Dynamic Causal Modelshelper.ipam.ucla.edu/publications/mbi2004/mbi2004_3847.pdfV5 SPC Wellcome...
Transcript of Dynamic Causal Modelshelper.ipam.ucla.edu/publications/mbi2004/mbi2004_3847.pdfV5 SPC Wellcome...
![Page 1: Dynamic Causal Modelshelper.ipam.ucla.edu/publications/mbi2004/mbi2004_3847.pdfV5 SPC Wellcome Department of Imaging Neuroscience, ION, UCL, UK. Contents • Neurodynamic model •](https://reader034.fdocuments.us/reader034/viewer/2022050122/5f527bcb42705b0f1913c179/html5/thumbnails/1.jpg)
Dynamic Causal Models
Will Penny
Olivier David, Karl Friston, Lee Harrison, Andrea Mechelli, Klaas Stephan
Mathematics in Brain Imaging, IPAM, UCLA, USA, July 22 2004.
V1
V5
SPC
V1
V5
SPC
Wellcome Department of Imaging Neuroscience, ION, UCL, UK.
![Page 2: Dynamic Causal Modelshelper.ipam.ucla.edu/publications/mbi2004/mbi2004_3847.pdfV5 SPC Wellcome Department of Imaging Neuroscience, ION, UCL, UK. Contents • Neurodynamic model •](https://reader034.fdocuments.us/reader034/viewer/2022050122/5f527bcb42705b0f1913c179/html5/thumbnails/2.jpg)
Contents• Neurodynamic model
• Hemodynamic model
• Model estimation and comparison
• Attention to visual motion
• Single word processingFriston et al.(2003) Neuro-Image, 19 (4), pp. 1273-1302.
![Page 3: Dynamic Causal Modelshelper.ipam.ucla.edu/publications/mbi2004/mbi2004_3847.pdfV5 SPC Wellcome Department of Imaging Neuroscience, ION, UCL, UK. Contents • Neurodynamic model •](https://reader034.fdocuments.us/reader034/viewer/2022050122/5f527bcb42705b0f1913c179/html5/thumbnails/3.jpg)
Contents• Neurodynamic model
• Hemodynamic model
• Model estimation and comparison
• Attention to visual motion
• Single word processing
![Page 4: Dynamic Causal Modelshelper.ipam.ucla.edu/publications/mbi2004/mbi2004_3847.pdfV5 SPC Wellcome Department of Imaging Neuroscience, ION, UCL, UK. Contents • Neurodynamic model •](https://reader034.fdocuments.us/reader034/viewer/2022050122/5f527bcb42705b0f1913c179/html5/thumbnails/4.jpg)
Single region
1 11 1 1z a z cu= +
u2
u1
z1
z2
z1
u1
a11c
![Page 5: Dynamic Causal Modelshelper.ipam.ucla.edu/publications/mbi2004/mbi2004_3847.pdfV5 SPC Wellcome Department of Imaging Neuroscience, ION, UCL, UK. Contents • Neurodynamic model •](https://reader034.fdocuments.us/reader034/viewer/2022050122/5f527bcb42705b0f1913c179/html5/thumbnails/5.jpg)
Multiple regions
1 11 1 1
2 21 22 2 2
00
z a z ucz a a z u
= +
u2
u1
z1
z2
z1
z2
u1
a11
a22
c
a21
![Page 6: Dynamic Causal Modelshelper.ipam.ucla.edu/publications/mbi2004/mbi2004_3847.pdfV5 SPC Wellcome Department of Imaging Neuroscience, ION, UCL, UK. Contents • Neurodynamic model •](https://reader034.fdocuments.us/reader034/viewer/2022050122/5f527bcb42705b0f1913c179/html5/thumbnails/6.jpg)
Modulatory inputs
1 11 1 1 12
2 21 22 2 21 2 2
0 0 00 0
z a z z ucu
z a a z b z u
= + +
u2
u1
z1
z2
u2
z1
z2
u1
a11
a22
c
a21
b21
![Page 7: Dynamic Causal Modelshelper.ipam.ucla.edu/publications/mbi2004/mbi2004_3847.pdfV5 SPC Wellcome Department of Imaging Neuroscience, ION, UCL, UK. Contents • Neurodynamic model •](https://reader034.fdocuments.us/reader034/viewer/2022050122/5f527bcb42705b0f1913c179/html5/thumbnails/7.jpg)
Reciprocal connections
1 11 12 1 1 12
2 21 22 2 21 2 2
0 00 0
z a a z z ucu
z a a z b z u
= + +
u2
u1
z1
z2
u2
z1
z2
u1
a11
a22
c
a12
a21
b21
![Page 8: Dynamic Causal Modelshelper.ipam.ucla.edu/publications/mbi2004/mbi2004_3847.pdfV5 SPC Wellcome Department of Imaging Neuroscience, ION, UCL, UK. Contents • Neurodynamic model •](https://reader034.fdocuments.us/reader034/viewer/2022050122/5f527bcb42705b0f1913c179/html5/thumbnails/8.jpg)
Neurodynamics
i ii
= + +∑z Az u B z Cu
InputsChange inNeuronalActivity
NeuronalActivity
IntrinsicConnectivityMatrix
ModulatoryConnectivityMatrices
InputConnectivityMatrix
V1
V5
SPC
![Page 9: Dynamic Causal Modelshelper.ipam.ucla.edu/publications/mbi2004/mbi2004_3847.pdfV5 SPC Wellcome Department of Imaging Neuroscience, ION, UCL, UK. Contents • Neurodynamic model •](https://reader034.fdocuments.us/reader034/viewer/2022050122/5f527bcb42705b0f1913c179/html5/thumbnails/9.jpg)
Contents• Neurodynamic model
• Hemodynamic model
• Model estimation and comparison
• Attention to visual motion
• Single word processing
![Page 10: Dynamic Causal Modelshelper.ipam.ucla.edu/publications/mbi2004/mbi2004_3847.pdfV5 SPC Wellcome Department of Imaging Neuroscience, ION, UCL, UK. Contents • Neurodynamic model •](https://reader034.fdocuments.us/reader034/viewer/2022050122/5f527bcb42705b0f1913c179/html5/thumbnails/10.jpg)
Hemodynamics
( , , )( )g z
y b==x x h
x
Hemodynamicvariables
For each region:
[ , , , ]s f v q=x
Hemodynamicparameters
Seconds
Dynamics
![Page 11: Dynamic Causal Modelshelper.ipam.ucla.edu/publications/mbi2004/mbi2004_3847.pdfV5 SPC Wellcome Department of Imaging Neuroscience, ION, UCL, UK. Contents • Neurodynamic model •](https://reader034.fdocuments.us/reader034/viewer/2022050122/5f527bcb42705b0f1913c179/html5/thumbnails/11.jpg)
Hemodynamic saturation
Sub-linear and super-linearresponses to pairs of stimuli
Equivalent input-outputfunctions
Neuronal impulse
Neuronal impulses
![Page 12: Dynamic Causal Modelshelper.ipam.ucla.edu/publications/mbi2004/mbi2004_3847.pdfV5 SPC Wellcome Department of Imaging Neuroscience, ION, UCL, UK. Contents • Neurodynamic model •](https://reader034.fdocuments.us/reader034/viewer/2022050122/5f527bcb42705b0f1913c179/html5/thumbnails/12.jpg)
Why have explicit models for neurodynamicsand hemodynamics ?
44332211 ucucucucazz ++++=
For 4 event types u1, u2, u3, u4 :
In a GLM for a single region, y=Xβ+e, with 3 basis functions per event type (canonical,shifter, stretcher) there are 12 parameters to estimate. These relate hemodynamics directly to each stimulus.
In a (single region) DCM there are 4 neuronal efficacy parameters relating neuronal activity to each stimulus
And 5 hemodynamic parameters relating neuronal activity to the BOLDsignal.
A total of 9 parameters.
),( hzby =
![Page 13: Dynamic Causal Modelshelper.ipam.ucla.edu/publications/mbi2004/mbi2004_3847.pdfV5 SPC Wellcome Department of Imaging Neuroscience, ION, UCL, UK. Contents • Neurodynamic model •](https://reader034.fdocuments.us/reader034/viewer/2022050122/5f527bcb42705b0f1913c179/html5/thumbnails/13.jpg)
DCM PriorsHemodynamics
Rate of signal decay: 0.65Elimination rate: 0.41Transit time: 0.98Grubbs exponent: 0.32Oxygenation fraction: 0.34
E[h]Cov[h]
Neurodynamics
Stability priors ensure principal Lyapunov exponent is less than zerowith high probability.
![Page 14: Dynamic Causal Modelshelper.ipam.ucla.edu/publications/mbi2004/mbi2004_3847.pdfV5 SPC Wellcome Department of Imaging Neuroscience, ION, UCL, UK. Contents • Neurodynamic model •](https://reader034.fdocuments.us/reader034/viewer/2022050122/5f527bcb42705b0f1913c179/html5/thumbnails/14.jpg)
Contents• Neurodynamic model
• Hemodynamic model
• Model estimation and comparison
• Attention to visual motion
• Single word processing
![Page 15: Dynamic Causal Modelshelper.ipam.ucla.edu/publications/mbi2004/mbi2004_3847.pdfV5 SPC Wellcome Department of Imaging Neuroscience, ION, UCL, UK. Contents • Neurodynamic model •](https://reader034.fdocuments.us/reader034/viewer/2022050122/5f527bcb42705b0f1913c179/html5/thumbnails/15.jpg)
Bayesian Estimation2( ) ( ; , )p pp Nθ θ µ σ=
Relative Precision Weighting
2( | ) ( ; , )ep y N y θ σθ =
2( | ) ( ; , )p y Nθ θ µ σ=
2 2 2
22 2
1 1 1
1 1
e p
pe p
y
σ σ σ
µ σ µσ σ
= +
= +
y eθ= +
pµ
yµ
Normal densities
![Page 16: Dynamic Causal Modelshelper.ipam.ucla.edu/publications/mbi2004/mbi2004_3847.pdfV5 SPC Wellcome Department of Imaging Neuroscience, ION, UCL, UK. Contents • Neurodynamic model •](https://reader034.fdocuments.us/reader034/viewer/2022050122/5f527bcb42705b0f1913c179/html5/thumbnails/16.jpg)
Multiple parameters
(1)θ
(2)θ
( ) ( ; , )p pp N=θ θ µ C
( | ) ( ; , )ep N=y θ y Xθ C
( | ) ( ; , )p N=θ y θ µ C
( )1 1 1
1 1
Te p
Te p p
− − −
− −
= +
= +
C X C X C
µ C X C y C µ
= +y Xθ e
One-step if Ce, Cp and µp are known
GeneralLinear Model
![Page 17: Dynamic Causal Modelshelper.ipam.ucla.edu/publications/mbi2004/mbi2004_3847.pdfV5 SPC Wellcome Department of Imaging Neuroscience, ION, UCL, UK. Contents • Neurodynamic model •](https://reader034.fdocuments.us/reader034/viewer/2022050122/5f527bcb42705b0f1913c179/html5/thumbnails/17.jpg)
Nonlinear models( )b= +y θ e
( )( ) ( ) ( )
( )
( )
ii i
i
i
i
bb b
b
r y br
∂= + −
∂∂
=∂
∆ = −= −= ∆ +
µθ µ θ µθ
µJθ
θ θ µµ
J θ e
( ) ( ; , )
( ) ( ; , )p p
p i p
p N
p N
=
∆ = ∆ −
θ θ µ C
θ θ µ µ C
( | ) ( ; ( ), )( | ) ( ; , )
e
e
p N bp N
=
∆ = ∆
y θ y θ Cr θ r J θ C
,i iµ C
( )( )1 1 1
1
1 11 1
Ti e p
Ti i i e p p i
− − −+
− −+ +
= +
= + + −
C J C J C
µ µ C J C r C µ µ
Gauss-Newton ascent with priors
Linearization
CurrentEstimates
{ , , , }=θ A B C h
θ∆
r
Friston et al.(2002) Neuro-Image, 16 (2), pp. 513-530.
![Page 18: Dynamic Causal Modelshelper.ipam.ucla.edu/publications/mbi2004/mbi2004_3847.pdfV5 SPC Wellcome Department of Imaging Neuroscience, ION, UCL, UK. Contents • Neurodynamic model •](https://reader034.fdocuments.us/reader034/viewer/2022050122/5f527bcb42705b0f1913c179/html5/thumbnails/18.jpg)
Model Comparison I
{ , , , }=θ A B C h
( | , ) ( | )( | , )( | )
p m p mp mp m
=y θ θθ y
y
V1
V5
SPC
( | ) ( | , ) ( | )p m p m p m d= ∫y y θ θ θ
Model, m Parameters:
PriorPosterior Likelihood
Evidence
Laplace, AIC, BIC approximations
Model fit + complexityPenny et al. (2004) NeuroImage, 22(3), pp. 1157-1172.
![Page 19: Dynamic Causal Modelshelper.ipam.ucla.edu/publications/mbi2004/mbi2004_3847.pdfV5 SPC Wellcome Department of Imaging Neuroscience, ION, UCL, UK. Contents • Neurodynamic model •](https://reader034.fdocuments.us/reader034/viewer/2022050122/5f527bcb42705b0f1913c179/html5/thumbnails/19.jpg)
Model Comparison II{ , , , }=θ A B C h
( | , ) ( | )( | , )( | )
p m p mp mp m
=y θ θθ y
y
V1
V5
SPC
Model, mParameters:
PriorPosterior Likelihood
( | ) ( )( | )( )
p m p mp mp
=yy
y
PriorPosterior Evidence
Parameter Parameter
Model Model
![Page 20: Dynamic Causal Modelshelper.ipam.ucla.edu/publications/mbi2004/mbi2004_3847.pdfV5 SPC Wellcome Department of Imaging Neuroscience, ION, UCL, UK. Contents • Neurodynamic model •](https://reader034.fdocuments.us/reader034/viewer/2022050122/5f527bcb42705b0f1913c179/html5/thumbnails/20.jpg)
Model Comparison III
V1
V5
SPC
( | ) ( | , ) ( | )p m i p m i p m i d= = = =∫y y θ θ θ
( | ) ( | , ) ( | )p m j p m j p m j d= = = =∫y y θ θ θ
Model, m=i
V1
V5
SPC
Model, m=j
Model Evidences:
Bayes factor:
( | )( | )ijp m iBp m j
==
=yy
1 to 3: Weak3 to 20: Positive20 to 100: Strong>100: Very Strong
![Page 21: Dynamic Causal Modelshelper.ipam.ucla.edu/publications/mbi2004/mbi2004_3847.pdfV5 SPC Wellcome Department of Imaging Neuroscience, ION, UCL, UK. Contents • Neurodynamic model •](https://reader034.fdocuments.us/reader034/viewer/2022050122/5f527bcb42705b0f1913c179/html5/thumbnails/21.jpg)
Contents• Neurodynamic model
• Hemodynamic model
• Bayesian estimation
• Attention to visual motion
• Single word processing
![Page 22: Dynamic Causal Modelshelper.ipam.ucla.edu/publications/mbi2004/mbi2004_3847.pdfV5 SPC Wellcome Department of Imaging Neuroscience, ION, UCL, UK. Contents • Neurodynamic model •](https://reader034.fdocuments.us/reader034/viewer/2022050122/5f527bcb42705b0f1913c179/html5/thumbnails/22.jpg)
Attention to Visual MotionAttention to Visual Motion
STIMULISTIMULI250 250 radiallyradially moving dots at 4.7 degrees/smoving dots at 4.7 degrees/s
PREPRE--SCANNINGSCANNING5 x 30s trials with 5 speed changes (reducing to 1%)5 x 30s trials with 5 speed changes (reducing to 1%)Task Task -- detect change in radial velocitydetect change in radial velocity
SCANNINGSCANNING (no speed changes)(no speed changes)6 normal subjects, 4 100 scan sessions;6 normal subjects, 4 100 scan sessions;each session comprising 10 scans of 4 different conditioneach session comprising 10 scans of 4 different condition
1. Photic2. Motion3. Attention
Experimental Factors
Buchel et al. 1997
![Page 23: Dynamic Causal Modelshelper.ipam.ucla.edu/publications/mbi2004/mbi2004_3847.pdfV5 SPC Wellcome Department of Imaging Neuroscience, ION, UCL, UK. Contents • Neurodynamic model •](https://reader034.fdocuments.us/reader034/viewer/2022050122/5f527bcb42705b0f1913c179/html5/thumbnails/23.jpg)
Specify regions of interest
Identify regions of Interest eg. V1, V5, SPC
GLM analysis
V1
V5
SPC
Motion
Photic
Att
Model 1
![Page 24: Dynamic Causal Modelshelper.ipam.ucla.edu/publications/mbi2004/mbi2004_3847.pdfV5 SPC Wellcome Department of Imaging Neuroscience, ION, UCL, UK. Contents • Neurodynamic model •](https://reader034.fdocuments.us/reader034/viewer/2022050122/5f527bcb42705b0f1913c179/html5/thumbnails/24.jpg)
V1
V5
SPC
Motion
Photic
Att
Model 1−10
−5
0
5
10
−4
−2
0
2
4
0 200 400 600 800 1000 1200−4
−2
0
2
4
V1
V5
Estimation
SPC
Time (seconds)
![Page 25: Dynamic Causal Modelshelper.ipam.ucla.edu/publications/mbi2004/mbi2004_3847.pdfV5 SPC Wellcome Department of Imaging Neuroscience, ION, UCL, UK. Contents • Neurodynamic model •](https://reader034.fdocuments.us/reader034/viewer/2022050122/5f527bcb42705b0f1913c179/html5/thumbnails/25.jpg)
Posterior Inference
−0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.70
1
2
3
4
5
6
B321
P(B
3 21|y
)
γ
How muchattention (input 3) changes connection
fromV1 (region 1)
to V5 (region(2)
![Page 26: Dynamic Causal Modelshelper.ipam.ucla.edu/publications/mbi2004/mbi2004_3847.pdfV5 SPC Wellcome Department of Imaging Neuroscience, ION, UCL, UK. Contents • Neurodynamic model •](https://reader034.fdocuments.us/reader034/viewer/2022050122/5f527bcb42705b0f1913c179/html5/thumbnails/26.jpg)
V1
V5
SPC
Motion
Photic
Att
Model 1
Motion
Photic
Att
V1
V5
SPC
Model 2
Bayes Factor B12 > 1019
VeryStrong
![Page 27: Dynamic Causal Modelshelper.ipam.ucla.edu/publications/mbi2004/mbi2004_3847.pdfV5 SPC Wellcome Department of Imaging Neuroscience, ION, UCL, UK. Contents • Neurodynamic model •](https://reader034.fdocuments.us/reader034/viewer/2022050122/5f527bcb42705b0f1913c179/html5/thumbnails/27.jpg)
V1
V5
SPC
Motion
Photic
Att
Model 1
V1
V5
SPC
Motion
Photic
Att
Model 3
Bayes Factor B13=3.6
Positive
![Page 28: Dynamic Causal Modelshelper.ipam.ucla.edu/publications/mbi2004/mbi2004_3847.pdfV5 SPC Wellcome Department of Imaging Neuroscience, ION, UCL, UK. Contents • Neurodynamic model •](https://reader034.fdocuments.us/reader034/viewer/2022050122/5f527bcb42705b0f1913c179/html5/thumbnails/28.jpg)
V1
V5
SPC
Motion
Photic
Att
Model 1
Motion
Photic
Att
Att
V1
V5
SPC
Model 4
Bayes Factor B14=2.8
Weak
Penny et al. (2004) NeuroImage, Special Issue.
![Page 29: Dynamic Causal Modelshelper.ipam.ucla.edu/publications/mbi2004/mbi2004_3847.pdfV5 SPC Wellcome Department of Imaging Neuroscience, ION, UCL, UK. Contents • Neurodynamic model •](https://reader034.fdocuments.us/reader034/viewer/2022050122/5f527bcb42705b0f1913c179/html5/thumbnails/29.jpg)
Contents• Neurodynamic model
• Hemodynamic model
• Bayesian estimation
• Attention to visual motion
• Single word processing
![Page 30: Dynamic Causal Modelshelper.ipam.ucla.edu/publications/mbi2004/mbi2004_3847.pdfV5 SPC Wellcome Department of Imaging Neuroscience, ION, UCL, UK. Contents • Neurodynamic model •](https://reader034.fdocuments.us/reader034/viewer/2022050122/5f527bcb42705b0f1913c179/html5/thumbnails/30.jpg)
Single word processing
SPM{F}
“dog”“radio”“gate”….
10,15,30,60 and 90words perminute
![Page 31: Dynamic Causal Modelshelper.ipam.ucla.edu/publications/mbi2004/mbi2004_3847.pdfV5 SPC Wellcome Department of Imaging Neuroscience, ION, UCL, UK. Contents • Neurodynamic model •](https://reader034.fdocuments.us/reader034/viewer/2022050122/5f527bcb42705b0f1913c179/html5/thumbnails/31.jpg)
A2
WA
A1
0.92
0.380.47
-0.62
-0.51
0.37
Input 1:Word Presentation
Input 1
Estimated Model
Intrinsic connections: Full connectivityassumed – only significant connectionsshown.
Modulatory connections: modulation of all self-connections, only A1 and WA significant
![Page 32: Dynamic Causal Modelshelper.ipam.ucla.edu/publications/mbi2004/mbi2004_3847.pdfV5 SPC Wellcome Department of Imaging Neuroscience, ION, UCL, UK. Contents • Neurodynamic model •](https://reader034.fdocuments.us/reader034/viewer/2022050122/5f527bcb42705b0f1913c179/html5/thumbnails/32.jpg)
Hemodynamic saturation in A1
Seconds
y
Summed individual responses
Responseto pair
IndividualStimuli
Pair ofStimuli
![Page 33: Dynamic Causal Modelshelper.ipam.ucla.edu/publications/mbi2004/mbi2004_3847.pdfV5 SPC Wellcome Department of Imaging Neuroscience, ION, UCL, UK. Contents • Neurodynamic model •](https://reader034.fdocuments.us/reader034/viewer/2022050122/5f527bcb42705b0f1913c179/html5/thumbnails/33.jpg)
Neuronal Saturation in A1
Seconds
zt modulationof A1 selfconnection
Withor
without
A1
-0.62
![Page 34: Dynamic Causal Modelshelper.ipam.ucla.edu/publications/mbi2004/mbi2004_3847.pdfV5 SPC Wellcome Department of Imaging Neuroscience, ION, UCL, UK. Contents • Neurodynamic model •](https://reader034.fdocuments.us/reader034/viewer/2022050122/5f527bcb42705b0f1913c179/html5/thumbnails/34.jpg)
Identifiability
−−
−=
11
1
3231
2321
1312
aaaaaa
A τ
Estimation of relative intrinsic connections and modulatoryconnections is robust to errors inestimation of hemodynamics due toeg. slice timing problems
Indeterminacy in neurodynamic and hemodynamic time constants is soaked up in τ
u2
z1
z2
u1
-τc
a12
a21
b21
-τ
![Page 35: Dynamic Causal Modelshelper.ipam.ucla.edu/publications/mbi2004/mbi2004_3847.pdfV5 SPC Wellcome Department of Imaging Neuroscience, ION, UCL, UK. Contents • Neurodynamic model •](https://reader034.fdocuments.us/reader034/viewer/2022050122/5f527bcb42705b0f1913c179/html5/thumbnails/35.jpg)
Summary
• Neurodynamic model• Hemodynamic model• Bayesian estimation• Attention to visual motion• Single word processing
![Page 36: Dynamic Causal Modelshelper.ipam.ucla.edu/publications/mbi2004/mbi2004_3847.pdfV5 SPC Wellcome Department of Imaging Neuroscience, ION, UCL, UK. Contents • Neurodynamic model •](https://reader034.fdocuments.us/reader034/viewer/2022050122/5f527bcb42705b0f1913c179/html5/thumbnails/36.jpg)
Neuronal Transients and BOLD
300ms 500ms
More enduring transients produce bigger BOLD signals
SecondsSeconds
![Page 37: Dynamic Causal Modelshelper.ipam.ucla.edu/publications/mbi2004/mbi2004_3847.pdfV5 SPC Wellcome Department of Imaging Neuroscience, ION, UCL, UK. Contents • Neurodynamic model •](https://reader034.fdocuments.us/reader034/viewer/2022050122/5f527bcb42705b0f1913c179/html5/thumbnails/37.jpg)
BOLD is sensitive to frequencycontent of transients
Seconds
SecondsRelative timings of transients areamplified in BOLD
Neurodynamics Hemodynamics