Aging, Memory and Alzheimer’s Disease Kinga Szigeti, MD, PhD.
Research on Learning & Memory with Neuromodulation in CLMN lab€¦ · Learning & Memory Clinical...
Transcript of Research on Learning & Memory with Neuromodulation in CLMN lab€¦ · Learning & Memory Clinical...
ResearchonLearning&MemorywithNeuromodulationinCLMNlab
PresentedbySungshinKimJune,30,2018http://clmnlab.com
Taxonomyofbiologicalmemory
3
WhymotorLearning?
OurlifeisacontinuumofmotorlearningLearningnewmotorskills
Adaptingtochanges
Regaininglostmotorskills
Twotypesofmotorlearning2.Motorskilllearning1. Motoradaptation
(1)visuomotor adaptation(kinematics)
(2)force-fieldadaptation(dynamics)
Advantageofmotorlearningresearch:ü Easytoquantifytheamountoflearningü Possibletomeasureprogressoflearningü Theoretical&computationalapproacheswithmodelsareencouragedBut,within-subjectexperimentdesignisdifficult
Effectsoftaskschedule onmotorlearningInterferencebetweentasks- Contextualinterferenceeffect
Timedecayofmemory- Spacingeffect
Grade:A
Grade:B
Grade:B
Grade:C
1h
Grade:A
1h1h1h1h 1h1h1h
4h 4h
8h
Acomputationalmodelofmotorlearning&memory
f s( 1) ( 1) ( 1)y n x n x n+ = + + +f( )/
f f f( 1) ( ) ( )T nx n x n e e nt b-+ = +
xs (n+1) = xs (n)e−T (n)/τ s +βse(n) ⋅c(n)
Fastmemory
SlowMemory
Forgetting Learning
Trials
Learning
Slowmemory
Fastmemory
Motoroutput
xs
xf
y
Fast memory
Slow memory
e+
y
d-
c
Bestschedulesforbothtwotasks?231 ≈2×109 possibleschedules!
Model-basedoptimalschedulesearchingAlgorithm(e.g.,Geneticalgorithm)
JYLee,YOh,SSKim,RScheidt,NScweighofer,NeuralComput,2016
…
Evolution(Crossover,Mutation)
Insearchofoptimallearningschedule
Neuralsubstratesofmemorieswithmultipletimescales(modeling)
SKim,KOgawa,JLv,NScweighofer,HImamizu,PLoS Biology,2015
o 2RMSE 4.96 ,R 0.981= =
xk (n+1) = xk (n)exp(−T (n) / τ k )+βk ⋅e(n) ⋅c(n) Forgetting Learning
βk =rτ kq (k =1,..,30)
y(n) = xk (n)T
k=1
30
∑ c(n)
Neuralsubstratesofmemorieswithmultipletimescales(fMRIresults)
SKim,KOgawa,JLv,NScweighofer,HImamizu,PLoS Biology,2015
Fourprincipalnetworkswithdifferenttimescales
Applyingthestate-of-the-artsparsesingularvaluedecompositionmethod
SKim,KOgawa,JLv, NScweighofer,HImamizu,PLoS Biology,2015
Multi-voxelpatternclassificationwithmachinelearningtechniques
Pattern Classifier
Task1
Task2
ClassificationofTask1vs.Task2- LinearSupportVectorMachine,averagedclassificationaccuracyreported
Non-invasiveneuromodulation :Transcranial MagneticStimulation(TMS)
Maedaetal.,ClinicalNeurophysiol,2000
facilitation
inhibition
TargetedTMSselectivelyactivatehippocampal-corticalmemorynetwork
Kimetal.,ScienceAdvances,2018
Stimulationlocation
Targetlocation
Current&FutureResearch
Multimodalapproachtolearning&memory
Behavior
Measurementoflearning,forgetting,interference
Modeling
Neuro-modulation
Computationmodelsforbehavioraldataandprediction
CausalinterventiontolearningusingTMS
fMRI
Investigationonneuralmechanisms
Learning&Memory
Clinicalstudy
Parkinson’sdisease,Stroke,Alzheimer’sdisease
Setupofthelaboratory
ReachingtaskinfMRIscanner MR-compatibledataglove
Visuomotor experimentsetup TMSwithNeuronavigation system
Ourteammembers
Sungshin KimPrincipalInvestigator
Kyusung LimPostdoc
Yera ChoiTo-be-hired
Yunha ShinTo-be-hired
Nayeon KwonStudentIntern
Topic1:Neuralcorrelatesofreward-basedmotorskilllearninginhigh-dimensionalspace
ThefirstfMRIexperimentoflearninganewmotorskillfromscratch
PreliminarybehavioralandfMRIresults
Early Late
FirstfMRIdemonstrationinhigh-dimensionalLearningmotorskillfromscratchWefoundstrongfMRIactivitiesinbilateralputamenmodulatingrewardduringmotorskilllearning
Topic2:TMSmodulationofmotorlearning&memory
SKim,etal.,PLoS Biology,2015
Fast memory
Slow memory
e+
y
d-
c
+
Disruptingplanning
Disruptinglearning
Group1 Group2
Thankyou
CLMN