Brain Connectivity and Model Comparison

38
Brain Connectivity and Model Comparison Will Penny Will Penny Wellcome Trust Centre for Neuroimaging, Wellcome Trust Centre for Neuroimaging, University College London, UK University College London, UK 26 26 th th November November 20 20 10 10

description

Brain Connectivity and Model Comparison. Will Penny. Wellcome Trust Centre for Neuroimaging, University College London, UK. 26 th November 20 10. Dynamic Causal Models. Neural state equation :. inputs. Dynamic Causal Models. Neural state equation :. MEG. Neural model: - PowerPoint PPT Presentation

Transcript of Brain Connectivity and Model Comparison

Page 1: Brain Connectivity and  Model Comparison

Brain Connectivity and Model Comparison

Will PennyWill Penny

Wellcome Trust Centre for Neuroimaging,Wellcome Trust Centre for Neuroimaging,University College London, UKUniversity College London, UK

2626thth November November 20 201010

Page 2: Brain Connectivity and  Model Comparison

),,( uxFx Neural state equation:

inputs

Dynamic Causal Models

Page 3: Brain Connectivity and  Model Comparison

),,( uxFx Neural state equation:

Neural model:8 state variables per region

nonlinear state equationpropagation delays

MEG

inputs

Dynamic Causal Models

Page 4: Brain Connectivity and  Model Comparison

),,( uxFx Neural state equation:

Electric/magneticforward model:

neural activityEEGMEGLFP

(linear)

Neural model:8 state variables per region

nonlinear state equationpropagation delays

MEG

inputs

Dynamic Causal Models

Page 5: Brain Connectivity and  Model Comparison

),,( uxFx Neural state equation:

Electric/magneticforward model:

neural activityEEGMEGLFP

(linear)

Neural model:1 state variable per regionbilinear state equationno propagation delays

Neural model:8 state variables per region

nonlinear state equationpropagation delays

fMRI MEG

inputs

Dynamic Causal Models

Page 6: Brain Connectivity and  Model Comparison

),,( uxFx Neural state equation:

Electric/magneticforward model:

neural activityEEGMEGLFP

(linear)

Neural model:1 state variable per regionbilinear state equationno propagation delays

Neural model:8 state variables per region

nonlinear state equationpropagation delays

fMRI MEG

inputs

Hemodynamicforward model:neural activityBOLD(nonlinear)

Dynamic Causal Models

Page 7: Brain Connectivity and  Model Comparison

Dynamic Causal Models

DCM for ERP/ERFDCM for Steady State Spectra

DCM for fMRI

DCM for Time Varying SpectraDCM for Phase Coupling

Page 8: Brain Connectivity and  Model Comparison

Synchronization

Gamma sync synaptic plasticity, forming ensembles

Theta sync system-wide distributed control (phase coding)

Pathological (epilepsy, Parkinsons)

Phase Locking Indices, Phase Lag etc are useful characterising systems in their steady state

Page 9: Brain Connectivity and  Model Comparison

For studying synchronization among brain regions Relate change of phase in one region to phase in others

Region 1

Region 3

Region 2

??

( )i i jj

g

Weakly Coupled Oscillators

Page 10: Brain Connectivity and  Model Comparison

One Oscillator

f1

Page 11: Brain Connectivity and  Model Comparison

Two Oscillators

f1

f2

Page 12: Brain Connectivity and  Model Comparison

Two Coupled Oscillators

f1

)sin(3.0 122 f

0.3

Page 13: Brain Connectivity and  Model Comparison

Stronger coupling

f1

)sin(6.0 122 f

0.6

Page 14: Brain Connectivity and  Model Comparison

Mutual Entrainment

)sin(3.0 122 f

0.30.3

)sin(3.0 211 f

Page 15: Brain Connectivity and  Model Comparison

3

2

1

DCM for Phase Coupling

)sin( jij

ijii af

Page 16: Brain Connectivity and  Model Comparison

3

2

1

DCM for Phase Coupling

)sin( jij

ijii af

])[sin( jij

ijkk

ii kaf

Page 17: Brain Connectivity and  Model Comparison

3

2

1

DCM for Phase Coupling

)sin( jij

ijii af

])[sin( jij

ijkk

ii kaf

])[cos(])[sin( jij

ijkk

jij

ijkk

ii kbkaf

Phase interaction function is an arbitrary order Fourier series

Page 18: Brain Connectivity and  Model Comparison

MEG Example

Fuentemilla et al, Current Biology, 2010

Page 19: Brain Connectivity and  Model Comparison

Delay activity (4-8Hz)

Duzel et al. (2005) find different patterns of sensor-space theta-coupling in the delay period dependent on task. We are now looking at source space and how this coupling evolves.

Page 20: Brain Connectivity and  Model Comparison

Data Preprocessing

Pick 3 regions based on source reconstruction

1. Right MTL [27,-18,-27] mm2. Right VIS [10,-100,0] mm3. Right IFG [39,28,-12] mm

Project MEG sensor activity onto 3 regions with fewer sources than sensors and known location, then pinv will do (Baillet et al., 2001)

Page 21: Brain Connectivity and  Model Comparison

Data Preprocessing

Pick 3 regions based on source reconstruction

1. Right MTL [27,-18,-27] mm2. Right VIS [10,-100,0] mm3. Right IFG [39,28,-12] mm

Project MEG sensor activity onto 3 regions with fewer sources than sensors and known location, then pinv will do (Baillet et al., 2001)

Bandpass data into frequency range of interest

Hilbert transform data to obtain instantaneous phase

Page 22: Brain Connectivity and  Model Comparison

Data Preprocessing

Pick 3 regions based on source reconstruction

1. Right MTL [27,-18,-27] mm2. Right VIS [10,-100,0] mm3. Right IFG [39,28,-12] mm

Project MEG sensor activity onto 3 regions with fewer sources than sensors and known location, then pinv will do (Baillet et al., 2001)

Bandpass data into frequency range of interest

Hilbert transform data to obtain instantaneous phase

Fit models to control data (10 trials) and memory data (10 trials).

Each trial comprises first 1sec of delay period.

Page 23: Brain Connectivity and  Model Comparison

QuestionWhich connections are modulated by memory task?

MTL

VISIFG

2.89

2.46

?

?

This question can be answered using Bayesian parameter inference

Page 24: Brain Connectivity and  Model Comparison

MTL

VISIFG

MTL

VISIFG

MTL

VISIFG

MTL

VISIFG

MTL

VISIFG

MTL

VISIFG1

MTL

VISIFG2

3

4

5

6

7

Master-Slave

PartialMutualEntrainment

TotalMutualEntrainment

MTL Master VIS Master IFG Master

Q. How do we compare these hypotheses ? A. Bayesian Model Comparison

Page 25: Brain Connectivity and  Model Comparison

LogEv

Model

1 2 3 4 5 6 70

50

100

150

200

250

300

350

400

450

LogBF model 3 versus model 1 > 20

Page 26: Brain Connectivity and  Model Comparison

MTL

VISIFG

2.89

2.46

0.89

0.77

Model 3

Page 27: Brain Connectivity and  Model Comparison

MTL-VIS

IFG

-V

IS

Control

Page 28: Brain Connectivity and  Model Comparison

MTL-VIS

IFG

-V

IS

Memory

Page 29: Brain Connectivity and  Model Comparison
Page 30: Brain Connectivity and  Model Comparison

Jones and Wilson, PLoS B, 2005

Recordings from rats doing spatial memory task:

Page 31: Brain Connectivity and  Model Comparison
Page 32: Brain Connectivity and  Model Comparison

Summary

Differential equation models of brain connectivity

Bayesian inference over parameters and models

DCM for Phase Coupling

Page 33: Brain Connectivity and  Model Comparison
Page 34: Brain Connectivity and  Model Comparison

Connection to Neurobiology:Septo-Hippocampal theta rhythm

Denham et al. 2000: Hippocampus

Septum

11 1 1 13 3 3

22 2 2 21 1

13 3 3 34 4 3

44 4 4 42 2

( ) ( )

( ) ( )

( ) ( )

( ) ( )

e e CA

i i

i e CA

i i S

dx x k x z w x Pdtdx x k x z w xdtdx x k x z w x Pdtdx x k x z w x Pdt

1x

2x 3x

4xWilson-Cowan style model

Page 35: Brain Connectivity and  Model Comparison
Page 36: Brain Connectivity and  Model Comparison

Four-dimensional state space

Page 37: Brain Connectivity and  Model Comparison

Hippocampus

Septum

A

A

B

B

Hopf Bifurcation

Page 38: Brain Connectivity and  Model Comparison

cossin)( baz

For a generic Hopf bifurcation (Erm & Kopell…)

See Brown et al. 04, for PRCs corresponding to other bifurcations