ArticlePDF Available

Abstract and Figures

In the next decade, high performance computing is projected to reach exascale capacity (10 18 floating­point operations per second), a region deemed suitable for emulating one or more continuous­learning spiking neuron models of the human brain in real­time [Markram 2012]. Alternatively, specialized hardware may make an exascale computer system unnecessary. Our human brain contains 86 billion neurons, nearly as many glial cells, 100 trillion synapses and over 100 different chemical neurotransmitters [Herculano­Houzel 2009]. Adequate preservation of the human brain after death such that the potential information represented by all of those parts is not lost is difficult, though progress is being made on several fronts. Capturing the full human brain connectome, with all relevant connectome data of a given person, appears to be very difficult and may remain one of the greatest technological hurdles in the coming years. In 2012, Kenneth J. Hayworth emphasized a specific technique for SEM scanning of brain slices for whole brain emulation [Hayworth 2012]. Hayworth's paper is recommended as a complement to the present paper. Our survey has a more general approach to method and a broader scope. The earlier report on whole brain emulation by Sandberg and Bostrum [Sandberg 2008] recognized the potential importance of body and environment simulations.
Content may be subject to copyright.
WholeBrainEmulationStateoftheArt
©2015GaryFeierbachandRandalA.Koene
Abstract
Inthenextdecade,highperformancecomputingisprojectedtoreachexascalecapacity(10
18
floatingpointoperationspersecond),aregiondeemedsuitableforemulatingoneormore
continuouslearningspikingneuronmodelsofthehumanbraininrealtime[Markram2012].
Alternatively,specializedhardwaremaymakeanexascalecomputersystemunnecessary.Ourhuman
braincontains86billionneurons,nearlyasmanyglialcells,100trillionsynapsesandover100
differentchemicalneurotransmitters[HerculanoHouzel2009].Adequatepreservationofthehuman
brainafterdeathsuchthatthepotentialinformationrepresentedbyallofthosepartsisnotlostis
difficult,thoughprogressisbeingmadeonseveralfronts.Capturingthefullhumanbrainconnectome,
withallrelevantconnectomedataofagivenperson,appearstobeverydifficultandmayremainoneof
thegreatesttechnologicalhurdlesinthecomingyears.In2012,KennethJ.Hayworthemphasizeda
specifictechniqueforSEMscanningofbrainslicesforwholebrainemulation[Hayworth2012].
Hayworth'spaperisrecommendedasacomplementtothepresentpaper.Oursurveyhasamore
generalapproachtomethodandabroaderscope.Theearlierreportonwholebrainemulationby
SandbergandBostrum[Sandberg2008]recognizedthepotentialimportanceofbodyandenvironment
simulations.
Introduction
Recently,therehasbeenaflurryofnewsconcerningprogressinartificialintelligence(AI).Wesee
possiblesignsofintelligenceinSiri(Apple),Watson(IBM)andtheselfdrivingcar(Google).Atthe
sametime,newtoolsarebeingdevelopedtounderstandbiologicalneuronalnetworks.Here,we
examineprogresstowardsthetwogoalsof(1)acquiringthedatathatrepresentstheneuronalnetwork
ofaspecifichumanbrainand(2)recreatingthatneuronalnetworkinoperationalform,forexampleby
acomputeremulationwithmemories,emotionsandpersonalityintact.Effortsareunderwayinthe
UnitedStatesandintheEuropeanUniontotacklethehumanbrainatthelevelofphysiologyand
neuronalcircuits[EFPL2014,NIH2013,NIH2014].Thebyproductsoftheseefforts,braintissue
preservationandpreparation,neuronandglialcellbehaviorandneuralnetworkunderstandingand
modeling,havedirectbearingonwholebrainemulation.
Thefundamentalsforwholebrainemulationthroughsystemidentificationinneuronaltissuewere
laidoutinpriorwork[Koene2012a],andaroadmapoftechnologicalrequirementswaspostulated
[Koene2012b,Deca2014].Thepresentpaperisconstrainedtoareviewofthestateoftheart,andwe
refertothosepriorpublicationsfordetailsaboutthestructuredapproachforwhichthetechnologies
reviewedbelowareneeded.
Technologyfortissuepreservation,preparation,circuittracingandemulationatthescaleofa
humanbrainarealsodiscussed.Asuccessfulemulationwillneedtobeembodiedwithinputandoutput
ineitheravirtualorphysicalmanner.
Toconstrainthisconcretestudy,wedonotaddressmatterssuchasunderstandinghow
consciousnessworks,definitionsorintuitivenotionsofpersonalidentity,dualistphilosophies(e.g.
'souls'),legalconsiderations,moralconsiderationsoranyothermetaphysicalpoints.Thesemaybe
importantissues,butarewellbeyondthescopeofthistechnologicalsurvey.
InvivoBrainScanning
A.MagneticResonanceImaging(MRI)
Theidealistoobtainahighresolutionscanofthecompleteconnectomeinvivoandwithoutharm.
Toadegree,theHumanConnectomeProject[HCP2012]takesthatapproach,usingahighlymodified
3and7TeslaMRIwithcustomizedheadcoils[Ugurbil2013].Thisyieldsa1.25mmvoxelresolution
andcanidentifyonlylargebundlesofnervefiber(whitematter).Acube1.25mmoneachsidecan
contain170,000neuronsandmillionsofsynapses.MRIresolutionwouldhavetoincreasebyfiveor
moreordersofmagnitudetoseeallofthesedetails(leavingasideotherconstraintsaboutthethings
MRIcanorcannotdetect).Thisresolutionhasbeenusefultolearnabouthowvariousbrainsubsystems
areconnectedandhashelpedresearchersidentifyfunctionalunitsandtosomeextentthepurposeof
thoseunits.9.4TMRImachinesarenowavailableandawholebody11.7Tmachineisexpected
[Wada2010].The11.7Tmachinesareexpectedtoresolve100μmvoxels. 
ResolutionofcurrentMRIimaginghasbeenimprovedbyalmostanorderofmagnitudeby
applyingdiffusiontechniques[Hagmann2007,CohenAdad2012]andfurtherimprovementmaybe
possiblewithbetterMRIdetectionandprocessingalgorithms.AheadonlyMRImachinemight
achievehigherresolutionat25T.MRImachinesfornonhumanuseareavailableupto17T[Wada
2010].InvivoMRIscanshavetodealwithseveraldifficulties[Passingham2013].(1)Movement:
Clampingapersonsheadmayrestrictoverallheadmotiontoamillimeterorlessbutrespirationand
bloodpressurechangescanalsomoveneuronssignificantfractionsofamillimeter,makingitdifficult
toincreaseresolution.(2)Theneedformorepowerfulmagnets:Toreachtherequiredresolution(~40
nm)[Marblestone2013],themagnetswouldrequiretechnologybeyondanythingwecanconceiveof
today.Theeffectsofultrastrongmagneticfieldsonhumanneuronalsystemsareunknown.Magnets
usedincurrentMRIsystemsareneverturnedoffandcaninducesmallcurrentsinneuralcircuitsasa
personisloadedintothemachine.PeoplehavereportedvertigoinandaroundMRImachines,
indicatingpossibleinterferencewiththevestibularsystem.Strongermagnetswillinducestronger
currentsduringmovementwithinthefield,leadingtoincreasedsafetyconcerns.Thehumanbody
consistsof65%water,adiamagnetic.Water’sdiamagneticpropertiesallowpowerfulmagnetsto
levitatesmallmammals.
B.MagneticResonanceSpectroscopy(MRS)
MRImachinesat3TandabovecanapplyMagneticResonanceSpectroscopy(MRS)toidentify
neuraltransmitterssuchasgammaaminobutyricacid(GABA),Nacetylaspartatylglutamate(NAAG),
Acetylcholine(Acl),myoinositol,andglutamate/glutamine(Glu)fromsmallmoleculenuclear
resonance[Foerster2013].AsinMRI,thesetechniquesaremanyordersofmagnituderemovedfrom
theresolutionthatisneededfordetailedneuronalcircuittracing,butareusefulfortracinglong
myelinatednervebundlesfromonepartofthebraintoanother.
C.UltrasoundImaging
XiangZhangatLawrenceBerkeleyLabisexperimentingwith10GHzultrasoundimaging,
currentlyusedon2Dmaterialsat1,000timestheresolutionoftypicalultrasoundimaging.Ultrasound
maybecomecompetitivewithMRIforfineneuralstructure.
D.CarbonNanotubeDetectors
JunichiroKonoofRiceandFrançoisLéonardofSandiaNationalLaboratoriesdemonstrate
promisingresultsusingcarbonnanotubeterahertzdetectorsthatmayreplaceMRIformedicalimaging
athigherresolution[He2014].Thetechniqueisalsoaffectedbysmallheadmotionsandchangesin
bloodpressure.Theseeffectsmaybemitigatedbyanesthetizingthesubjectandscanningbetween
carotidpulses.
Someproposalsenvisionusinggeneticallyengineered/modifiedvirusesthatcouldfollowneural
pathwaysandrecordthepathbyextendingDNAwitha‘barcode’,usingsetsofbasepairsforthebars,
tobedecodeduponcollection[Zador2014,Kramer2013,Callaway2008].Atechnologycurrently
usedonsomesmallanimalsistheuseoflightemittingtracers,florescentlabelsandquantumdots
addedtotagneurotransmittersandstructures.Inthiscase,boneandintermediatetissuewithpoorlight
transmittancehastoberemoved.Othertechnologies,suchasfunctionalmagneticresonanceimaging
(fMRI),diffuseopticaltomography(DOT),magnetoencephalography(MEG),singlephotonemission
computerizedtomography(SPECT)andpositronemissiontomography(PET)canbeusedtocorrelate
regionsofthehumanbrainwithparticularmentaltasks,butataresolutionseveralordersofmagnitude
belowthatrequiredfordetailedneuronalcircuittracing.Anewtechnique,usingahighlymodified
longtippedhighspeedatomicforcemicroscope(LTHSAFM),allowsonetowatchthemorphology
ofliveneuralcellschangeundervariousstimuli[Shibata2015].
PostmortemBrainPreservation
Theproblemwithwaitingfordeath,beforecapturingthatperson’sconnectome,isirreversible
braindamagemayhaveoccurredbefore(e.g.Alzheimer'sdisease,stroke)andafterdeath(e.g.
ischemicdamage),orinthebrainpreservationprocess(e.g.icecrystalformation).Ifbraindamagedue
toconditionssuchasAlzheimer'sdiseaseorstrokeoccursthenthereisagreatlyreducedpossibilityof
recreatingthatperson’spriorbrainconnectivity.
Thecurrentpreservationmethodofreplacingbloodwithanagentthatdoesnotformicecrystals
andpreservesbrainconformationhasbeenaccomplishedwithsmallbrains,butitstillfallsshortfor
brainsthesizeofahumanbrain.TheBrainPreservationFoundation(www.brainpreservation.org)has
offereda$100,000challengeforareliablebrainpreservationtechnique.Therearetwoserious
challengersplustwonewmethods,onedevelopedatStanfordUniversitythatreplacesthelipidswitha
hydrogelthathasmetwithsuccessonratbrainstheotherdevelopedatMIT[Chen2015]using
expansionmicroscopy(ExM).
Thefirsttechnique,usedattheMaxPlanckInstitutetopreservedawholemousebrainused
chemicalpreservationandplasticembedding[Mikula2012].Thesecondtechnique,usedby21
st
CenturyMedicine,usescryopreservationfreezingandreplacingbloodandwaterbasedfluidswithan
antifreezetypefluidcalledM22thatdoesn'texpandandformsharpcrystalsonfreezing[Fahy12004,
Fahy22004].ThethirdisStanfordUniversitydevelopedhydrogelmethodcalledCLARITYthat
createsatransparentblockofbraintissuewithlipidmembranesremovedthatiseasiertocircuittrace
[Cheng2013].Theforthandnewestadditionisexpansionmicroscopy[Chen2015]thatenlargesthe
specimenbysynthesizingaswellablepolymernetworkwithinthespecimengivinga4.5linear
expansionin3dimensionsproviding~70nmlateraland~200nmaxialresolutionusingaconfocal
spinningdiskmicroscope.
Therearedifferentgoalsforpreservinginternalorgans.Preservationdowntothemolecularlevelis
necessaryfororgansthataretobetransplanted.Generallycryopreservationhasbeenthemethod
chosenforreanimationofbiologictissueandsomesuccesseswithsmallanimalsandorgansincluding
fishandflyembryos.MorerecentlyFahy[Fahy22004]hashadsuccesswithlowtemperature
vitrificationfollowedbycontrolledthawingandtransplantingofarabbit'skidney.
Inthelastdecade,considerableprogresshasbeenmadeinpreservingthehumanbrainafterdeath,
butthehopeofbringingabrainbacktolifelooksremote.Currently,perfusingthehumanbrainwith
cryopreservationagentsdoessignificantdamageatthemolecularlevel.However,thismaynotmatter
forbrainconnectometracinganddownloading.ForthepurposesofSEM,Clarityorexpansion
microscopyaldehydestabilizedcryopreservation,accordingtoMcIntyreandFahy[McIntyre2015],
enablesindefinitestorageofwholebrainswhilepreservingtheultrastructurewithminimaldistortion
andisscalabletoanysizedbrain.Ifneuralwallstructuresareintactsothatdendrites,neuralcelltypes,
somaconfigurationandaxonscanbetracedalongwiththesynapticjunctionsthenonecanreconstruct
anetwork.Additionally,ifassociatedinformationsuchasneurotransmittertypeandsynapticstrength
areidentified,thenitmaybepossibletoestablishthekeyinformationthatcouldbeusedtoreconstitute
thatperson'smemoryandpersonalitywhenneuroscienceisabletoreliablymapsuchdatatofunctional
representations.
UnderstandingBrainPhysiology
Itmaynotbenecessarytounderstandhowthewholebrainorevensubsectionsofthebrainworkin
ordertoemulateahumanbrain.However,itwillbenecessarytounderstandhowthebasicbuilding
blocks,thevariousneurontypesandmorphologies,workandhowtofunctionallysimulatethem.This
involvesidentifyingandunderstandingneurotransmitters,synapticjunctions,neuronsandtheir
connectivity,theinfluenceofglialcells,etc.Significantprogresshasbeenmadeintheseareasbutnew
detailsandcomplexitiesarebeingrevealedcontinuously.
StartingwithLapicque'sspikingmodelofaneuronin1907[Abbott1999],progresswashampered
bytheinabilitytoprobeneuralnetsofanycomplexitytoascertaintheiractualfunction.Thelasttwenty
yearshavebroughtaboutaremarkablechangeinourabilitytoprobeneurons,circuittraceneural
circuits,understandthefunctioningofalargearrayofneurotransmittersandlightuptheinner
workingsoftheinnermachineryoftheneuron.Methodsforprobingneuralactivityhaveimproved
fromglasspipettestoParyleneCcoatedselfentangledcarbonnanotubeswithtipstrimmedusinga
Ga+focusedionbeam[Yoon2013]forintraandextracellularrecording.Arraysofimplanted
electrodes(upto127)usedforsimultaneousrecordingofupto14brainareasinthemacaquemonkey
havebeendeveloped[Feingold2012]forpermanentimplant.CMOSsensorarraysforextracellular
recordingcannowrecord16Kpixelsataframerateof2kiloframespersecondwithapixelpitchof
7.8μmx7.8μm[Eversmann2003].Thishasbeenusedforimagingactivityinabrainslice.Membrane
currentsgeneratedbyneuronscanbemeasuredinintracellularspace[Nadasdy1998]andthesource
neuronscangenerallybeascertainedwithsomesignalprocessingifanarrayofsensorsareused.
RomainBrette[Brette2009]bringsthespikebasedmodelnearlyuptodatewithcomputer
simulationsandnowwithimprovedtoolssuchasinvitroandinvivoexperiments,multichannel
monitoring,andcomputersimulationonecanshowthatinvivoneuronsappeartoexhibitashorter
timeconstant(afewmillisecondsorless)thaninvitroneurons(afewtensofmilliseconds)anda
neuron’soutputiscorrelatedtothetimingoftheinputpulses.Spikingneuronsaredifferentthanthe
perceptronstylelevelinput(thelevelthoughtofasafunctionofspikerepetitionrate)artificialneuron,
variantsofwhichhavebeenusedinmanyAIneuralnetworks.Computingtheforwardpropagationofa
perceptronnetworkismorethananorderofmagnitudesimplersinceonesetofcomputationscan
representaseriesofspikecomputations.Furthermore,spikingnetworksaresensitivetospike
correlationsontheirinputsindicatingthattheremaynotbeasimpletransformationbetweenspiking
modelsandnonspikingmodels.Ifonewantstoemulatehumanneuralnetworksanduseparameters
derivedfromhumanneuralnetworksthen,itseems,itissafesttousespikingmodelstoachievethe
sameresult.Althoughnotallpropertiesofbiologicalneuronsareimportantforneuralcomputation,
clearlymanyare[Ostojic2009],andtheremaybemanysubtletiesthatcontributetoreplicatinghuman
behavior. 
Weareatatimeofacceleratingdiscoveryandactivityinneuroscience,andcouldsaythatweare
onatrajectorythatresemblesMoore'sLaworBell'sLawphenomena.Recentdevelopmentsin
nanoscaleanalysistools[Alivisatos2013]andinthedesignandsynthesisofnanomaterialshave
generatedoptical,electrical,andchemicalmethodsthatcanreadilybeadaptedforuseinneuroscience.
TheultimateintentoftheBrainActivityMapping(BAM)[Alivisatos2013]projectistobeableto
monitortheactivityofeveryneuroninamousebrainsimultaneouslyat1mssampleresolution.Thisis
currentlybeingtackledwithimproved2Delectrodearrayswithoutcellpenetration,buttheneedfora
“spikesorting”algorithmthatcanhandlerecordingfromalargefractionofthecellsinapieceof
neuraltissueisstillaproblem.New3Dmicrofabricatedelectrodearraysmayovercomesomeofthe
problemsof2Darrays.Thesearrayscanstimulate,aswellasrecordfromneurons.Thisresearch
coupledwithoptical[Micheva2010]andothertechniques,includingneuralnetsimulations,isleading
toanunderstandingofindividualneuronbehavioratthemolecularlevel.Inaddition,thisshouldlead
toanunderstandingofvariousneuralsubsystems.Thisworkisthenverifiedbyneuralsimulation
[Dethier2011]thatshould,ultimately,manifestthesamebehaviorifthenetworkistracedcorrectly
andtheneuronsaremodeledandinitializedcorrectly.
Notallthisresearchisrelevanttowholebrainemulationbutit'stooearlytobesurewhatisand
whatisn't.Currentactiveresearchindicatesthatglia[Fields2013],astrocytesinparticular,
communicateandperformanactivebutnotthoroughlyunderstoodroleinnetworkactivationand
plasticity.
CapturingtheFullConnectomeofaBrain
Thereareseveralmeansforpreparingbraintissueforelectronmicroscopesandsuperresolution
lightmicroscopes.Toidentifythoseparametersnecessaryforconnectomereconstructionand
behavioralsimulationitcurrentlyappearsthat40nmvoxelsarenecessary.Bothelectronmicroscopes
andsuperresolutionlightmicroscopesarecapableof40nmorlessvoxelresolution[Denk2012,
Briggman2012]andbothmaybenecessarytoidentifyeverythingthatisnecessaryfornotonlytracing
theconnectomebutcountingvesicles,identifyingtypesofneurotransmitters,enzymes,anddistribution
ofionchannels.Thetechnologiesinvolvedincludeserialsectionelectronmicroscopy,serialblockface
electronmicroscopyandsuperresolutionlightmicroscopy.
Serialsectionelectronmicroscopyhastwomainsubdivisions,transmissionelectronmicroscopy
(TEM)withsamplespreparedonTEMgrids,andsectionscollectedonsolidsupportforserialelectron
microscope(SEM)imaging.TEMadvantagesinclude:tissuesmaybestainedaftercutting,sectionscan
bereimagedandparallelimagingispossible.TEMhasthehighestlateralresolutionandthefastest
imagingrate.SEMcanuseautomatedsectioncollectionandlargesectionareaswithalowerlikelihood
ofsectionloss.Bothmethodsarecurrentlysubjecttosectionfolding,distortionandloss.Heavymetal
stainingisusedtooutlinemembranousstructures.
Serialblockfaceelectronmicroscopy(SBFSEM)hastwomaingroups,diamondknifecutting
SEMandfocusedionbeamablationFIBSEM[Knott2008,Marblestone2013].Theadvantagesareno
lossofsections,littledistortionandautomatedcuttingandimageacquisition.Thedisadvantageis
sectionsaredestroyedduringcuttingorablation.TheSBFSEMcurrentlyhasaproblemofsometimes
leavingresidueonthenewlycleavedsurface.TheFIBSEMhasthedisadvantageofbeinglimitedto
rectangularblocks12μmononesideduetothedepthoffocusoftheionablationbeam.
Superresolutionlightmicroscopesalsofallintotwogeneralcategories,stimulatedemission
depletionnanoscopy(STED)[Persson2011]andphotoactivationlocalizationmicroscopy(STORMor
PALM)[Huang2008].Also,twophotonexcitationmicroscopy(TPEM)[So2002]isofinterestsince
thesamplesufferslessphotodamagealthoughtheresolutionisn'tasgood.AdvantagesofSTEDand
STORMincludecolortodistinguishdifferentfluorophorestaggingfeaturesofinterestandtheability
togetresolutionsdownto2030nmlateralresolutionalthoughtheaxialresolutionis5060nm.
Disadvantagesincludethedifficultyofgettingasufficientlyhighfluorophoreconcentrationina
section,thesectionthicknessandmultipleimagesaregenerallyrequired.Theseopticalmethodsrival
electronmicroscopyandinsomecasescanbeusedonlivebiologicalsystems.
Automatedtapecollectedlatheultramicrotomy(ATLUM)[Hayworth2012]slicesablockoftissue
intosectionsdownto25nmthick.Largelateralsectionscanbeachievedof2.5x6mm.Thetissue
blockcanbepretreatedwithfluorophoreandmetallicstainingenablinguseofbothopticalandSEM
imaging.Zeiss[ZEISS2013]isdevelopingmultibeamSEMs(mSEM)andwillhavea61beamunit
commerciallyavailablein2015.Thiswillgivea10xprocessingspeedimprovement.Usingassembly
lineprocessingATLUM>STORM>mSEM>imageprocessingwithmultipleassemblylinesit
appearsconceivablethatahumanbraincanbeprocessedinthismannerinareasonabletimeframeof
perhaps3yearswith100processinglinesandexascaledataprocessingandstoragefacilities.This
assumesa50x50x50nmvoxelandonegigavoxelpersecondthroughputperlinecreatingadatabaseof
10petavoxels.
Hayworth[Hayworth2012]hasevaluatedseveralcurrenttechnologieswiththeviewtoward
automatingvariousformsoftissuepreparationanduseofscanningelectronmicroscopes.Hesuggested
thattheplasticembeddedbrainbesubdividedinto3x180mmstrip20μmthick.Thisprocedurewould
create250,000tissuestrips.Thesearethenprocessedinto20x20μm,180mmlongtissuepillars.so
thattheycanbeparceledouttomultipleFIBSEMmachines.Inthiscase,losttissuebetweencylinders,
layeralignmentbetweencylindersanddistortionduetothecuttingprocessmaybeproblems.In
additiontheFIBprocessdestroystissuesoitisoneshot.Itisnotknownwhetherthiscanbecombined
withopticaltechniquesforidentifyingotherkeyelementsoftheconnectomesothatitcanbesimulated
withbehaviorintact.
Kreshuketal.[Kreshuk2013]usedatransmissionelectronmicroscopecameraarray(TEMCA)to
getfasterthroughputwhileusing40x40x40nmvoxels.Theyachievedreasonableresultsonadesktop
(12core3.2GHzXeonES1630CPU)usingrandomforestpatternrecognitionin3Dforsynapse
identification,butitstillfellshortofexperthumanidentificationofconnectomeinformation.
Automatedserialtwophoton(STP)tomography[Ragan2012]hasbeenusedsuccessfullyonwhole
mousebrains.ItisablockfacesetuponanXYZstage,laserscanningandwithabuiltin
vibratingblademicrotome.ThestageelevatesintheZdirectionin50μmstepsforblockfaceslicing
andapiezoelectricallycontrolledobjectiveallowsvoxelsZdimensionsassmallas2.5μm.Awhole
mousebraincanbeimagedbymovingthestageintheXandYdirectiongetting260overlapping
coronalsections.At0.5x0.5μmXYresolutionand2.5μminZ,thewholemousebraincanbe
automaticallyimagedwithin7days.Thiscanbecomplicatedwhenusingmultiplefluorophores.
Further,thedesiredresolution,40nmforcompleteconnectomeinformationanddeterminingsynaptic
strength(e.g.imagingsynapticvesicles)forsimulation,maynotbeattainable.
Anotherwaytocapturetheconnectomeisusingthepseudorabiesvirus(PRV)whichisknownto
travelneuralpathways[Callaway2008,Kramer2013].Othervirusesarealsoknowntospreadinthis
fashionbuttheBarthstrainofPRVspreadsslowerwithreducedcytopathiceffectsanddoesnotaffect
humansorotherprimates.EvidencesuggeststhatPRVmaybeexclusivelytranssynapticmakingit
idealforthispurpose.
Largescaleexperimentsneedtobeconductedtodeterminewhichofthesetechniquesarecapable
ofhighlyparalleloperation,highreliability,automatedimaging/processing,canacquiresynaptictype
(neurotransmitters)andstrengthinformation(synapticvesiclesandreceptordistributions),andcan
makenewconnections(dendriticspines[Bonhoeffer2002]).
PatternRecognitionandCapturingtheConnectome+
Accuratesimulationofthehumanconnectomeinvolvednotjustthemorphologicalneuralmapof
our85billionneuronsbutshouldincludealltheinformationrelatedtotheirtypeandlocality,their
associatedglialcells,synaptictypesandstrengthsandwhateverotherinformationmaybefound
necessaryinthefutureforafullbehavioralsimulation.Thiswillbecalledtheconnectome+.
Currentlymuchconnectomepatternrecognitionattheneuralleveliseitherperformedbycomputer
aidedhumans,byusingpatternrecognitionsoftwareorlabelingneuronswithfluorescentproteins
(Brainbow)[Livet2007].Acrowdsourcedhumanrecognitionofneuralmorphologyusingatoolcalled
EyeWire[Eyewire2014]iscurrentlybeingusedtocreateatrainingdatabaseforwhatwillbecome
fullyautomatedsoftware[Turaga2014,Lang2011].Evenifhumanscouldrecognizeaneuronora
synapticjunctioneveryseconditwouldtakemorethan10billionpersonhourstocompileahuman
connectomeinthismanner.Clearly,thisprocessmustbecomeentirelyautomatedandextremelyfast.
Earliereffortsatneuralcircuittracingfromconfocalimagesusedthefilamenttracersoftwareof
Imaris(Bitplane,Zurich)[Schmitt2004,Evers2005]totracethecenterlineofneuron'snetcomplete
soma,axonanddendrites.AccuracywasimprovedbySchmittetal.byfittingcenterlinesandradiito
thecorrespondingimageofeachsection,constructingaskeletonoftheneuron.Thesurfacewas
createdusingcylindricalsectionsusingtheradiiassociatedwitheachsection.Theprocedurestill
requiredmanualinterventiontocorrecterrorsinbuildingtheskeleton.
Currentpatternrecognitionsoftwareusesanumberoftechniquesthathaverecentlyemergedas
powerfulpatternrecognitionprocedures.Theseincludeconvolutionnets(CN)[Turaga2014],random
forest(RF)[Andres2014],Markovrandomfields(MRF),NeuralNetworks(NN)andanumberof
otherstatisticalandadhoctechniques.Thefirsttaskistopartitionasectionbyconnectingrelated
pixels.Thisiscalledanaffinitygraph.BothTuragaetal.usingCNsandAndres,etal.usingRF
achievedsimilarresultscomparabletohumansandasizableimprovementovertheothertechniques,
butthatstilllefteighttotenpercentofthevoxelsindisagreementwiththehumananalysisofthetest
set.Thisresultisbetterthanonemightexpectbecauseitisjusttheclassificationofindividualvoxels
andnottheoverallpicture.Iftheseareedgevoxelsbetweentwodifferentneurons,thiscouldleadtoa
mergingofthetwoneuronsorapartitioningofonegivingafalseresult.Weareledtotheconclusion
thatmorethanlocalknowledgemightbenecessarytoarriveatanaccurateresult.Forexample,finding
thatanetworkhasmorethanonecellnucleuswouldindicatethatthesenetworkswereincorrectly
combined.Voxelsthatwereweaklyclassifiedasnonboundaryvoxelsthatcouldseparatethenetworks
wouldthenbesuspect.EvenifoneascertainsthefullhumanCNSwithnear100percentaccuracythere
isstilltheproblemoftheadditionalinformationtomaketheemulationofahumanneuralnetfunction
suchassynapsetypesandconnectionstrength.
SEMsarepowerfultoolsbuttheyprovideverylimitedinformationineachvoxel,sevenbitsof
intensitylevel.Oneoftenneedopticalmethodsusingfluorescenttagstoprovideinformationon
neurotransmitters,vesicles,neuraltypes,etc.thatcanbecorrelatedwithEMsections.Tapedsections
thatcanbefeedthroughbothatwophotonmicroscopeandanSEMappearstobetheonlywaytodo
this.Thisdoesn'tnecessarilyruleouttheuseofFIBSEMsandSBFSEMsiftheelectronbeamand
photonbeamcansharethesamemountedsample.Theproblemisboththeelectrongunandthe
microscopeobjectivelenshavetobequiteclosetothesection.Apossiblesolutionistorotatethe
sampleholderbetweentwostops.Gravitationaldistortionshouldbeminimalforcylindricalsamplesof
afewmillimeterssquare.
Opticalmethodscanbeusedtotraceneuralfibersinmyelinsheathsusingpolarizedlightimaging
(PLI)[Palm2010].Thepolarizedlight,duetothebirefringenceofthemyelinsheathsofthenerve
fibersrevealsathirddimensionforplacingthefibersina3Dspace.ChristophPalm,etal.wasableto
usethistotracenervefibers1μmto20μmindiameterbundles.
Anoveltechniqueistophysicallyexpandthesampleasmuchas4.5foldlinearexpansionusinga
polyelectrolytegel[Chen2015]todosuperresolutionmicroscopybyincreasingthespeedofvoxel
captureusingdiffractionlimitedmicroscopesinstead.Onecanalsoincreasethepossibleresolutionof
superresolutionmicroscopyusingthismethod. 
Thereisthepossibilitythatoneoracombinationoftechniqueswillbeabletorisetotheoccasion
sincemostofthesuperresolutionopticaltechniquesareveryrecentdevelopments.Onewouldexpecta
combinationoftechniquesusingsuperdefinitionopticalmicroscopywillhavethecapabilitytocircuit
traceandimagealltheparametersnecessaryforneuralemulationintenyearssinceanorderof
magnitudeofopticalresolutionhasbeenattainableinthelastdecadeandmajorimprovementsarestill
possible.
Indetailedstudiesofsmallsectionsofthehumanbrain(e.g.cerebralcortex)weknowweare
dealingwithrepeatingpatternsofneuralconnections.Thereareloadsofexceptionstotherepeating
patternsbutthegeneralstructureinagivenareacanbearoughguidecombinedwithotherinformation
forbothdecidingwhetheragivensynapseshouldbeinhibitoryorexcitatoryandthetypeof
neurotransmittersinvolved.
Dependingontheresourcesmadeavailablethetimeframeforthefirsthumanbraintobefaithfully
capturedcouldrangefromacoupleofdecadestomanydependingonwhetheritbecomesanational
priority.Duetotheincrediblecomplexityofthehumanbrainonemustassumetherewillbeseveral
yearsoffailuresbutthesefailuresshouldteachusagreatdeal.Success,withinthistimeframe,
dependsonwhethertheinformationnecessaryforemulationisknownorobtainableusingreasonable
extensionsofcurrenttechnology.Furthermore,itdependsonexceedinglyaccurateconnectome+
captureandemulationbecausesmallerrorsincriticalplacesmaycausepoorresults.
BrainEmulation 
In2008Koggeetal.[Kogge2008]laidouttherequirementsandproblemstobuildanexascale
computersystem.ThisbecameapriorityfortheUSgovernmentforavarietyofapplicationsranging
fromparticlesimulationsofstarformationtoquantumchemistry[Hopkins2011].Inanotherdecade
weshouldhavesupercomputersthatwillprocessexaflops(10
18
floatingpointoperationseachsecond).
TheplansofIntelandAMD[Intel2012,AMD2014,Tirias2014]targetprocessorsinthe2020time
framethatwillhave25timestheenergyefficiencyofcurrentprocessors.Energyefficiencyisakey
factorsinceIntel[Sodani2011]statedthatthethencurrent(2011)supercomputersystemsinthe10
petaflops/secondrangeburn12megawattsofpower.Theyseeanexaflopmachinein2020usingnot
morethan20megawatts(thehumanbrain,bycomparison,consumesbetween2540watts).In2013
KoggeandShalf[Kogge2013]pointedoutthatanotherchallengeforexascalesystemsisdatalocality.
Shufflingdataathighspeedandlowlatencybetweenprocessorsonchip,betweenchipsandbetween
circuitboardswillbecomemoreimportantthatjusttherawprocessingpower.Theyalsopointedout
thatbuildingcompilersthatrearrangedataandprocessingpowertoachievemoredatalocalitywill
becomevitalassystemsscalewithanordersofmagnitudemoreprocessorcores.Theexafloptargetis
importantsincesuchmachineswillbeabletoemulateone[Markram2012]ormorehumanbrains
usingspikingneuralmodelsinrealtimeandaverysophisticatedvirtualworldaswell.Thiswill
probablyhappenusingthesamesiliconbasedCMOS(ComplementaryMetalOxideSemiconductor)
processorpathwe'vefollowedsincetheRCA1802(COSMAC)microprocessorin1976.
Transistorcircuitshavebeenmadewithcarbonnanotubes[Bachtold2001,Sun2011]],graphene
[Liu2013]andMoS
2
[Mak2013]thatmaybecapableofverylowpowerandhighperformance.These
technologiesmaybetenyearsoutbeforereachingusefullevelsofintegrationandspeedorenduplike
siliconcarbide,diamond,galliumarsenideandgermaniumwithlimitedapplicationspaces.
Inthelastdecade,therehasbeenmanynewdevicesthatmayhavearoleinfuturebrainemulators
includingthememrister(HP),ReRAM(UniversityCollegeLondon),correlatedoxidediode(Harvard
University),32Gbittwistedradiobeams(UniversityofSouthernCalifornia),phasechangematerials
(PCMs)andmanygraphenestylenanomaterialsthathavememoryandswitchproperties.Verylow
powerismoreimportantthanspeedsinceitlimitsthepackingdensityandoperatingcostsofan
emulator.Connectivityanddevicesonchiphavemanyordersofmagnitudehigherreliabilitythanoff
chipconnectionsanddevices,sothemoredevicesonecanputonasubstratethebetterthechanceof
buildingareliablelargescalesystem.
Currentneuralsimulationmodels(theBrianprogram)[Goodman2008]usethebasic
Hodgkin–Huxleymodelwithasynchronoussystemclock.Thereis,infact,evidencethatspiking
correlationisusedforretinaandothertimevaryingsensoryinputfordatainterpretationandmaybea
fundamentalaspectoftheirutility.Therearemanycurrenteffortstomakeneuralnetworksimulations
thatarebiologicallyrealistic.PietroMazzonietal.[Mazzoni1990]showedthatspikingneuralmodels
couldlearnusingreinforcementratherthanthebiologicallyunrealisticbackpropagation.Asimple
modelofspikingneuronsthatreproducedtherichbehaviorofbiologicalcorticalneuronswas
demonstratedbyEugeneIzhikevich[Izhikevich2003].Thismodelreproducedspiking,bursting,
mixedmodefiringpatterns,postinhibitory(rebound)spikesandbursts,continuousspikingwith
frequencyadaption,spikethresholdvariability,bistabilityofrestingandspikingstates,and
subthresholdoscillationsandresonance.TchumatchenkoandClopath[Tchumatchenko2014]
demonstratedhowinhibitoryinterneuronsinthepresenceofasubthresholdresonanceselectively
amplifiesfiringrateresponsetoselectfrequenciesalreadypresentinisolatedneurons.Othermodels
andframeworksincludeGENESIS2.4,LFPy,MvaSpike,PyNN,NeuroML,NineML,SpineML
[Richmond2013].
It'snotclearthataspikingmodelwillbenecessaryforallneuralcircuitsbutcurrentlythereisno
clearcircuitandweightmappingformulafromspikingtononspikingtoachievethesameresultant
behavior.Thereis,infact,evidencethatspikingcorrelationisusedinmanyneuralcircuits[Goodman
2008].Currentnonspikingimplementationsofneuralnetsareapproximately100timesless
computationallyintensive.Othersimplificationmayalsobepossiblebyreplacingsomeneuralcircuits
withothermoreefficientcomputationalalgorithmsthatachievethesamefunctionality.
Muchcurrentemphasishasbeenonthehumanneocortexformachinelearningalgorithm
development[Numenta2011,Schmidhuber2013].Numenta'spurposeistodistilloutprinciplesfor
developingalgorithmsfromthewaythehumanneocortexseemstowork.Theirgoalisnottoreplicate
aparticularhumanneocortexbuttocreateacomputationalmodelforAIuse.Whattheyhaverealized
isthesparsedistributedrepresentationencodingandneuralnethierarchiesleadtowhattheyterm
HierarchicalTemporalMemory(HTM)fortheirclassofmodels.Thenextstepwastomakethebottom
layerneuralnetsrecurrent(RNN)orbidirectionalrecurrentneuralnets(BRNN)[Karpathy2014]since
thiscausedthelowerlevelstobecomemorestableandlongtermrecognizers.Thisclassofmodelshas
beenabreakthroughinAIandgivesusasimplifiedunderstandingofwhatgoesonintheneocortexbut
itisnotcurrentlypurposedforconvertinganactualhumanneuralnetwithassociatedsynapticweights
foremulation.
Anotherapproachistomakehardwarethatisarchitecturallydesignedtoemulateneuralnetworks.
ThishasbeenaccomplishedatIBM[Merolla2014,Cassidy2014]witha5.4billiontransistorchip
(TrueNorth)thatemulates4096neurosynapticcores.Ithandles1millionspikingneuronsand256
millionconfigurablesynapses.Thesechipscanbetiledin2dimensionsanduseonly65milliwattsor3
ordersofmagnitudelesspowerthanusingstateoftheartgeneralpurposemicroprocessors.Thisisa
verysignificantsteptowardadedicatedneuralemulatorandmaybeabetterfitthanageneralpurpose
supercomputer.However,itmaybesomewhatlessversatilewhenotheralgorithmsareavailableto
replacesomebrainparts.Anotherproblemisthischipismodeledonacorticalcolumnofamacaque
cerebralcortex.Ahumanequivalentcorticalcolumnismorecomplexbutshouldbereachablesince
thischipwasimplementedin28nmCMOSandthecurrentstateoftheart(Intel)is14nmCMOS.
TherearestillsomeproblemareassincePurkinjecellscanbranchouttoadendriticarborof200,000
fibersandapyramidalcellreceives30,000excitatoryinputsand1700inhibitoryinputs.Thereisstill
theproblemofhowglialcellsmodulateneuralactivity.Thisapproachhaspromisebutmayrequirea
piggybackmemoryandconnectionchiptoemulatetheselargerhumancellsandtheirconnectivity.To
reduceexternalchipconnectionsonereallyneedanentirehumancorticalstack(hypercorticalcolumn)
emulatedononechip.Itmaytakeastackofpiggybackchipsormoreactivesiliconlayerstodothat.
GeneralVisionhasacommerciallyavailablechipandboardsetsfeaturingoneormoreoftheir
CM1Kchipsthatcontain1024‘neurons’.Ratherthanaspikingneuronmodelthischipusesnumerical
valuesforinputsmakingitusefulforAIapplicationsbutprobablylessusefulforreplicatinghuman
neuralnetworks.
Otherspecializedneuralsimulationsystemsincludeamixedanalogdigitalmultichipsystem
(Neurogrid)[Benjamin2014],theUniversityofManchesterprojectSpiNNaker,theIBMSyNAPSE
project(GoldenGateChip),andtheHeidelbergUniversityBrainScaleproject(HICANNchip).Of
these,theNeurogridusestheleastchipareaandtheleastpowerandisonaparwithTrueNorthabove.
Currentlyitmakessensetomakesurethatthevarioustypesofhumanhypercorticalcolumns,and
otherlowerbrainsystemsoperatethesameinageneralpurposesupercomputersasinvivo.The
flexibilityofgeneralpurposecomputersisneededtomakesurethateverythingnecessaryforafaithful
simulationcanbeincludedthenspecialpurposesystemscanbedevelopedtoreplicatethosesimulated
features.Unfortunately,mostAIemphasisisjustonthecerebralcortexandthehypercorticalcolumns
ortheirmammaliancorticalcolumnancestors.Forthepurposesofbrainemulationtheentirehuman
brainisimportant,notjustthecerebralcortex.
VirtualEmbodiment
Initially,emulatedindividualsmayliveinasimulatedenvironmentinwhichtheycaninteractwith
eachotherandalsointeractwiththerealworldviatheInternet[Abrash2014].Evenwithtoday's
technology,suchanenvironmentwillbefarricherthanSecondLifebutwouldcurrentlyfallshortofa
realworldsensualexperience.Theresolutionofvariousvirtualworldsisincreasingatarapidratein
linewithcomputingpowerandisreachingthegeneralpublicintermsofinteractiveonlinegaming.
Sensoryandhapticbodysuits,omnidirectionaltreadmillsandsophisticatedheadgeararebeginningto
emergealongwithsupportingsoftware.Theemulatedhuman'sneuralinterfaceshouldbeabletotake
advantageofthesamevirtualinterfacethatrealhumanswouldbepluggedinto.
Realistic3Dvirtualreality(VR)worldsmaybecomewidespreadwiththeOculusRiftandother
VRviewersavailabletoconsumers.Theseviewerscanoffersufficientresolution,accuratetranslation
andlowpersistencetoavoidpreviousmotionsicknessproblemsforhumanviewers.Atfirst,
applicationswillbeprimarily3Dgamingbutlateritmaybeenvironmentswithhighlyevolvedcitiesin
virtualspacewithrealisticavatarsspeakingintheuser’svoiceoracopyoftheuser’svoice
synthesized.Thiswillbeaspaceformeetingpeopleprivatelyorpublicly,forshopping,recreationand
sightseeing.Eventuallyuserwillbeabletobecomeabirdorotheranimal,mythicalorrealinthis
world.Asavirtualperson(brainuploaded)thereisnoreasononecannotbeinthisworldwiththe
additionoffullsensoryfeedbacktothepersonsvirtualneuralsystem.Therealdangerofsuchavirtual
worldisbecominghabituatedtoitandhavingtherealworldseemblandbycomparison.
AndroidEmbodiment
Portableroboticsisadvancingatafairratebutisn'tfollowingMoore'sLawbyprogressingatan
exponentialrate.Therearenowwalkingrobotswithhighdexteritybutportablepowerandcompact
lowpowerhighefficiencyartificialmuscleshavebeenaproblem.Thefirstandroidsforbrain
emulatedindividualswillhavetobeRFlinkedtoasupercomputerfortheirbrainpowerandmayalso
haveapowertetherforlocomotion.Formoreportability,powercouldbesuppliedbyaplutonium238
radioisotopethermoelectricgenerator(RTG).RTGs,althoughtheyfoundwideuseinthe1970s,are
nowonlysuitedtospacecraftduetotheirexpense,threatofradioactivecontaminationafter
decommissioningandthescarcityofplutonium238.Adirectethanolfuelcellisadistinctpossibility
givenafewmoreyearsofdevelopment.Ifchemicalrechargeablebatterytechnologyimproves
markedlythenwellplacedfastrechargingstations,chairsorbedscouldsolvethisproblem.Atotally
independentandroidexistencemayinvolvereducinganexascaleoralternativecustomsupercomputer
andpowersourcetofitsomewhereinahumanbodysizeframecompletewithalltheotherpartsthat
wouldmaketheandroidrobotlifelikeandacceptabletotheuploadedhumanandtothoseheorshe
willinteractwith[Ishiguro12010,Ishiguro22010].Humansturnedandroidwillwanttolookandfeel
ashumanaspossibleandwillwantothertoreacttothemwithoutvisitingtheuncannyvalley(aterm
coinedbyMasahiroMoritodescribetheuncomfortablefeelingonehaswhenencounteringsomething
nearhumanlookingbutnotquitehumaninmotionorlikeness).
Hapticfeedbacktofeelvirtualheat,pressureandtexturearestillveryrudimentary[Varalakshmi
2012].Currentlygaming,robotics,surgicaltraining(laparoscopy,endoscopyandendovascular
procedures),telemanipulationandartificiallimbsprovidethemotivationforresearchinthisarea.For
thepurposeofhumanbrainemulationjustthedevelopmentofsensorsthatreplicatethehumansensory
systemsisimportant.Thesenseoftouchforanartificialhandisupto20sensors[Tan2014].The
majordrawbackisnotthesensorsbutconnectingthesetoappropriatenervesintheupperarm.Of
course,thisisnonissueinanandroidconnectedtoawholebrainemulator.Thehumanbodyhasfive
typesofcutaneousreceptors:LayeredcapsulePaciniancorpuscle(vibration),Layeredcapsule
Meissnercorpuscle(touch),thincapsuleRuffiniending(pressure),nonencapsulatedathairroot
(touch),andnonencapsulatedMerkelendings(touch).Afreenerveendingsensespain,temperature
andtouch[Nolte1999].Thesensordensityvariesfrom200MeissnerandMerkelunits/cm
2
onthe
fingertips,tongueandlipstoabout40onthepalmsandface.Organicfilmtransistorsandsensors
shouldallowthisdensityofsensorsinmaterialsthatcanbeformedintofingersandlipswithin5years
buttherehastobeaneedwithsufficientfinancingtodevelopthetechnology.
Humanmusclesarecompactandfairlyefficient;seeTable1from[Mirfakhrai2007].Thereare
manytypesofactuatorsthatcanactasartificialmusclesincludinghydraulic,pneumatic,thermal
shapedmemoryalloys,linearandrotarymotors,piezoelectricpolymers,ferroelectricpolymersand
manyothers.Tomakeandroidscordlessartificialmusclesneedactuatorsthataremorecompactand
uselesspowerthatcurrenttechnology.Astepinthatdirectionincludesrecentworkwithelectroactive
polymers(EAPs),conductivepolymers,carbonnanotubes(CNTs)[Mirfakhrai2007]andtwistedfibers
[Baughman2014,Haines12014,Haines22014,Lima2012].Aninexpensivenylonversionoftwisted
fibers,heatactivated,isfarmorepowerfulthanhumanmusclebyweightandsize.Usingcarbon
nanotubefilamentyarnyieldsanevenmorepowerfulcontractileforce.Asourceofheatisrequiredto
makethisworksothatisabitinconvenient,otherwise,thesmoothactionandsizemakesitidealfor
androidfacialandhandmuscles.
Table1Propertiesofmammalianskeletalmuscles
Property
TypicalValue
MaximumValue
Strain(%)
20
>40
Stress(MPa)
.01(Sustainable)
0.35
Workdensity(kj/m
3
)
8
Density(kg/m
3
)
1037
StrainRate(%/s)
500
Powertomass(W/kg)
50
200
Efficiency(%)
40
CycleLife
10
9
Modulus(MPa)
10to60
Visionandhearingcanuseofftheshelftechnologywithconventionaltechnologytoconvertthis
intoappropriateneuralspiketrains.Speechcanuseavocaltractsimulator.Thehumansenseoftaste
includessalt,bitter,sour,sweetandumami[Trivedi2012].Smellmayconsistofasmanyas1000
differentkindsofolfactoryneurons,allsimilarbutwitheach'binding'andifferentarrayofodorants.
Althoughhumantastehasbeencrudelyreplicated,smellhasbeenmorechallenging.Inthelastdecade
nanocompositesensortechnologieshaveadvancedtothepointthatcompanieslikeSensigentsellsa32
sensorarraysensitivetoawideassortmentofvolatileorganiccompounds.Moresensitivesensorsuse
metalligandbondingcolorimetric5x5and6x6sensorarrays[Suslick2004].Thisisbasedonthe
strongandrelativelyspecificbondingbetweentheanalytesandachemoresponsivedye.Sensitivityof
subppmsensitivityisreadilyrealizablewiththistechnologyusinganinexpensiveCMOSorCCD
imagingarrayorscanner.
Summary
Therearethreeinterdependentareasthatwillrequirethemostfocus,agoodunderstandingofbrain
physiologyatthelocallevel(corticalcolumn,etc.)pertinenttoemulation,highreliabilitybraintissue
scanningautomationandfast3Dpatternrecognitionthat'sconsiderablymoreaccuratethanhuman
recognition.
Therearemanyneurontypes,synapsetypes,synapsegrowthmechanisms,neurotransmittersand
variousglialcellsthatneedtobefuntionallyunderstood.Patternrecognitionsoftwaremustrecognize
andrejectjunksuchasdamagedordeadneuronsandganglia,plaques,tumors,lesions,calcium
deposits,vascularanomalies,bloodclots,strokedamage,unimportantglialcellsandotherirrelevant
cellularmaterialorforeignmatter.Itwasn'tuntilthelate80sandearly90sthatnitricoxide(NO),
hydrogensulfide(H
2
S)andcarbonmonoxide(CO)wererecognizedaslocalbroadcasttype
neurotransmitters[Dawson1994].Beforethatnoonesuspectedthatgasescouldbeneurotransmitters.
JustrecentlyFieldshasobservedatypeofglialcell(astrocyte)communicationusingcalciumion
wavesbetweenthemselvesandnearbyneuronsthepurposeofwhichisstillnotwellunderstood[Fields
2013].Anotherunsuspectedinconvenienceisthatsomehippocampalpyramidalneuronshavebeen
discoveredwithdendritesthathavegrowntheirownaxons[Thome2014]thatfireindependentlyofthe
pyramidalcell.It'sbeenacardinalrulethataneuronhasoneandonlyoneaxonforacenturybutthese
recentdiscoveriesdemonstratethatneuralbiologystillhasfundamentalsurprises.
Thesecondareaofconcernishighlyreliableautomatedsectionpreparationsystemwithlowtissue
distortionandnolossofthefeatureinformationnecessaryforsimulation.Thisalsoinvolvesreliable
marking,dyingandcataloging(ifsampleisnotdestroyed)andstoragetechniques.Beyondthis,highly
parallelscanningmicroscopesmustbedevelopedalongwithmultiexabytestoragefacilitiesforthe
datathatwillbecollected(30x30x30nmvoxelforanaveragehumanmalebrainofvolume1260cm
wouldrequire4.7x10
19
bytesat1bytepervoxel).Notethattheentiredatabasewouldhavetobesaved
tomakecorrectionsbasedonadditionalinformation,betterpatternrecognitionalgorithmsoralgorithm
learning.
Thethirdhurdleisthepatternrecognitionofallfeaturesandinterconnectivityrelevanttoafaithful
reproductionoftheindividualhuman'sconnectome.Itisdifficultforawelltrainedhumantocircuit
traceevenaverysmallneuralnet.Ittookyearstoteaseoutthestructureofthe302(hermaphrodite)
and384(male)neuronwiringdiagramofnematodeC.elegans.Notethatjustrecently[Li2014]itwas
discoveredthatoneoftheneuronsinC.eleganswasactuallyprovidingtwodifferentoutputs:
locomotionspeedanddirection.Thisshowsthatapparentlysimpleneuralsystemscancontain
unsuspectedcomplexity.
HenryMarkram’sBlueBrainProjecthassimulatedajuvenilerat’ssomatosensorycortexconsisting
of31kneuronsand37millionsynapses[Markram2015].Thisisbasedonpatchclampstudiesthat
identified55layerspecificmorphologicaland207morphoelectricalneuronsubtypes.Thecapturing
ofahumancorticalcolumn(~100kneurons)thoughttobe700to1000minicolumnsisinprogress.
Capturingthecerebralcortexofanindividualisordersofmagnitudemoredifficultsinceinvivo
studiesaredifficultandasinotheranimalseverycorticalcolumnnetworkisdifferent(thoughmany
areverysimilar)sothisindicatesthattheremaybeagenericbiologicallyapproximatedsimulated
artificialintelligencewellbeforeanindividualbraincanbeuploaded.Onewouldexpectthatthe
humancorticalcolumnnetworkwouldprovideabettermodelforAIusethaneitheraratora
macaque’scorticalcolumn.
Inareaswherethefunctionofagroupofneuronscanbefullycharacterizedanalgorithmic
replacementmightbemadeandcommittedtodigitallogic.Aprosthesisreplacementforpartofthe
hippocampusisbeingdevelopedinTheodoreBerger’sLab[Berger2012]atUniversityofSouthern
California.Theintentionistodevelopabiomimeticmicroelectronicdevicetorestorelostmemory.
Otherbiomimeticdeviceslikethecochlearimplantforthedeafhavebeenaroundformanydecades.
Areaslikeseeing,movement/intentionarealsounderbiomimeticmicroelectronicdevelopmentat
variousuniversitylabswithprototypedevicesinafewhumansubjects.Thiscanworkwellforareas
wherelittleornoadultlearningisinvolvedsotheI/Oresponseisbasicallystatic.
TherearemanycriticsofboththeEuropeanandUScurrenteffortsinlargebudgetbrainresearch
dealingwithlargescalecomputersystemsforsimulations[Fregnac2014]sincemanybasicscience
neuroscientistsfeltleftoutoftheprocessandthecurrentroadmap.Thecriticsclaimthiswilldelay
resultsusefulforcuringbraindiseases.Ifthecriticshavetheirwayitwilllikelydelaybotheffortsat
extractingwhatisusefulforAIandwholebrainemulation.Thisissimilartotheuproaroverthehuman
genomeprojectbegunin1990thatfeltmappingthehumangenomewouldnotleadtocuresfordisease.
Insteadithadprovideduswithabigpictureunderstandingofgeneticdiseases,geneticpredispositions
tovariousdiseasesandopenedtodoortomodernepigeneticsthatwouldhavepreviouslybeenwritten
offasLysenkoism.
Table2containsguesstimatesofwheneachbenchmarkmaybeaccomplished.Thisassumes
fundingisavailable.The“CapturingIndividualHumanBrainConnectome”entryisdependentonthe
priortableentriesandthewill(andmoney)tocreateasectioningandscanningfactory.
Akeypartoftheproblemwasnotincludedintheguesstimatesorinthedetaileddiscussioninthis
paper:Theproblemofbuildinganemulatorthatcontainstherightfunctions,arrangedintheright
manner,withsufficientlycorrectfunctionparametersderivedbysometransformationfromthevarious
measurementsmadeinabrain.Atpresent,thereisnosuchemulator,noristhereamethodfor
populatingsuchanemulatorwithparametervalues.Itisacrucialareathatdemandsfurtherresearch
anddevelopment.
DevelopmentLeadTimes
HumanBrainPreservationsufficientforstorageand
uploading
1015years
UnderstandingBrainPhysiologyatdepthnecessary
forbehavioralemulation
10–15years
AutomatedBrainsectioningandscanning
1015years
CapturingIndividualHumanBrainConnectome
2550years*
ComputationPotentialforHumanBrainEmulation
(LevelOutput).Probablywon'tworkbutusefulforAI
Currentmultipetaflopsupercomputers
ComputationPotentialforBrainEmulation(Spiking
Output)
10–20years(ExascaleSuper
Computers)orSpecializedarchitectures
VirtualWorldExistence(~computationalpower)
Possiblenow,mixofAIsandgamers
AndroidExistence(RFlinkandpartiallytethered)
1015Years
IndependentAndroidExistence(selfcontained)
2550years
*AnApollolikeprojectcouldgetearlyresultsbuteconomicandpoliticalconsiderationscoulddelayresultsformany
years.
Table2
References
[Abbott1999]L.F.Abbott,“Lapicque’sintroductionoftheintegrateandfiremodelneuron(1907)
BrainResearchBulletin,Vol.50,Nos.5/6,pp.303–304,1999
[Abrash2014]MichaelAbrash,“WhatVRcould,should,andalmostcertainlywillbewithintwo
years”ValveatSteamDevDaysinSeattleFebruary2014
[Alivisatos2013]A.PaulAlivisatos,etal.,“NanotoolsforNeuroscienceandBrainActivity
Mapping”ACSNano7(3)pp18501866
[AMD2014]"AMDAcceleratesEnergyEfficiencyofAPUs,DetailsPlanstoDeliver25xEfficiency
Gainsby2020
http://www.amd.com/enus/pressreleases/Pages/amdacceleratesenergy2014jun19.aspx
[Andres2014]BjörnAndres,etal.,“SegmentationofSBFSEMVolumeDataofNeuralTissueby
HierarchicalClassification”DOI:10.1007/9783540693215_15Conference:PatternRecognition,
30thDAGMSymposium,Munich,Germany,June1013,2008,Proceedings
[Bachtold2001]AdrianBachtold,PeterHadley,TakeshiNakanishi,CeesDekker,“LogicCircuitswith
CarbonNanotubeTransistors”ScienceVol2949November2001pp13171320
http://www.cs.duke.edu/courses/spring06/cps296.5/papers/BHND01.pdf
[Baughman2014]RayBaughman,“SpinningYarnintoMuscles”Science20February2014
[Benjamin2014]BenVarkeyBenjamin,etal.,“Neurogrid:MixedAnalogDigitalMultichipSystem
forLargeScaleNeuralSimulations”Vol102,No.5,May2014ProceedingsoftheIEEE
[Berger2012]TheodoreW.Berger,etal.,“TowardaCognitiveHippocampalNeuralProthesis:
ImplantableBiomimeticMicroelectronicstoRestoreLostmemory”CNS*2012WorkshoponMethods
ofSystemIdentificationForStudyingInformationProcessinginSensorySystems.July25,2012
Atlanta/Decatur,GA.
http://www.bionet.ee.columbia.edu/workshops/cns/methods/2012/identification
[Bonhoeffer2002]TobiasBonhoefferandRafaelYuste,“SpineMotility:Phenomenology,
Mechanisms,andFunction”Neuron,Vol.35,1019–1027,September12,2002
[Brette2009]RomainBrette,“SpikedBasedModelsofNeuralComputation”Thesis2009–Université
ParisV
[Briggman2012]“KevinL.BriggmanandDavidD.Bock,“VolumeElectronMicroscopyfor
NeuronalCircuitReconstruction”CurrentOpinioninNeurobiology201222:154161
[Callaway2008]EdwardM.Callaway,“TransneuronalCircuitTracingwithNeurotropicViruses”
CurrentOpinioninNeurobiology2008,18:617623
[Cassidy2014]AndrewS.Cassidy,etal.,“RealtimeScalableCorticalComputingat46GigaSynaptic
OPS/Wattwith~100XSpeedupinTimetoSolutionand~100,000XReductioninEnergytoSolution”
SC14,November1621,2014,Neworleans9781479955008/14/$31.00©2014IEEE
[Chen2015]FeiChen,PaulW.Tillberg,EdwardS.Boyden,“ExpansionMicroscopy”Sciencexpress
15January2015,DOI:10.1126/science.1260088
[Cheng2013]KwanghunCheng,etal.,“StructrualandMolecularInterrogationofIntactBiological
Systems”NatureDOI10.1038/12107
[CohenAdad2012]JulienCohenAdad,etal.,“ImprovedQBallimagingusinga300mT/mhuman
gradient”
http://www.humanconnectomeproject.org/wpcontent/uploads/2012/06/CohenAdad__abstract_ISMR
M12_QBall3.pdf
[Dawson1994]TedM.DawsonandSolomonH.Snyder,“GasesasBiologicalMessengers:Nitric
OxideandCarbonMonoxideintheBrain”TheJournalofNeuroscienceSeptember1994,14(9):
51475159
[Deca2014]RandalA.KoeneandDianaDeca(2014).Editorial:WholeBrainEmulationseeksto
ImplementaMindanditsGeneralIntelligencethroughSystemIdentification,JournalofArtificial
GeneralIntelligence.Volume4,Issue3,Pages1–9,ISSN(Online)19460163,DOI:
10.2478/jagi20130012,April2014
[Denk2012]WinfriedDenk,KevinLBriggman,andMoritzHelmsteadter.“Structuralneurobiology:
missinglinktoamechanisticunderstandingofneuralcomputation,”NatureReviewsNeuroscience
13.5(2012),351–8
[Dethier2011]JulieDethier,etal.,“SpikingNeuralNetworkDecoderforBrainMachineInterfaces”
Proceedingsofthe5thInternationalFrD1.14IEEEEMBSConferenceonNeuralEngineeringCancun,
Mexico,April27May1,2011
[EPFL2014]TheBlueBrainProjectEPFLhttp://bluebrain.epfl.ch/page58124en.html
[Evers2004]EversJF,SchmittS,SibilaM,DuchC.“Progressinfunctionalneuroanatomy:precise
automaticgeometricreconstructionofneuronalmorphologyfromconfocalimagestacks.J
Neurophysiology.2005Apr;93(4):233142.Epub2004Nov10.
[Eversmann2003]BjornEversmann,etal.,“A128x128CMOSBiosensorArrayforExtracellular
RecordingofNeuralActivity”IEEEJouralofSolidStateCirsuits,Vol.38,No.12,December2003
DOI:10.1109/JSSC.2003.819174
[Eyewire2014]Eyewire/ArtificialIntelligencehttp://wiki.eyewire.org/en/Artificial_Intelligence
[Fahy12004]GregoryM.Fahy,etal.,“CryopreservationofOrgansbyvitrification:Perspectivesand
RecentAdvances”CryobiologyDOI10.1016/j.cryobiology.2004.02.002PMID15094092
[Fahy22004]GregoryM.Fahy,BrianWowk,JunWu,andSharonPaynter,“Improvedvitrification
solutionsbasedonthepredictabilityofvitrificationsolutiontoxicity”Cryobiology48(2004)2235
[Feingold2012]JosephFeingold,etal.,"AsystemforRecordingNeuralActivityChronicallyans
simultaneouslyfromMultipleCorticalandSubcorticalRegionsinnonhumanprimates"DOI:
10.1152/jn.00625.2011(2012)
[Fields2013]
http://rdouglasfields.wordpress.com/2013/03/06/braincellsthatcommunicatewithoutelectricitycalci
umwavesinglia/
[Foerster2013]FoersterBR,etal.,“Animbalancebetweenexcitatoryandinhibitoryneurotransmitters
inamyotrophiclateralsclerosisrevealedbyuseof3Tprotonmagneticresonancespectroscopy”
JAMANeurol.2013Jun24:18.
[Fregnac2014]YvesFregnac,GillesLaurent“Neuroscience:WhereisthebrainintheHumanBrain
Project”Nature/Comment,03September2014
[Goodman2008]DanGoodman,RomainBrette,“Brian:ASimulatorforSpikingNeuralNetworksin
Python”FrontiersinNeuroinformatics,18November2008,DOI:10.3389/neuro.11.005.2008
[Hagmann2007]PatricHagmann,etal.,“MappingHumanWholeBrainStructuralNetworkswith
DiffusionMRI”PLoSONE.2007;2(7):e597.PublishedonlineJul4,2007.
DOI:10.1371/journal.pone.0000597
[Haines12014]CarterS.Haines,etal.,“ArtificialMusclesfromFishingLineandSewingThread”
Science343,868(2014)DOI:10.1126/science.1246906
[Haines22014]CarterS.Haines,etal.,“SupplementaryMaterialsforArtificialMusclesfromFishing
LineandSewingThread”21February2014,Science343,868(2014),DOI:10.1126/science.1246906
[Hayworth2012]KennethJ.Hayworth,"ElectronImagingTechnologyforWholeBrainNeuralCircuit
Mapping"InternationalJournalofmachineConsciousnessVol.4,No.1(20112)1250006DOI:
10.1142/S1793843012500060
[HCP2012]HumanConncetomeProjecthttp://www.humanconnectome.org/about/project
[He2014]XiaoweiHe,etal.,“CarbonNanotubeTerahertzDetector”NanoLett.,2014,14
(7),pp
3953–3958DOI:10.1021/nl5012678May29,2014
[HerculanoHouzel2009]SuzanaHerculanoHouzel,“TheHumanBraininNumbers:alinearly
scaledupprimatemodel”FrontiersinHumanNeuroscienceDOI10.3389/neuro.09.031
[Hopkins2011]CurtHopkins,“ObamaBudgetIncludes$126MillionforExascaleComputing”
http://readwrite.com/2011/02/21/obama_budget_includes_126_million_for_exascale_com#awesm=~oI
liLPMVOQJLxo
[Huang2008]BoHuangetal.,“ThreedimensionalSuperresolutionImagingbyStocasticOptical
ReconstructionMicroscopy”,Science2008February8;319(5864):810813.
DOI:10.1126/science.1153529.
[Intel2012]IntelEuropeanExascaleLabsReport2012
http://www.exascalecomputing.eu/wpcontent/uploads/2013/09/IntelEuropeanExascaleLabsAnnual
Report2012.pdf
[Ishiguro12010]HiroshiIshiguro,
http://spectrum.ieee.org/robotics/humanoids/hiroshiishigurothemanwhomadeacopyofhimself
[Ishiguro22010]HiroshiIshiguro,
http://spectrum.ieee.org/automaton/robotics/humanoids/040310geminoidfhiroshiishigurounveilsne
wsmilingfemaleandroid 
[Izhikevish2003]EugeneM.Izhikevich,“SimpleModelofSpikingNeurons”IEEETransactionson
NeuralNetworks”,Vol.14,No.6,November2003
[Karpathy2014]AndrejKarpathyandLiFeiFei,“DeepVisualSemanticAlignmentsforGenerating
ImageDescriptions”,StanfordUniversityTechnicalReportCS231n/2014
[Knott2008]GrahamKnott,HerschelMarchman,DavidWall,BenLich,“SerialSectionScanning
ElectronMicroscopyofAdultBrainTissueUsingFocusedIonBeamMilling”TheJournalof
Neuroscience,19March200828(12):29592964;DOI:10.1523/jneurosci.318907.2008
[Koene2012a]RandalA.Koene(2012).FundamentalsofWholeBrainEmulation:State,Transition
andUpdateRepresentations.InternationalJournalofMachineConsciousness,Vol.4,No.1(2012)
1250002,doi:10.1142/S1793843012500023
[Koene2012b]RandalA.Koene(2012).ExperimentalResearchinWholeBrainEmulation:TheNeed
forInnovativeInVivoMeasurementTechniques.InternationalJournalofMachineConsciousness,
Vol.4,No.1(2012)1250004,doi:10.1142/S1793843012500047
[Kogge2008]PeterKogge,etal.,“ExaScaleComputingStudy:TechnologyChallengesinAchieving
ExascaleSystems”September28,2008AFRLcontract#FA865007C7724
[Kogge2013]PeterKogge,JohnShalf,“ExascaleComputingTrends:Adjustingtothe“NewNormal”
forComputerArchitecture”ComputinginScience&Engineering,November/December2013
CISE156Shalf.Indd
[Kramer2013]TanKramer,LynnW.Enquist,“DirectionalSpreadofAlphaherpesvirusesinthe
NervousSystem”Viruses2013,5,678707;DOI:10.3390/v5020678
[Kreshuk2014]AnnaKreshuk,etal.,“AutomatedDetectionofSynapsesinSerialSection
TransmissionElectronMicroscopyImageStacks”PLoSONE9(2):e87351.
DOI:10.1371/journal.pone.0087351
[Lang2011]StephanLang,etal.“FastextractionofneuronmorphologiesfromlargescaleSBFSEM
imagestacks”10.3389/neuro.11.005.2008JournalofComputationalNeuroscience(ImpactFactor:
2.44).03/2011;31(3):53345.DOI:10.1007/s1082701103161(2011)
[Li2014]ZhaoyuLi,etal.,“EncodingofBothAnalogandDigitallikeBehavioralOutputsbyOneC.
elegansInterneuron”DOI:http://dx.doi.org/10.1016/j.cell.2014.09.056(2014)
[Lima2012]Marciod.Lima,etal.,“Electrically,Chemically,andPhotonicallyPoweredTorsionaland
TensileActuationofHybridCarbonNanotubeYarnMuscles”Science338,928(2012);
DOI:10.1126/science.1226762
[Liu2013]GuanxiongLiu,etal.,“GrapheneBasedNonBooleanLogicCircuits”Departmentof
ElectricalEngineering,UniversityofCalifornia–Riverside1308.2931v1.pdf
[Livet2007]JeanLivet,etal.,“Transgenicstrategiesforcombinatorialexpressionoffluorescent
proteinsinthenervoussystem”NatureVol450,1November2007DOI:10.1038/nature06293
[Mak2013]KinFaiMak,etal.,“TheValleyHallEffectinMoS2Transistors”
[Marblestone2013]AdamH.Marblestone,etal.,“Conneconomics:TheEconomicsofLargeScale
NeuralConnectomics”bioRxivpostedonlineDecember10,2013,DOI:
http://dx.DOI.org/10.1101/001214
[Markram2012]H.Markram,‘‘Thehumanbrainproject,’’Sci.Amer.,vol.306,no.6,pp.50–55,May
2012.
[Markram2015]HenryMarkram,etal.,“ReconstructionandSimulationofNeocorticalCirsuitry”Cell
Volume163,Issue2,p456492,8October2015DOI:http://dx.doi.org/10.1016/j.cell.2015.09.029
[Mazzoni1990]PietroMazzoni,etal.,“AMoreBiologicallyPlausibleLearningRuleforNeural
Networks”Proc.Natl.Acad.Sci.USAVol.88pp.44334437,May1991Neurobiology
[McIntyre2015]RobertL.McIntyre,GregoryM.Fahy,“AldehydestabilizedCryopreservation”
Cryobiology71(2015)448458,http://dx.doi.org/10.1016/j.cryobiol.2015.09.003
[Merolla2014]PaulA.Merolla,etal.,“AMillionSpikingNeuronIntegratedCircuitwithaScalable
CommunicationsNetworkandInterface”Science8August2014:Vol.345no.6197pp.668673,
DOI:10.1126/science.1254642
Suplementarytext:
http://www.sciencemag.org/content/suppl/2014/08/06/345.6197.668.DC1/Merolla.SM.rev1.pdf
[Micheva2010]KristinaD.Micheva,“SingleSynapseAnalysisofaDiverseSynapsePopulation:
ProteomicImagingMethodsandMarkers”NeuronDOI10.1016/j.neuron.2010.09.024
[Mikula2012]ShawnMikula,JonasBinding,WinfriedDenk,“Stainingandembeddingthewhole
mousebrainforelectronmicroscopy”NatureMethodspublishedonline21October2012;
DOI:10.1038/NMETH.2213
[Mirfakhrai2007]TissaphernMirfakhrai,JohnD.W.MaddenandRayBaughman“PolymerArtificial
Muscles”MaterialsTodayApril2007Vol10No.4,ISDN13697021ElsevierLtd.2007
[Nadasdy1998]ZoltanNádasdy,etal.,“ExtracellularRecordingandAnalysisofNeuronalActivity:
FromSingleCellstoEnsembles”,Chapter2
[NIH2013]AdvisoryCommitteetotheNIHDirectorINTERIMREPORT“BrainResearchthrough
AdvancingInnovativeNeurotechnologies(BRAIN)WorkingGroup”September16,2013
[NIH2014]BRAIN2025:“AScientificVisionBrainResearchthroughAdvancingInnovative
Neurotechnologies(BRAIN)”WorkingGroupReporttotheAdvisoryCommitteetotheDirector,NIH
June5,2014
[Nolte1999]JohnNolte“TheHumanBrain–AnIntroductiontoIt'sFunctionalAnatomy”Forth
Edition,SensoryReceptorsandThePeripheralNervousSystem,pp191214
[Numenta2011]Numenta,“HierarchialTemporalMemoryincludingHTMCorticalLearning
Algorithms.”Version0.2.1September12,2011
[Ostojic2009]Ostojic,S.;Brunel,N.;Hakim,V.(2009)."HowConnectivity,BackgroundActivity,
andSynapticPropertiesShapetheCrossCorrelationbetweenSpikeTrains".JournalofNeuroscience
19August2009,29(33):10234–10253.DOI:10.1523/JNEUROSCI.127509.2009
[Palm2010]ChristophPalm,etal.,“TowardsUltraHighResolutionFibreTractMappingofthe
HumanBrain–RegistrationofPolarisedLightImagesandReorientationofFibreVectors”FrontHum
Neurosci.2010;4:9.PMCID:PMC2866503PublishedonlineApr23,2010.PrepublishedonlineSep
28,2009.DOI:10.3389/neuro.09.009.2010
[Passingham2013]RichardE.Passingham,“Whatwecanandcannottellaboutthewiringofthe
humanbrain”DOI:10.1016/j.neuroimage.2013.01.010
[Persson2011]F.Persson,etal.,“FluorescenceNanoscopyofSingleDNAMoleculesbyUsing
StimulatedEmissionDepletion(STED)”DOI:10.1002/anie.201100371AngewandteChem.Int.Ed.
2011,50,5581–5583
[Ragan2012]Ragan,Yimothy,etal.,“Serialtwophotontomography:anautomatedmethodfor
exvivo
mousebrainimaging”NatMethods.;9(3):255–258.DOI:10.1038/nmeth.1854.
[Richmond2013]PaulRichmond,etal.,“FromModelSpecificationtoSimulationofBiologically
ConstrainednetworksofSpikingNeurons”Neuroinformatics2014;12(2):307323.doi:
10.1007/s120210139208z2013
[Sandberg2008]AndersSandberg,NickBostrom,“WholeBrainEmulation–ARoadmap”,Futureof
HumanityInstitute,OxfordUniversity,Technicalreport20083(2008)
[Schmidhuber2013]JurgenSchmidhuber,“MyFirstDeepLearningSystemof1991+DeepLearning
Timeline19622013”TecnicalreportarXiv:1312.5548v1[cs.NE]
[Schmitt2004]SchmittS,EversJF,DuchC,ScholzM,ObermayerK."Newmethodsforthe
computerassisted3Dreconstructionofneuronsfromconfocalimagestacks.”Neuroimage.2004
Dec;23(4):128398.
[Shibata2015]MikihiroShibata,etal.,“LongtipHighSpeedAtomicForceMicroscopyfro
Nanometerscaleimaginginlivecells”,NatureScientificReports5,8724;DOI:10.1038/srep08724
(2015).March4,2015.
[So2002]PeterT.C.So(MIT),“TwophotonFluorescenceLightMicroscopy”EncyclopediaofLife
Sciences,2002MacmillanPublishersLtd,NaturePublishingGroup
[Sodani2011]AvinashSodani,“RacetoExascale:OpportunitiesandChallenges”
http://www.microarch.org/micro44/files/MicroKeynoteFinalAvinashSodainash.pdf
[Sun2011]DongmingSun,etal.,“Flexiblehighperformancecarbonnanotubeintegratedcircuits”
NatureNanotechnology6,156161***
[Suslick2004]KennethS.Suslick,“AnOptoelectronicNose:“Seeing”SmellsbyMeansof
ColorimetricSensorArrays”MRSBULLETINOctober2004pg720725
[Tan2014]DanielW.Tan,etal.,Aneuralinterfaceprovideslongtermstablenaturaltouch
perception”SciTranslMed8October2014:Vol.6,Issue257,p.257ra138Sci.Transl.Med.DOI:
10.1126/scitranslmed.3008669
[Tchumatchenko2014]TatjanaTchumatchenkoandClaudiaClopath,“Oscillationsemergingfrom
noisedrivensteadystateinnetworkswithelectricalsynapsesandsubthresholdresonance”2NATURE
COMMUNICATIONS|5:5512|DOI:10.1038/ncomms6512
[Thome2014]ChristianThome,etal.,“AxonCarryingDendritesConveyPrivilegedInputin
HippocampalNeurons”DOI:http://dx.dol.org/10.1016/j.neuron.2014.08.013
[Tirias2014]TiriasResearch“AMDTargetsAcceleratingEnergyEfficiencyGains”June2014
[Trivedi2012]BijalP.Trivedi,“Neuroscience:Hardwiredfortaste”Nature
486,S7–S9(21June
2012)DOI:10.1038/486S7a
[Turaga2014]SrinivasC.Turaga,etal.,“ConvolutionalNetworksCanLearntoGenerateAffinity
GraphsforImageSegmentation.”NeuralComputation22.2(2014):511538.MIT
[Ugurbil2013]KamilUgurbil,etal.,“Pushingspatialandtemporalresolutionforfunctionaland
diffusionMRIintheHumanConnectomeProject”
http://dx.DOI.org/10.1016/j.neuroimage.2013.05.012
[Varalakshmi2012]B.D.Varalakshmi,etal.,“Haptics:StateoftheArtSurvey”InternationalJournal
ofComputerScienceIssues,pp234244(2012)http://ijcsi.org/papers/IJCSI953234244.pdf
[Wada2010]HitoshiWada,etal.,“ProspectofHighFieldMRI”IEEETransactionsonApplied
Superconductivity20No.3,115122
[Yoon2013]InhoYoon,etal.,“IntracellularNeuralRecordingwithPureCarbonNanotubeProbes”
DOI:10.1371/journal.pone.0065715(2013)
[Zador2014]IanD.Peikon,DianaI.GizatullinaandAnthonyM.Zador,“InVivoGenerationofDNA
SequenceDiversityforCellularBarcoding”NucleicAcidsResearch,2014DOI:10.1093/nar/gku604
[ZEISS2013]ZeissNews8.10.2013Dr.ChristianJagerOverviewproductPortfolioZEISS
Microscopy
Article
Full-text available
Visualization of morphological dynamics of live cells with nanometer resolution under physiological conditions is highly desired, but challenging. It has been demonstrated that high-speed atomic force microscopy is a powerful technique for visualizing dynamics of biomolecules under physiological conditions. However, application of high-speed atomic force microscopy for imaging larger objects such as live mammalian cells has been complicated because of the collision between the cantilever and samples. Here, we demonstrate that attaching an extremely long (~3 μm) and thin (~5 nm) tip by amorphous carbon to the cantilever allows us to image the surface structure of live cells with the spatiotemporal resolution of nanometers and seconds. We demonstrate that long-tip high-speed atomic force microscopy is capable of imaging morphogenesis of filopodia, membrane ruffles, pit formation, and endocytosis in COS-7, HeLa cells and hippocampal neurons.
Article
Full-text available
Oscillations play a critical role in cognitive phenomena and have been observed in many brain regions. Experimental evidence indicates that classes of neurons exhibit properties that could promote oscillations, such as subthreshold resonance and electrical gap junctions. Typically, these two properties are studied separately but it is not clear which is the dominant determinant of global network rhythms. Our aim is to provide an analytical understanding of how these two effects destabilize the fluctuation-driven state, in which neurons fire irregularly, and lead to an emergence of global synchronous oscillations. Here we show how the oscillation frequency is shaped by single neuron resonance, electrical and chemical synapses.The presence of both gap junctions and subthreshold resonance are necessary for the emergence of oscillations. Our results are in agreement with several experimental observations such as network responses to oscillatory inputs and offer a much-needed conceptual link connecting a collection of disparate effects observed in networks.
Article
Full-text available
Touch perception on the fingers and hand is essential for fine motor control, contributes to our sense of self, allows for effective communication, and aids in our fundamental perception of the world. Despite increasingly sophisticated mechatronics, prosthetic devices still do not directly convey sensation back to their wearers. We show that implanted peripheral nerve interfaces in two human subjects with upper limb amputation provided stable, natural touch sensation in their hands for more than 1 year. Electrical stimulation using implanted peripheral nerve cuff electrodes that did not penetrate the nerve produced touch perceptions at many locations on the phantom hand with repeatable, stable responses in the two subjects for 16 and 24 months. Patterned stimulation intensity produced a sensation that the subjects described as natural and without "tingling," or paresthesia. Different patterns produced different types of sensory perception at the same location on the phantom hand. The two subjects reported tactile perceptions they described as natural tapping, constant pressure, light moving touch, and vibration. Changing average stimulation intensity controlled the size of the percept area; changing stimulation frequency controlled sensation strength. Artificial touch sensation improved the subjects' ability to control grasping strength of the prosthesis and enabled them to better manipulate delicate objects. Thus, electrical stimulation through peripheral nerve electrodes produced long-term sensory restoration after limb loss.
Article
Full-text available
This paper presents a novel approach to the understanding of Haptic and its related fields where haptics is used extensively like in display systems, communication, different types of haptic devices, and interconnection of haptic displays where virtual environment should feel like equivalent physical systems. There have been escalating research interests on areas relating to haptic modality in recent years, towards multiple fields. However, there seems to be limited studies in determining the various subfields and interfacing and related information on haptic user interfaces and its influence on the fields mentioned. This paper aims to bring forth the theory behind the essence of Haptics and its Subfields like haptic interfaces and its applications.
Article
Full-text available
Europe's €1-billion science and technology project needs to clarify its goals and establish transparent governance, say Yves Frégnac and Gilles Laurent.
Article
Full-text available
Heterogeneity is a ubiquitous feature of biological systems. A complete understanding of such systems requires a method for uniquely identifying and tracking individual components and their interactions with each other. We have developed a novel method of uniquely tagging individual cells in vivo with a genetic ‘barcode’ that can be recovered by DNA sequencing. Our method is a two-component system comprised of a genetic barcode cassette whose fragments are shuffled by Rci, a site-specific DNA invertase. The system is highly scalable, with the potential to generate theoretical diversities in the billions. We demonstrate the feasibility of this technique in Escherichia coli. Currently, this method could be employed to track the dynamics of populations of microbes through various bottlenecks. Advances of this method should prove useful in tracking interactions of cells within a network, and/or heterogeneity within complex biological samples.
Technical Report
We present a model that generates free-form natural language descriptions of image regions. Our model leverages datasets of images and their sentence descriptions to learn about the inter-modal correspondences between text and visual data. Our approach is based on a novel combination of Convolutional Neural Networks over image regions, bidirectional Recurrent Neural Networks over sentences, and a structured objective that aligns the two modalities through a multimodal embedding. We then describe a Recurrent Neural Network architecture that uses the inferred alignments to learn to generate novel descriptions of image regions. We demonstrate the effectiveness of our alignment model with ranking experiments on Flickr8K, Flickr30K and COCO datasets, where we substantially improve on the state of the art. We then show that the sentences created by our generative model outperform retrieval baselines on the three aforementioned datasets and a new dataset of region-level annotations.
Article
We present a first-draft digital reconstruction of the microcircuitry of somatosensory cortex of juvenile rat. The reconstruction uses cellular and synaptic organizing principles to algorithmically reconstruct detailed anatomy and physiology from sparse experimental data. An objective anatomical method defines a neocortical volume of 0.29 ± 0.01 mm(3) containing ∼31,000 neurons, and patch-clamp studies identify 55 layer-specific morphological and 207 morpho-electrical neuron subtypes. When digitally reconstructed neurons are positioned in the volume and synapse formation is restricted to biological bouton densities and numbers of synapses per connection, their overlapping arbors form ∼8 million connections with ∼37 million synapses. Simulations reproduce an array of in vitro and in vivo experiments without parameter tuning. Additionally, we find a spectrum of network states with a sharp transition from synchronous to asynchronous activity, modulated by physiological mechanisms. The spectrum of network states, dynamically reconfigured around this transition, supports diverse information processing strategies. Paperclip: VIDEO ABSTRACT.
Article
Inspired by the brain’s structure, we have developed an efficient, scalable, and flexible non–von Neumann architecture that leverages contemporary silicon technology. To demonstrate, we built a 5.4-billion-transistor chip with 4096 neurosynaptic cores interconnected via an intrachip network that integrates 1 million programmable spiking neurons and 256 million configurable synapses. Chips can be tiled in two dimensions via an interchip communication interface, seamlessly scaling the architecture to a cortexlike sheet of arbitrary size. The architecture is well suited to many applications that use complex neural networks in real time, for example, multiobject detection and classification. With 400-pixel-by-240-pixel video input at 30 frames per second, the chip consumes 63 milliwatts.