 sec segmentof a continuousEEG recording, recordedwith33 electrodes. The taskof thesubjectswastocountintheirhead: theyhad toadd numbers. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 102 www.itk.ppke.hu Figure49.–FromthecontinuousEEGpresentedonthepreviousfigurearecuttedout2secondlongsegmentscalledsweepsorepochs.Thesesweepsaretimelockedtotheonsetoftheactualtasks.Sointhiscasethecountingtaskbeganatzerotimeandfortheanalysisoftheevokedpotentialsa2secondlongpartoftheEEGisused. sweep2.png ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 103 www.itk.ppke.hu Figure50.–AonesecondlongaverageofEEGsweepsfromaGO/NOGOtask.AfteraveragingtheevokedpotentialburiedinthebackgroundEEGactivitycanbeexamined. avg.png ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 104 www.itk.ppke.hu Figure51.–TheevokedpotentialsoftwosubjectsdoingaGO/NOGOtaskattheFzelectrodecomparedtoeachother.Asignificantdifferenceintheamplitudesofthelatecomponentoftheevokedpotentialcanbeobserved(around400ms). avg2.png ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 105 www.itk.ppke.hu NONLINEAR TECHNIQUES Lineartechniquesweredesignedtodetectpropertiesorrelationshipswithinorbetweentimeseriesgeneratedbylinearsystemsbuttheseanalysistechniquescanfailwhenappliedtononlinearsystem.Thenumeroussuccesfulresultsachievedwithlinearmethodsshows,thatsometimesalinearprocessisagoodapproximationofthesystem’sbehavior,butthebraincanbeconsideredasanonlinearsystem,sothereisasignificantneedfornovelsignalprocessingtoolsforstudyingnonlinearrelationshipsinphysiologyaswellasacriticalnecessitytoevaluatethetoolsthathavebeendevelopedoverthepastdecades. NonlineartechniquescanbeusedtotheanalysisoftheEEGandthereareattemptstopredictanddetectepilepticseizuresortoautomaticallydetectthedifferentstagesofsleepwithnonlinearmethods. TwononlinearmethodsusedinEEGanalysis,OmegacomplexityandSynchronizationlikelihood,willbedemonstratedinthefollowingsection. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 106 www.itk.ppke.hu NONLINEAR TECHNIQUES –OMEGA COMPLEXITY .complexityisameasureofcomplexityofspatio-temporaldynamicsoftheelectricactivityofthebrain.ItcanbeusedwithmultichannelEEGrecordings.LetsassumewehaveKelectrodes.The.complexityiscalculatedfromtheKeigenvaluesofthecovariancematrixoftheEEGdata: where.i(i=1…K)aretheeigenvaluesofthecovariancematrix..complexityisminimal(withvalue1)ifthereisonlyasinglegeneratorpresentonalloftheKmeasuringpointsanditreachesitsmaximumifKuncorrelatedgeneratorsexistwithequalpower,oneforeachoftheKelectrodes.(Figure52.leftside)Thiscomplexitymeasurecanbeusedtostudybrainmacrostateswithdurationsofseveralminutesorlonger,forexampletoisolatedifferentsleepstages. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 107 www.itk.ppke.hu NONLINEAR TECHNIQUES –SYNCHRONIZATION LIKELIHOOD Thesynchronizationlikelihood(SL)givesastraightforwardnormalizedestimateofthedynamicalinterdependenciesbetweentwoormoresimultaneouslyrecordedtimeseries.Itissuitablefortheanalysisofnon-stationarydata,liketheEEGorMEG.TheSLisameasurewhichdescribeshowstronglychannelkattimeiissynchronizedtoallotherM-1channels.TheSLhasarangebetweenareferencevaluePand1,whereP<<1(usually0.05).IncaseofSLisequaltoP,thanallMtimeseriesareuncorrelated.IfSLisequaltoone,thanthesynchronizationbetweenallMtimeseriesismaximal.(Figure52.rightside)ModificationsofSLcanbeobtainedbyaveragingoverthetimeindexi,averagingoverthechannelindexkorboth. ExamplestotheuseoftheSLareshownonFigures53-55. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 108 www.itk.ppke.hu NONLINEAR TECHNIQUES –SYNCHRONIZATION LIKELIHOOD The synchronization likelihood has the following properties: • SLincreasesifthecouplingbetweentwosystemsincreases • SLcandetectnon-linearcouplingbetweensystems •SLcandetectachangebetweenthedynamicalsystemswithhightimeresolution • SLisfairlyrobustinthecaseofconsiderablynoisydata ApplicationfieldsofSLtoEEGandMEGdata • Epilepsy:synchronizationbetweenchannelsduringandbeforetheseizures • Synchronizationchangesduringeyeopeningandclosinginthealphaband • GammabandsynchronizationinMEGdata. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 109 www.itk.ppke.hu roland1.bmp Figure52.ValuesofOmegacomplexityandSynchronizationlikelihoodincaseofthreedifferentsignaltypes.Incaseofwhitenoise(a),thesynchronizationbetweenthechannelsisminimal,soSLhasasmallvalue,whiletheOmegacomplexitiyismaximal.Usingsinewaves(b)withthesamefrequency,thatarealsointhesamephase,meansmaximalsynchronizationwithSL=1(maximalvalue)and„Omegacomplexity”=1(minimalvalue).ApplyingthetwomeasurestoEEGdata(c),theirvalueswillbesomewherebetweenthevaluesshownbythewhitenoise(nosynchronization)andsinewaves(maximalsynchronization). ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 110 www.itk.ppke.hu SL_matrixok_gamma_b.bmp Figure53.MapsofSynchronizationlikelihoodinthegammaband(30-50Hz)betweenallofthe33electrodes,whentheeyeswereclosed(left)oropened(middle).TheSLisaveragedoverthetimedimension.ThedifferenceofthetwoSLmapsisdisplayedontherightsideofthefigure.Redareasmeanhighersynchronization,whileblueareasmeanslightornosynchronizationbetweentheappropriatechannels.TheSLbetweenthesameelectrodesarealwaysmaximal(redcoloredmaindiagonalonthefirsttwoimages),sotheirdifferenceiszero(bluecoloredmaindiagonalonthedifferenceimage).Thenumbersnexttomatricesaretheelectrodenumbers. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 111 www.itk.ppke.hu Figure54.TopographicbrainmapsconstructedfromtheSLdifferencemapsindifferentfrequencybandsreferencedtothecentralelectrodes(Fz,Cz,Pz).Wegetthesemapsbyselectingonerow(orcolumn,becausethedifferencematrixissymmetrical)fromthedifferencematrixthatcorrespondstotheactualreferenceelectrode.Thisrowcontainshowmuchthereferenceelectrodeissynchronizedtotheotherelectrodes.Redcolorsmeanhighersynchronization,whilebluecolorsindicateslightornosynchronizationonthegivenbrainregion. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY ido_syn_1sz_ny_8-11Hz_.bmp www.itk.ppke.hu 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 112 ido_syn_1sz_cs_8-11Hz_.bmp Figure55.ThechangeofSLintimeateveryrecordingsiteaccordingtoareferenceelectrodeinthealpha1band(8-11Hz)incaseofeyesclosed(bottomleftimage)oreyesopen(bottomrightimage).Theredcolorsindicatehighersynchronization,whilethebluecolorsmarkslightornosynchronizationatagiventime-point.Theupperimagesaretheaveragesoftherowsofthebottomimages:ontheleft,thechangeoftheSLaverageovertimeincaseofeyesclosedisdisplayed,whileontheleftthetimeevolutionoftheaverageSLisshown,whentheeyeswereopen. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 113 www.itk.ppke.hu NEUROMETRICS (NM) The method seveloped by Dr. E. Roy John (1924-2009) at the Brain Reserach Laboratories, New York University Medical Center. Neurometrics is a multivariate statistical method for the evaulation of EEG and ERP (event related potentials) changes. It uses standardized qualitative properties for describing the deviation from a normal database. It is a „statistical help” for diagnosing neurological and psychiatric disorders. Statistics can be applied to the analysis of pathological EEG signals. In the EEG recorded at relaxed state the amplitude and frequency-changes happen seemingly random, but actually these are lead by statistically regular and törvényszerû processes. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 114 www.itk.ppke.hu NEUROMETRICS Advantages of Neurometrics against conventional EEG: • Itgives a precise, quantitative and reproducable estimation of the deviation from the normal brain activity • Convertingthe data into a standardized dimension makes it possible to create topographical statitistical brain maps • Quantitativecharacterization of some brain disorders is also possible • Multivariatestatistical techniques can be applied • Subgroups of the patients can be determined • Thedevelopment of the disorders can be quantitatively characterized ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 115 www.itk.ppke.hu STEPS OF NEUROMETRICS • Selecting at least a 2 minute long representative segment from a 20 minute length EEG recording: The relative spectrum is calculated from a minimum 20 second long artifact-free EEG period. For frequency analysis a minimum 60 second length EEG segment is used. For coherency and measuring the asymmetry between hemispheres the whole 120 second long segment is used. • Calculating the frequency-spectrum with FFT (with a frequency-resolution of 0.2-0.4 Hz) • Calculatingthe quantitative parameters from the values of the FFT ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 116 www.itk.ppke.hu NEUROMETRICS –CALCULATING THE PARAMETERS 1. Absolute power: Calculating the power density spectrum and integrating this. 2. Relative power: The absolute power of a given interval divided by the absolute power of the whole frequency range. 3. Hemispheric asymmetry: The quotients of absolute power values of the symmetrical recording electrode-pairs (right/left) 4. Coherency: The size of the consistency of phaseshifts between the recording sites. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 117 www.itk.ppke.hu RECORDING TECHNIQUES USED IN NM Monopolarrecording sites: F1,F2, F3, F4, C3, C4, P3, P4, O1, O2, F7, F8, T3, T4, T5, T6, Fz, Cz, Pz Combined monopolar features: Left side: F7, T3, T5Right side: F8, T4, T6 Left medial: F1, F3, C3, P3, O1Right medial: F2, F4, C4, P4, O2 Left frontal area: F1, F3, F7Right frontal area: F2, F4, F8 Left central area: C3, T3Rigth central area: C4, T4 Left posterior ares: P3, O1, T5Right posterior area: P4, O2, T6 Left hemisphere: F1, F3, C3, P3, Right hemisphere: F2, F4, C4, P4, O1, F7, T3, T5 O2, F8, T4, T6 Central area: Fz, Cz, Pz Frontal area: F1, F2, F3, F4, F7, F8, Fz Medialarea: C3, C4, T3, T4, Cz Posterior area: P3, P4, O1, O2, T5, T6, Pz Whole head: all the 19 monopolar electrode recording sites ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 118 www.itk.ppke.hu RECORDING TECHNIQUES USED IN NM Bipolarrecording setup: Cz-C3, Cz-C4, T3-T5, T4-T6, O1-P3, O2-P4, T3-F7, T4-F8 Combined bipolar features: Whole head: all the 8 bipolar recording electrode-pairs Left hemisphere: Cz-C3, T3-T5, O1-P3, T3-F7 Right hemisphere: Cz-C4, T4-T6, O2-P4, T4-F8 Frontal area: T3-T5, T4-T6, T3-F7, T4-F8 Posterior area: Cz-C3, Cz-C4, O1-P3, O2-P4 ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 119 www.itk.ppke.hu CONSTRUCTING THE DATABASE • EEG parameters Gaussian distribution • Regression analysis:• Dependent variable: age • Independent variable: EEG parameters • Normal averagesAge dependent linear regression equations • Different physical dimensionStandardized (Z transformed) parameters (Figure 56.) (Figure 57.) ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 120 www.itk.ppke.hu neurometrics1.png Figure 56. Transform toGaussian distribution ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 121 www.itk.ppke.hu Z TRANSFORM qEEG –(normal) mean (estimated) deviation Properties of Z transformed values: Every parameter (of healthy people) has a zero expected value and a deviation with value 1. The database has a relative characteristics! neurometrics2.png Figure 57. Z transform of the EEG data. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 122 www.itk.ppke.hu neurometrics4.png Figure 58. Multivariate probability estimation ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 123 www.itk.ppke.hu NORMAL VALUES Processing the EEG data of numerous healthy individuals: normal database Who can not be considered as healthy? • Neurological or psychiatric disorder • Headtrauma • EEG deflection earlier • Taking medications within 3 weeks before the examination • Drinking alcohol or taking drugs • IQ not between the normal values Displaying the results: Neurometrical matrix Probability brain maps ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 124 www.itk.ppke.hu isz3 Figure59.Exampleofaneurometricmatrixfromahealthyperson.Thenumbersrepresentdeviations.Deviationsabove2orbelow-2canbeconsideredpathological. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 125 www.itk.ppke.hu Figure60.Probabilitybrainmapsofdifferentpatientgroups(relativepoweroffrequnencybands) brain_map_neurometric.png ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 126 www.itk.ppke.hu NM –DECREASING BRAIN ACTIVITY • First sign of the decreasing brain activity: Theta is increasing (at Global deterioration score (GDS) 2, subjective complaints to forgetfulness: familiar names, places of objects, but no demonstrable sign) • Next stage: Delta activity is increasing • In advanced stage: Decrease of alpha and beta activity • The correlation of GDS with coherence and the hemispheric asymmetry do not show any significant deflections. • But there is a significant deflection in the case of both the absolute and relative powers (different brain regions, cognitive weaknesses) ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 127 www.itk.ppke.hu neurometrics3.png Figure 61. Normal and patient groups in the multi-dimensional space. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 128 www.itk.ppke.hu PERSPECTIVES OF NEUROMETRICS • Optimal drug dosage • Predicator functions: The reaction of the patient to different treatments • Electrophysiological measurements (brain activity): where the EEG is inconsistent or not sensitive enough • Solving challenging diagnostic tasks (for example the discrimination of the unipolar depression from old-age dementia) • Same symptoms, with different physiological data: differentiation. ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 129 www.itk.ppke.hu SOFTWARES FOR SIGNAL ANALYSIS EEG • Compumedics –Neuroscan • Applied Neuroscience –Neuroguide • EGI –Net Station • Brain Products –Brain Vision Analyzer Spike sorting • Plexon -Offline Sorter • CED -Spike 2 • Alphaomega –Alpha-Sort • OpenEx –OpenSorter Source localization • Compumedics –Curry • EGI –GeoSource • Applied Neuroscience -Neuroguide •BESA Research • EEGLAB –Matlab Toolbox • PhiTools –PRANA Software Suite • … • Axona –Tint • Klusters • Matlab Tools: OSort, Wave_clus FIND, NeuroMAX, Mclust … • EMSE Suite • Brain Innovation –BrainVoyager QX • … ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 130 www.itk.ppke.hu RECOMMENDED LITERATURE • Signal Processing for Neuroscientists: An Introduction to the Analysis of Physiological Signals; Wim van Drongelen; 2007, Academic Press • EEG Signal Processing ; Saeid Sanei, J. A. Chambers; 2007;Wiley • Electroencephalograpy: Basic Principles, Clinical Applications, and Related Fields; E. Niedermeyer, F. Lopes da Silva;2004; Lippincott Williams & Wilkins • FunctionalNeurosciencevol2 Neurometrics: ClinicalApplicationsof QuantitativeElectrophysiology; E.R.John, John Wiley& Sons, 1977 • More than 1000 books related to signal analysis on www.amazon.com eeg_signal_proc.jpg drongelen.jpg eeg_konyv.jpg ELECTROPHYSIOLOGICAL METHODS OF THE STUDY OF THE NERVOUS-AND NUSCULAR SYSTEMHISTORY OF ELECTROPHYSIOLOGY 2011.10.07.. TÁMOP –4.1.2-08/2/A/KMR-2009-0006 131 www.itk.ppke.hu REVIEW QUESTIONS 1. What type of bioelectric signals do you know? 2. What are artefacts? Can you mention some examples? 3. What is the signal-to-noise ratio? How can we measure it? 4. What are the main filter classes? 5. Why is signal averaging useful? 6. In which domain is the correlation used? And the coherence? 7. What are the steps of the spike sporting algorithm? 8. What are the difficulties related to spike sorting? 9. Where is current source density analysis used? What is it good for? 10. What is brain mapping? 11. What is the procedure for processing event-related potentials? 12. What were the two mentioned nonlinear methods? In what application fields are they useful? 13. What is the aim of Neurometrics?