Mapping of oscillopathies - Oscillotherapeutics – Time-targeted interventions in epilepsy andbe

Appropriate animal models for each disease or disorderare requiredtofacilitatedevelopmentofoscillotherapeutics. Appro-priaterecordingtechniquesfortheoscillatingneuronalactivities ofanimalmodelsandhumansareindispensable,asareefﬁcient analyticalmethods.Thus,hereweprovideanoverviewofhowthe pathologicaloscillationsofneurologicalandpsychiatricdisorders arerecordedanddetected(the‘diagnostic’inresearchandclinics).

3.1. Animalmodels

3.1.1. Animalmodelsforepilepsy

Animalmodelsforepilepsyresearcharethoroughly summa-rizedin thebook ofPitk ¨anen etal.(2017).Brieﬂy, theepilepsy modelsareclassiﬁedbyseizuretypes(generalizedorfocal,petit malorgrandmal),animalspecies(mice,rats,catsetc.),whether invitroorinvivo,whethergeneticoracquired,whetheracuteor chronic,andhoweachseizureisevoked(electrical,chemical, sen-soryinputs,spontaneousetc.).

Chronic spontaneous seizure modelsare typically employed forthedevelopmentoftime-targetedclosed-loopinterventions.

Tottering(tg)andStargazer(stg)mousestrainsareavailableto studyabsence(petitmal)seizureswithspike-and-wavedischarges.

These strainshave knownmutation onalphaand gamma sub-unitsofvoltage-dependentcalciumchannels,respectively(Upton and Stratton, 2003).Two inbred strains are for example avail-ableforratexperiments:thegeneticabsenceepilepsyratsfrom Strasbourg(GAERS)andtheWistarAlbinoGlaxostrain(WAG/Rij) (CoenenandvanLuijtelaar,2003;Danoberetal.,1998).The spike-and-wavedischargesareseeninordinaloutbredlaboratoryrats andeveninwild-caughtratsaswell(Tayloretal.,2019).Absence seizurescaninducedacutelybysystemicadministrationofasingle pharmacologicalcompound[4,5,6,7tetrahydroxyisoxazolo(4,5,c)

pyridine3-ol(THIP),lowdosepentylenetetrazole(PTZ),or gamma-hydroxybutyrate]inrats(Farielloand Golden,1987; Marescaux et al., 1984; Snead, 1988) and chronically by prepuberty sys-temicadministrationofAY-9944ormethylazoxymethanolacetate (MAM)-AYinrats(Cortezetal.,2002;Serbanescuetal.,2004).

ThesystemicinjectionofGABA_Areceptorantagonists(e.g.PTZ, bicuculline,picrotoxin)caninduceacutegeneralized convulsion seizuresinrodents(Mackenzieetal.,2002;Veliseketal.,1992;

Velíˇskováetal.,1991).Thesystemicinjectionofglutamate recep-toragonists (e.g.kainicacidandNMDA) ormuscarinicreceptor agonists(e.g.pilocarpine)canalsoinduceacute,generalized, con-vulsiveseizuresinrodents(Ben-Arietal.,1981;MareˇsandVeliˇsek, 1992;Turskietal.,1983).Inaddition,theinhalationofﬂurothylcan beused(Prichardetal.,1969),alongwithintracranialinjectionsof bicuculline,picrotoxin,kainicacidandotherdrugstoinduceacute convulsiveseizuresinrodents(Ben-Arietal.,1980;Sierra-Paredes andSierra-Marcu ˜no,1996;Velíˇskováetal.,1991).Achronic sponta-neous,limbicseizurerodentmodelcanbepreparedbytherepeated systemicinjectionofPTZ,kainicacid,orpilocarpine(Cain,1981;

Cavalheiroetal.,1991;Hellieretal.,1998),asingle intrahippocam-palinjectionofkainicacid(Braginetal.,1999),orrepeateddaily electricalstimulationofthelimbicstructure(e.g.theAMY,HPC) (Goddardetal.,1969;McIntyreandGilby,2009).Geneticmodels forspontaneousconvulsiveseizuresareavailablebothinmouse andratstrains(e.g.weavermice,NER/Kyorats)(Serikawaetal., 2015;UptonandStratton,2003).Auditorystimulationcan induce-convulsiveseizuresinthegenericallyepilepsy-pronerats(GEPRs) andDBA/2mice(DeSarroetal.,2017).

3.1.2. AnimalmodelsforParkinson’sdisease

Animal models of PD are classiﬁed into neurotoxin models and genetic models (Gubellini and Kachidian, 2015). 6-OHDA (6-hydroxydopamine) and MPTP (1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine)aretypicallyusedtomimic selectivelossof nigrostriataldopaminergicneuronsviamechanismbywhich mito-chondrialcomplexIisblocked(Tieu,2011).Bothneurotoxinsare usedinrodentandnon-humanprimateexperiments.

SystemicallyadministeredMPTPcaneasilycrossthebloodbrain barrierwhereas 6-OHDAshouldbestereotaxicallyinjectedinto thetargetbrainstructure(usuallythesubstantianigrapars com-pacta(SNc),themedialforebrainbundle,orthestriatum).6-OHDA andMPTPadministrationsleadtosigniﬁcantPD-likemotor symp-tomsincludingakinesia, freezing, bradykinesia, musclerigidity, abnormalposture,stereotypyandtremorassociatedwith signif-icantdegenerativelossofSNcdopaminergicneurons(Smeyneand Jackson-Lewis,2005).Neither6-OHDAnorMPTPadministration induces Lewy body-like inclusions withalpha-synuclein (Cenci etal.,2002).Importantly,boththe6-OHDArodentmodelandthe MPTP-treatedmonkeysexhibit pathological oscillationsintheir basalgangliaasfrequentlyobservedin humanpatients:tremor (4–7Hz),doubletremor(10Hz),andbeta(15–30Hz)(Deffainsand Bergman,2019;Heimeretal.,2006).Mutationsofcausalgenesor geneticriskfactorsofParkinson’sdiseasearemodelledinmiceand ratsincludingSNCA(alpha-synuclein),PRKN(parkin),PINK1 (PTEN-inducedputativekinase1),DJ-1(PARK7)andLRRK2(leucine-rich repeatkinase2).Thesemodelsofferwaystostudypathologyas Lewy-bodylikeinclusions,buttheyexhibitonlymildmotor symp-toms.Pathologicaloscillationsinthebasalgangliainthesegenetic modelshavenotbeenstudiedwellyet.

3.1.3. AnimalmodelsforAlzheimer’sdisease

MostADpatientsaresporadicandtherearesomeanimalmodels forsporadicADusingmetabolicandtraumaticbraininjury-induce damageetc.(Zhanget al.,2019).However,thevastmajority of currentADanimalmodelsaretransgenicrodents(mainly mice) andarebasedontheamyloidandtauhypotheses,andthe

genet-icsofthefamilialformofthedisease(MullaneandWilliams,2019).

Nearly170transgenic/knock-in/knock-outmodelsofADhavebeen developedtodate(ALZFORUMResearchModelsDatabase;https://

www.alzforum.org/research-models).Theyareprincipallyfocused onmutationsinAPP(Amyloidprecursorprotein),PSEN1 (prese-nilin 1),MAPT (microtubule-associated protein tau), and Trem2 (Triggering receptor expressed on myeloid cells 2), and APOE (apolipoprotein E), as well as the transfection of the amyloid processingenzyme,BACE1 (Beta-Secretase1)(Götzetal.,2018;

MullaneandWilliams,2019).Themodelanimalshavesingleor multiplemutationsofthesegenes.Forexample,3×Tgmice,which haveAPPKM670671NL(Swedish),MARPTP301L,andPSEN1M146 Vtriplemutations,showamyloidbetaplaque, hyperphosphory-latedtau,andneuroﬁbrillarytangleaspathologicalphenotypesand deﬁcitsinworking,spatial,andfearconditioningmemory(Oddo etal.,2003).5×FADmicehavethreemutationsonAPP(Swedish, Florida, London) and two mutations onPSEN1, and theyshow amyloid-betaplaqueandmemorydeﬁcitsassoonastwomonths old(Oakleyetal.,2006).Inthetaupathologymodel,rTg4510mice withMAPTP301Lmutationhaveneuroﬁbrillarytangle,neuronal lossandmemorydeﬁcitsasphenotypes(Santacruzetal.,2005).

TheoverexpressionofmutanthumanAPOE4protein(ariskfactor ofAD)inAPOE4-KImiceresultsinsigniﬁcantmemoryimpairment aswell(Sullivanetal.,2004).

Importantly,all these AD models(3×Tg,5 × FAD, rTg4510, APOE4-KI)consistentlyshowareductionofslowgammaoscillation intheCA1ofHPC(Boothetal.,2016;Gillespieetal.,2016;Iaccarino etal.,2016;Mablyetal.,2017),whichcontributestotheencoding andretrievalofmemory.CA1placecellrepresentationsofspace wereunstableinthesemiceandthedeﬁcitsinslowoscillations intheHPCwereconcomitantwithspatialmemory.Surprisingly, theoptogeneticactivationofparvalbumine(PV)-interneuronsat slowgammafrequencies(or40Hzlightﬂickersensory stimula-tion)reducedamyloid-betadepositionsinthebrainandrestored cognitiveimpairmentoftheADmicemodel(Iaccarinoetal.,2016;

Martorelletal.,2019).

In the APP23 × PS45 mouse model (Busche et al., 2008), thecoherenceofslowwavesbetweendifferentcorticalregions, thethalamus, andtheHPC iscompletelydisrupted in thelight anesthesia condition (Zott et al., 2018). This resembles dis-rupted,slow-waveoscillationsduringnaturalnon-REMsleepinAD patients(Wineretal.,2019).Thecoherentslowwaveoscillations were transiently disrupted in wild-type mice by the applica-tionofsolubleamyloid-beta,whichsuggestsacausalrelationship betweenamyloid-betaandthepathologicaloscillationpatternin AD(Buscheetal.,2015).

3.1.4. Animalmodelsforschizophrenia

Animalmodelsofschizophreniamostlyﬁtintooneoffour dif-ferentinductioncategories:developmental,drug-induced,lesion orgeneticmanipulationmodels(Jonesetal.,2011).Examplesof neurodevelopmentalmodelsincludegestationalMAMinjections, bacterial or viral infections, and post-weaning social isolation;

pharmacologicalmodelsincludeamphetamine-inducedpsychosis, NMDA antagonist [phencyclidine (PCP), MK-801, ketamine)]-inducedpsychosis; lesionmodels include neonatalventralHPC lesion;geneticmodelsincludevariousknock-outormutantmodels ofschizophrenia susceptibilitygenes,someof whichwere vali-datedbygenome-wideassociationstudies(SchizophreniaWorking GroupofthePsychiatricGenomicsConsortium,2014).

Thesemodelsresemblevariouscognitivesymptomsfoundin schizophreniapatientsincludingdeﬁcitsofsensorimotorgating, workingmemory, visio-spatialmemory, andobject recognition, aswellasdecreasedsocialinteraction,increasedlocomotion,and exaggeratedsensationetc.Themodelsalsoshowcellularor cir-cuit level alterations including decreased synaptic connections,

andspinedensities, thelossofprefrontalPV-positive interneu-rons, the loss of dendrites in cortical pyramidal neurons, and pathological oscillations (e.g. dysfunctional prefrontal gamma oscillations).Thepathological oscillationstiethecellularor cir-cuitlevelpathophysiologytoalterationsinlocalprocessingand large-scalecoordination,andinturnmayleadtocognitiveand per-ceptualdisturbancesobservedinschizophrenia(Pittman-Polletta etal.,2015;SenkowskiandGallinat,2015).

A number of schizophrenia-susceptibility genes have been identiﬁed on chromosome 22. These include DISC1 (disrupted-in-schizophrenia1),NRG1(neuregulin1)anditsreceptorERBB4 (erb-b2receptortyrosinekinase4,erbB-4),andCOMT (catechol-O-methyltransferase). DISC1 is a synaptic protein that plays a crucialrole insynaptogenesis(BennettAO,2008).Mutationsor thefunctionaldisturbanceofDISC1leadtothedisruptionof PV-positive interneuron cytoarchitecture and hypofunction in the cortexandHPC,whichiscriticalfornormaloscillatoryactivityin thebrain(Koyamaetal.,2013;Nakaietal.,2014).NRG1andERBB4 arealsosynaptogenicschizophreniasusceptiblegenes(Meiand Xiong,2008).Theirdisruptionresultedinabnormalgamma oscil-lationsintheHPCanddisruptedfunctionalcouplingbetweenthe ventralHPCandNAc (Koyamaet al.,2013;Nasonet al.,2011).

Reduced dysbindin-1 (another synaptic protein from suscepti-blegeneDTNBP1(DickmanandDavis, 2009))isassociated with reducedphasicactivationofPV-positiveinterneuronsandreduced gammaoscillations(Carlsonetal.,2011).

Oneofthelargestriskfactorsforschizophreniaisthe microdele-tion of chromosome 22q11.2 that wipes out up to 60 genes;

the 22q11.2 deletion syndrome results in facial abnormalities, heartdefects,andanumberofneuropsychiatricconditions(Jonas etal.,2014).Aquarterofthepatientsthat havethe microdele-tion of this chromosome develop schizophrenia. Importantly, in the Df(16)A^+/− mouse model of this micro delision, mice exhibited reduced PFC-HPC synchrony,represented by reduced phase-lockingofPFCneuronstoHPCthetaoscillationanddisrupted coherence across multiple frequency ranges (delta to gamma ranges)(Sigurdssonetal.,2010).

Aspharmacologicalmodels,NMDAreceptorblockerssuchas ketamineandPCPareknowntoinducedelusionsand hallucina-tionsinotherwisehealthysubjects(Krystaletal.,1994).Ketamine isknowntoattenuatebothbackgroundandsensoryevokedtheta powerintheCA3inmice.Itenhancesbothbackgroundandevoked gammapower,but itdecreasesrelative-inducedgammapower (Lazarewicz etal.,2010).Thissuggeststhat ketaminedecreases thesignal-to-noiseratioofgamma-bandactivityandpossiblyleads todisruptedpatternseparationfunctionintheCA3region, con-tributingitsdissociativefeeling.KetaminereducesNMDAreceptor functionpreferentiallyonPV-positiveinterneurons,whichimpairs HPCsynchrony,spatialrepresentations,andworkingmemoryin mice(Korotkovaetal.,2010).TheNMDAhypofunctionalsoreduces deltaandthetaactivityinthecortexandHPC(Kissetal.,2013).

Ketaminealsodisruptsthethetamodulationofgamma-band activ-ityand reducesnetworkresponsibility totheenvironmentin a computermodelofHPC(Neymotinetal.,2011).

Asa gene-environmentinteractionmodel,WISKET ratswere reportedasa selectivelybred linewithschizophrenia-like phe-notypes(reducedsensorimotorgating,hyperalgesia,andmemory deﬁcit)afterpost-weaningsocialisolationandchronicketamine treatmentover 15 generations (Bükiet al., 2018).The WISKET ratsshowedincreasedtheta,alpha,andbeta-bandactivitiesand reducedgamma-bandactivitiesinECoGrecordings(Horvathetal., 2016).

3.1.5. Animalmodelsforanxietyandtrauma-relateddisorders Animalmodelsfor anxietyand trauma-relateddisorders are classiﬁedintoﬁvemodels:experience-based,pharmacologic,

phar-macologicallesion,selectivelybredgenetic,andspeciﬁctransgenic (Hoffman,2016).Examplesof experience-basedmodelsinclude fearconditioningandextinction,pre-weaningstress,and mater-naldeprivation.Pharmacologicmodelsincludeyohimbine(alpha-2 adrenergicreceptorantagonist),CCKtetrapeptide(CCK-4,an anx-iogenic neuropeptide),caffeine (adenosinereceptor antagonist), m-chlorophenylpiperazine (serotonin 5-HT_2C receptor antago-nist), and FG7142 (benzodiazepine partial inverse antagonist).

Pharmacologicallesionmodelsincludethechronicinfusionofl -allylglycine(aninhibitorofglutamicaciddecarboxylase)intothe dorsomedial/perifornicalregionofthehypothalamus(DMH/PeF)in rats(JohnsonandShekhar,2012).Selectivelybredgeneticmodels includeRomanHighandLowAvoidancerats(Escorihuelaetal., 1999), Sardinian alcohol-preferringrats (Colombo et al., 1995), HighanxietybehaviorandLowanxietybehaviorrats (Yilmazer-Hanke et al., 2004), Floripa H and L rats (Ramos et al., 2003), Ultrasonicrats(BrunelliandHofer,2007),andHighanxiety behav-iormice(Erhardtetal.,2011).Speciﬁctransgenicmodelsinclude 5-HTtransporterknockoutmice,brain-derivedneurotrophic fac-tor (BDNF) Val66Met mice, COMT and monoamine oxidase A deﬁcient mice, 5-HT_1A knockout mice, corticotropin-releasing hormone overexpression mice, and neuropeptide Y-knockout mice.

Theirendophenotypescan bemeasuredasstartlereactivity, behavioralinhibition(viatheopenﬁeldandelevatedplusmazes, as wellas the light/dark, social interaction, and punished con-ﬂicttests),carbondioxidesensitivity(avoidanceofaCO₂-enriched environment,exploratorybehaviorafterexposuretoCO2-enriched air,tidalrespiratoryvolumeduringexposuretoCO₂-enrichedair), andfearover-generalization(discriminationofCS+andCS−stimuli afterfearconditioning).Recentstudieshaverevealedthatdistinct oscillatoryactivitiesinspeciﬁcPFC-AMY-HPCnetworksarerelated tobothfear/anxietyexpressionanditsregulations(C¸alıs¸kanand Stork,2019;Dejeanetal.,2016;Karalisetal.,2016;Likhtiketal., 2014).

3.1.6. Animalmodelsfordepressivedisorders

Animalmodelsfordepressivedisordersareclassiﬁedintoﬁve models: experience-based, pharmacologic, lesion, genetic, and gene-environmentinteraction(Hoffman,2016).Experience-based models include learned helplessness, chronic adult stress (e.g.

overnight illumination, water or food restriction, tilting cages, social isolation or crowding etc.), early life stress (e.g. mater-nal separation), and social stress (e.g. chronic social defeat).

Pharmacological modelsincludewithdrawal from psychostimu-lant use. Lesion modelsinclude bilateralolfactory bulbectomy.

Genetic models include selectively bred lines (e.g. the Rousen depressedmouseline,FlindersSensitiveLinerats,WisterKyoto rats, Fawn Hooded rats, SwLo/SwHi rats, cLH rat lines) (El Yacoubiet al.,2003; Hennand Vollmayr,2005; Overstreet and Wegener,2013;Rezvanietal.,2007;Willetal.,2003),and spe-ciﬁc transgenic lines (e.g. 5-HT transporter knock-outrats and mice, BDNF promoter IV-mutant mice, BDNF Met mice) (Chen etal.,2006;Sakataetal.,2010;Wisor etal.,2003).Their cogni-tive/behavioralphenotypessuchasanhedoniacanbemeasured using thesucrose preference test, conditionalplace preference, intracranialself-stimulation,variableprogressiveratio reinforce-ment, andresponse biasprobabilistic rewordtask; theycan be measuresasnegativeprocessingbiasusingincreasedreactivityto aversivestimuli,probabilisticreversallearning,andreactivityto emotionallyambiguouscues.Somephysiologicalendophenotypes (e.g.sleeppatternchanges)arerecapitulatedaswellintheserodent models.

Recently,stressvulnerabilityanddepressionsusceptibilitywere successfullydecodedfromlarge-scaleelectrophysiological record-ingsasdistinctoscillationpatternsinfreelymovingmice(Hultman

etal.,2018).Thespeciﬁcoscillationpatternsforthevulnerability and susceptibilityare consistentwiththeresults of pharmaco-logical(interferonadministration)andearlylifestress(maternal separation).

3.1.7. Animalmodelsfordrugaddiction

Animal models for drug addiction can be classiﬁed into models for the three stages: preoccupation/anticipation, binge/intoxication,and withdrawal/negativeaffectstages (Koob andVolkow,2010).

Animalmodelsforthepreoccupation/anticipationstageﬁtinto two categories: extinction-based and abstinence-based relapse models(Venniro et al., 2016).Extinction-based relapse models includedrug-,cue-,context-,stress-,andwithdrawalstate-induced relapses (Alleweireldt et al., 2001; Shaham et al., 2003), reac-quisition(Boutonetal.,2012),andresurgence(Winterbauerand Bouton,2011).Abstinence-based relapsemodels include forced abstinenceanddrugcravingincubation(Fuchsetal.,2006),adverse consequences-imposedabstinence(Cooperetal.,2007),and volun-taryabstinenceinducedbyintroducingnon-drugrewards(Caprioli etal.,2015).Inaddition,riskyandgamblingchoicetasksandthose withreward/aversionconﬂictscanbeusedtostudythe pathologi-caloscillationsunderlyinginappropriate,impulsiveandexecutive decisionmakinginaddictedstates(Passeckeretal.,2019;Verharen etal.,2018).

Animalmodelsofthebinge/intoxicationstageconsistof intra-venousandoraldrugself-administration(AhmedandKoob,1998), intracranialself-stimulation(MarkouandKoob,1992),conditional placepreference(Sanchis-Seguraand Spanagel,2006), drug dis-crimination(Stolermanetal.,2011),andgeneticmodelsofhigh addictionsusceptibility(Quintanillaetal.,2006).Inaddition,the drugtakinginthepresenceofaversiveconsequencesmodelcanbe usedtoﬁndpathologicaloscillationsgoverningcompulsivedrug takingbehavior(Vendruscoloetal.,2012).

Animalmodelsofthewithdrawal/negativeaffectstageinclude intracranialself-stimulation(rewarddecreases),conditionalplace aversion(Handetal.,1988),measuresofanxiety-likeresponses (e.g. the open ﬁeld and elevated plus mazes), and drug self-administration with extended access or in dependent animals (Ahmedetal.,2000).

3.2. Neuralactivityrecordings 3.2.1. Foranimalresearch

Large-scalebraindynamicsrecordingsasLFPsarevery pow-erfulfor investigatingoscillatoryactivitiesacrossmultiplebrain regions(HongandLieber,2019;Pesaranetal.,2018).Beyond sin-glesiterecordings,multi-siterecordingswithsiliconprobeshave allowedthegeometryofoscillatoryactivitiesinthebraintobe studied(WiseandNajaﬁ,1991).Linear16–32chrecordingprobes havebeenusedtomaplayerspeciﬁcoscillationsforexamplein thecortex(Minlebaev etal., 2011).Recent CMOS-basedprobes (Neuropixel)enableupto960chhigh-densityrecordingsona sin-gleshank(Junetal.,2017).Multi-shanklinearsiliconprobes(e.g.

buz256)cancapturetwo-dimensionalspatiotemporalstructureof oscillationsinabrainregion(Agarwaletal.,2014;Berényietal., 2014).Forexample,Olivaetal.foundthatsharp-waverippleinthe CA2subregionprecedesthoseintheCA1andCA3subregionsinthe ratHPC(Olivaetal.,2016).

Theinsertionofmultiplesiliconprobesand/orwireelectrodes intodistinctbrainregionsallowedoscillatoryinteractionsbetween brainregionstobeexploredduringspatialnavigation (Fernández-Ruizetal.,2017),goal-directedbehaviors(FujisawaandBuzsáki, 2011),epilepticseizures(Berényietal.,2012),anxiogenic condi-tions(Girardeauetal.,2017),depression(Hultmanetal.,2018), anddrugaddiction(Sjulsonetal.,2018).Matrixsiliconprobescan

beusedtoobtainhigh-densitythree-dimensionaloscillation activ-itiesinthebrain(Riosetal.,2016),andﬂexiblemeshelectronics usedinsteadof rigidelectrodesenableyear-long stable record-ings(Hongetal.,2018).Two-dimensionalelectrodesonﬂexible polymersheetsenablepotentialrecordingsfromthecortical sur-face(Khodagholy etal., 2015).Simultaneous recording ofbrain andotherphysiologicaloscillations(e.g.electrocardiogram, elec-tromyogram, and breathing) from freely-moving animals is an importanttechniquetostudypathophysiologicalrepresentations ofneuropsychiatricdisorders(Sasakietal.,2017).

3.2.2. Forclinicalpractice

Theinternational10–20or10–10EEGrecordingsystemsare widelyusedforstandarddiagnosisorstudyofavarietyof neu-ropsychiatricdisorders, includingepilepsy(Nuweret al.,1998).

OneoftheadvantagesofEEGrecordingsisitstimeresolution.This enablesfastoscillatoryactivitiestobeanalyzed(typically0.3–300 Hz).MEGrecordingshaveanevenhighertimeresolution(in mil-liseconds).Thefrequencyspectrumdensityineachrecordingsite andtherelationshipsbetweentherecordingsites(coherency, con-nectivity,causalityetc.)aretypicallyanalyzed.High-densityEEG recordings(64–256ch)increasespatialresolutionandallowsource imagingwithevensub-lobarprecision(Seeck etal.,2017).This enablesbetterspatialresolutionforseizurefocuspredictionwith tomography.fMRIrecordingsgivehigherspatialresolution(in mil-limeters)butlowertimeresolution(inseconds)comparedtoEEG recordings;theyprimarilyutilizetheblood-oxygen-level depen-dentcontrast,whichis complementarytoEEGrecordings.fMRI recordingscanbeusedtoinvestigateveryslowoscillatoryactivities withinandbetweenbrain regions.Invasiveelectrophysiological recordingsonorinthebrainarerequiredtoﬁndtheseizurefocus muchmorepreciselyortheoptimallocationofDBSelectrodesin thebasalgangliaofPDpatients.

3.3. Machinelearning-mediatedapproachesforanalysis

Itischallengingtoﬁnddiseaseordisorder-speciﬁcoscillation patternsinlarge-scaleneuronalactivitydata.Forexample, unsu-pervisedlearning techniqueshavebeenused toﬁndsigniﬁcant coherentresting-stateﬂuctuationsandfunctionalconnectivityof resting-statefMRIdata(Khoslaetal.,2019).Unsupervisedmethods likeindependentcomponentanalysis(ICA)andprincipal compo-nent analysis (PCA)decomposition are also used toﬁnd latent variable models in fMRI data. Deep learning methods such as

In document Oscillotherapeutics – Time-targeted interventions in epilepsy andbeyond Neuroscience Research (Pldal 5-9)