1. Models of effective connectivity & Dynamic Causal Modelling (DCM) Klaas Enno Stephan Laboratory for Social & Neural Systems Research Institute for Empirical Research in Economics University of Zurich Functional Imaging Laboratory (FIL) Wellcome Trust Centre for Neuroimaging University College London Methods & Models for fMRI data analysis in neuroeconomics November 2010

2. Overview Brain connectivity: types & definitions –anatomical connectivity –functional connectivity –effective connectivity Psycho-physiological interactions (PPI) Dynamic causal models (DCMs) –DCM for fMRI: Neural and hemodynamic levels –Parameter estimation & inference Applications of DCM to fMRI data –Design of experiments and models –Some empirical examples and simulations

3. Connectivity A central property of any system Communication systems Social networks (internet) (Canberra, Australia) FIgs. by Stephen Eick and A. Klovdahl; see http://www.nd.edu/~networks/gallery.htm

4. Structural, functional & effective connectivity anatomical/structural connectivity = presence of axonal connections functional connectivity =statistical dependencies between regional time series effective connectivity =causal (directed) influences between neurons or neuronal populations Sporns 2007, Scholarpedia

5. Anatomical connectivity neuronal communication via synaptic contacts visualisation by tracing techniques long-range axons  “association fibres”

6. Diffusion-weighted imaging Parker & Alexander, 2005, Phil. Trans. B

7. Parker, Stephan et al. 2002, NeuroImage Diffusion-weighted imaging of the cortico- spinal tract

8. Why would complete knowledge of anatomical connectivity not be enough to understand how the brain works?

9. Connections are recruited in a context-dependent fashion Synaptic strengths are context-sensitive: They depend on spatio-temporal patterns of network activity.

10. Connections show plasticity critical for learning can occur both rapidly and slowly NMDA receptors play a critical role NMDA receptors are regulated by modulatory neurotransmitters like dopamine, serotonine, acetylcholine synaptic plasticity =change in the structure and transmission properties of a chemical synapse Gu 2002, Neuroscience NMDA receptor

11. Short-term SP & neuromodulation NMDAR-independent –synaptic depression/facilitation –effects due to dendritic backpropagation & voltage- sensitive ion channels NMDAR-dependent –phosphorylation of AMPARs –modulation of EPSPs at NMDARs by DA, ACh, 5HT (gating) through phosphorylation NMDAR trafficking changes in membrane potential Reynolds et al. 2001, Nature Tsodyks & Markram 1997, PNAS peak PSP (mV)

12. Different approaches to analysing functional connectivity Seed voxel correlation analysis Eigen-decomposition (PCA, SVD) Independent component analysis (ICA) any other technique describing statistical dependencies amongst regional time series

13. Seed-voxel correlation analyses Very simple idea: –hypothesis-driven choice of a seed voxel → extract reference time series –voxel-wise correlation with time series from all other voxels in the brain seed voxel

14. Drug-induced changes in functional connectivity Finger-tapping task in first-episode schizophrenic patients: voxels that showed changes in functional connectivity (p<0.005) with the left ant. cerebellum after medication with olanzapine Stephan et al. 2001, Psychol. Med.

15. Does functional connectivity not simply correspond to co-activation in SPMs? No, it does not - see the fictitious example on the right: Here both areas A 1 and A 2 are correlated identically to task T, yet they have zero correlation among themselves: r(A 1,T) = r(A 2,T) = 0.71 but r(A 1,A 2 ) = 0 ! task T regional response A 2 regional response A 1 Stephan 2004, J. Anat.

16. Pros & Cons of functional connectivity analyses Pros: –useful when we have no experimental control over the system of interest and no model of what caused the data (e.g. sleep, hallucinations, etc.) Cons: –interpretation of resulting patterns is difficult / arbitrary –no mechanistic insight into the neural system of interest –usually suboptimal for situations where we have a priori knowledge and experimental control about the system of interest models of effective connectivity necessary

17. For understanding brain function mechanistically, we need models of effective connectivity, i.e. models of causal interactions among neuronal populations.

18. Some models for computing effective connectivity from fMRI data Structural Equation Modelling (SEM) McIntosh et al. 1991, 1994; Büchel & Friston 1997; Bullmore et al. 2000 regression models (e.g. psycho-physiological interactions, PPIs) Friston et al. 1997 Volterra kernels Friston & Büchel 2000 Time series models (e.g. MAR, Granger causality) Harrison et al. 2003, Goebel et al. 2003 Dynamic Causal Modelling (DCM) bilinear: Friston et al. 2003; nonlinear: Stephan et al. 2008

19. Overview Brain connectivity: types & definitions –anatomical connectivity –functional connectivity –effective connectivity Psycho-physiological interactions (PPI) Dynamic causal models (DCMs) –DCM for fMRI: Neural and hemodynamic levels –Parameter estimation & inference Applications of DCM to fMRI data –Design of experiments and models –Some empirical examples and simulations

20. Psycho-physiological interaction (PPI) We can replace one main effect in the GLM by the time series of an area that shows this main effect. E.g. let's replace the main effect of stimulus type by the time series of area V1: Task factor Task A Task B Stim 1 Stim 2 Stimulus factor T A /S 1 T B /S 1 T A /S 2 T B /S 2 GLM of a 2x2 factorial design: main effect of task main effect of stim. type interaction main effect of task V1 time series  main effect of stim. type psycho- physiological interaction Friston et al. 1997, NeuroImage

21. V1 attention no attention V1 activity V5 activity SPM{Z} time V5 activity Friston et al. 1997, NeuroImage Büchel & Friston 1997, Cereb. Cortex V1 x Att. = V5 V5 Attention Attentional modulation of V1→V5

22. PPI: interpretation Two possible interpretations of the PPI term: V1 Modulation of V1  V5 by attention Modulation of the impact of attention on V5 by V1 V1V5 V1 V5 attention V1 attention

23. Two PPI variants "Classical" PPI: –Friston et al. 1997, NeuroImage –depends on factorial design –in the GLM, physiological time series replaces one experimental factor –physio-physiological interactions: two experimental factors are replaced by physiological time series Alternative PPI: –Macaluso et al. 2000, Science –interaction term is added to an existing GLM –can be used with any design

24. Is the red letter left or right from the midline of the word? group analysis (random effects), n=16, p<0.05 corrected analysis with SPM2 group analysis (random effects), n=16, p<0.05 corrected analysis with SPM2 Task-driven lateralisation letter decisions > spatial decisions time Does the word contain the letter A or not? spatial decisions > letter decisions Stephan et al. 2003, Science

25. Bilateral ACC activation in both tasks – but asymmetric connectivity ! IPS IFG Left ACC  left inf. frontal gyrus (IFG): increase during letter decisions. Right ACC  right IPS: increase during spatial decisions. left ACC (-6, 16, 42) right ACC (8, 16, 48) spatial vs letter decisions letter vs spatial decisions group analysis random effects (n=15) p<0.05, corrected (SVC) Stephan et al. 2003, Science

26. PPI single-subject example b VS = -0.16 b L =0.63 Signal in left ACC Signal in left IFG b L = -0.19 Signal in right ant. IPS Signal in right ACC b VS =0.50 Left ACC signal plotted against left IFG spatial decisions letter decisions spatial decisions Right ACC signal plotted against right IPS Stephan et al. 2003, Science

27. PPI for event-related fMRI requires deconvolution Gitelman et al. 2003, NeuroImage (A  HRF)  (B  HRF)   (A  B)  HRF

28. Pros & Cons of PPIs Pros: –given a single source region, we can test for its context-dependent connectivity across the entire brain –easy to implement Cons: –very simplistic model: only allows to model contributions from a single area –ignores time-series properties of data –application to event-related data relies deconvolution procedure (Gitelman et al. 2003, NeuroImage) –operates at the level of BOLD time series sometimes very useful, but limited causal interpretability; in most cases, we need more powerful models

29. Overview Brain connectivity: types & definitions –anatomical connectivity –functional connectivity –effective connectivity Psycho-physiological interactions (PPI) Dynamic causal models (DCMs) –DCM for fMRI: Neural and hemodynamic levels –Parameter estimation & inference Applications of DCM to fMRI data –Design of experiments and models –Some empirical examples and simulations

30. Dynamic causal modelling (DCM) DCM framework was introduced in 2003 for fMRI by Karl Friston, Lee Harrison and Will Penny (NeuroImage 19:1273-1302) part of the SPM software package currently more than 100 published papers on DCM

31. Neural state equation: Electromagnetic forward model: neural activity  EEG MEG LFP Dynamic Causal Modeling (DCM) simple neuronal model complicated forward model complicated neuronal model simple forward model fMRI EEG/MEG inputs Hemodynamic forward model: neural activity  BOLD Stephan & Friston 2007, Handbook of Brain Connectivity

32. Invert model Make inferences Define likelihood model Specify priors Neural dynamics Observer function Design experimental inputs Inference on model structure Inference on parameters Bayesian inversion

33. LG left LG right RVFLVF FG right FG left LG = lingual gyrus FG = fusiform gyrus Visual input in the - left (LVF) - right (RVF) visual field. x1x1 x2x2 x4x4 x3x3 u2u2 u1u1 Example: a linear system of dynamics in visual cortex

34. LG = lingual gyrus FG = fusiform gyrus Visual input in the - left (LVF) - right (RVF) visual field. state changes effective connectivity external inputs system state input parameters LG left LG right RVFLVF FG right FG left x1x1 x2x2 x4x4 x3x3 u2u2 u1u1

35. Extension: bilinear dynamic system LG left LG right RVFLVF FG right FG left x1x1 x2x2 x4x4 x3x3 u2u2 u1u1 CONTEXT u3u3

36. endogenous connectivity direct inputs modulation of connectivity Neural state equation hemodynamic model λ x y integration BOLD yy y activity x 1 (t) activity x 2 (t) activity x 3 (t) neuronal states t driving input u 1 (t) modulatory input u 2 (t) t Stephan & Friston (2007), Handbook of Brain Connectivity   

37. Bilinear DCM Bilinear state equation: driving input modulation Two-dimensional Taylor series (around x 0 =0, u 0 =0):

38. - x2x2 stimuli u 1 context u 2 x1x1 + + - - - + u 1 Z 1 u 2 Z 2 Example: context-dependent decay u1u1 u2u2 x2x2 x1x1 Penny, Stephan, Mechelli, Friston NeuroImage (2004)

39. DCM parameters = rate constants The coupling parameter a thus describes the speed of the exponential change in x(t) Integration of a first-order linear differential equation gives an exponential function: Coupling parameter a is inversely proportional to the half life  of z(t):

40. The problem of hemodynamic convolution Goebel et al. 2003, Magn. Res. Med.

41. Hemodynamic forward models are important for connectivity analyses of fMRI data David et al. 2008, PLoS Biol. Granger causality DCM

42. stimulus functions u t neural state equation hemodynamic state equations Balloon model BOLD signal change equation important for model fitting, but of no interest for statistical inference 6 hemodynamic parameters: Computed separately for each area (like the neural parameters)  region-specific HRFs! The hemodynamic model in DCM Friston et al. 2000, NeuroImage Stephan et al. 2007, NeuroImage

43. stimulus functions u t neural state equation hemodynamic state equations Balloon model BOLD signal change equation The hemodynamic model in DCM Stephan et al. 2007, NeuroImage

44. A B C hh ε How interdependent are neural and hemodynamic parameter estimates? Stephan et al. 2007, NeuroImage

45. Bayesian statistics posterior  likelihood ∙ prior Bayes theorem allows one to formally incorporate prior knowledge into computing statistical probabilities. In DCM: empirical, principled & shrinkage priors. The “posterior” probability of the parameters given the data is an optimal combination of prior knowledge and new data, weighted by their relative precision. new data prior knowledge

46. stimulus function u modelled BOLD response observation model hidden states state equation parameters Combining the neural and hemodynamic states gives the complete forward model. An observation model includes measurement error e and confounds X (e.g. drift). Bayesian parameter estimation by means of a Levenberg-Marquardt gradient ascent, embedded into an EM algorithm. Result: Gaussian a posteriori parameter distributions, characterised by mean η θ|y and covariance C θ|y. Overview: parameter estimation η θ|y neural state equation

47. Gaussian assumptions about the posterior distributions of the parameters Use of the cumulative normal distribution to test the probability that a certain parameter (or contrast of parameters c T η θ|y ) is above a chosen threshold γ: By default, γ is chosen as zero ("does the effect exist?"). Inference about DCM parameters: Bayesian single-subject analysis

48. Bayesian single subject inference LG left LG right RVF stim. LVF stim. FG right FG left LD|RVF LD|LVF LD 0.34  0.14 -0.08  0.16 0.13  0.19 0.01  0.17 0.44  0.14 0.29  0.14 Contrast: Modulation LG right  LG links by LD|LVF vs. modulation LG left  LG right by LD|RVF p(c T  >0|y) = 98.7% Stephan et al. 2005, Ann. N.Y. Acad. Sci.

49. Likelihood distributions from different subjects are independent  one can use the posterior from one subject as the prior for the next Under Gaussian assumptions this is easy to compute: group posterior covariance individual posterior covariances group posterior mean individual posterior covariances and means “Today’s posterior is tomorrow’s prior” Inference about DCM parameters: Bayesian fixed-effects group analysis

50. Inference about DCM parameters: group analysis (classical) In analogy to “random effects” analyses in SPM, 2 nd level analyses can be applied to DCM parameters: Separate fitting of identical models for each subject Selection of bilinear parameters of interest one-sample t-test: parameter > 0 ? paired t-test: parameter 1 > parameter 2 ? rmANOVA: e.g. in case of multiple sessions per subject

51. Overview Brain connectivity: types & definitions –anatomical connectivity –functional connectivity –effective connectivity Psycho-physiological interactions (PPI) Dynamic causal models (DCMs) –DCM for fMRI: Neural and hemodynamic levels –Parameter estimation & inference Applications of DCM to fMRI data –Design of experiments and models –Some empirical examples and simulations

52. Any design that is good for a GLM of fMRI data. What type of design is good for DCM?

53. GLM vs. DCM DCM tries to model the same phenomena as a GLM, just in a different way: It is a model, based on connectivity and its modulation, for explaining experimentally controlled variance in local responses. No activation detected by a GLM → inclusion of this region in a DCM is useless! Stephan 2004, J. Anat.

54. Multifactorial design: explaining interactions with DCM Task factor Task A Task B Stim 1 Stim 2 Stimulus factor T A /S 1 T B /S 1 T A /S 2 T B /S 2 X1X1 X2X2 Stim2/ Task A Stim1/ Task A Stim 1/ Task B Stim 2/ Task B GLM X1X1 X2X2 Stim2 Stim1 Task ATask B DCM Let’s assume that an SPM analysis shows a main effect of stimulus in X 1 and a stimulus  task interaction in X 2. How do we model this using DCM?

55. Stim 1 Task A Stim 2 Task A Stim 1 Task B Stim 2 Task B Simulated data X1X1 X2X2 +++ X1X1 X2X2 Stimulus 2 Stimulus 1 Task ATask B ++++ + + –– Stephan et al. 2007, J. Biosci.

56. Stim 1 Task A Stim 2 Task A Stim 1 Task B Stim 2 Task B plus added noise (SNR=1) X1X1 X2X2

57. Example studies of DCM for fMRI DCM now an established tool for fMRI & M/EEG analysis >140 studies published combinations of DCM with computational models

58. Is the red letter left or right from the midline of the word? group analysis (random effects), n=16, p<0.05 corrected analysis with SPM2 group analysis (random effects), n=16, p<0.05 corrected analysis with SPM2 Task-driven lateralisation letter decisions > spatial decisions time Does the word contain the letter A or not? spatial decisions > letter decisions Stephan et al. 2003, Science

59. Theories on inter-hemispheric integration during lateralised tasks Information transfer (for left-lateralised task) Inhibition/CompetitionHemispheric recruitment LVFRVF T T T T + − − T T + + Predictions: modulation by task conditional on visual field asymmetric connection strengths Predictions: modulation by task only negative & symmetric connection strengths Predictions: modulation by task only positive & symmetric connection strengths  |LVF |RVF

60. LG left LG right FG right FG left RVFLVF B A B cond B ind LD VF LDB ind B cond intra inter 16 models LG left LG right FG right FG left LD RVF LVF LG left LG right RVF stim. LVF stim. FG right FG left LD LD,RVF LD|RVF LD LD,LVF LD|LVF VF LD B ind B cond LD RVF LVF LD|RVF LD|LVF VFLDB ind B cond D C

61. LG left LG right RVF stim. LVF stim. FG right FG left LD|RVF LD|LVF LD 0.25  0.04 0.03  0.03 0.12  0.02 0.02  0.02 0.36  0.06 0.16  0.05 Left lingual gyrus (LG) -12,-64,-4 Left fusiform gyrus (FG) -44,-52,-18 Right fusiform gyrus (FG) 38,-52,-20 Right lingual gyrus (LG) 14,-68,-2 mean parameter estimates  SE (n=12) significant modulation (p<0.05, uncorrected) non-significant modulation significant modulation (p<0.05, Bonferroni-corrected) LD>SD masked incl. with RVF>LVF p<0.05 cluster-level corrected (p<0.001 voxel-level cut-off) LD>SD, p<0.05 cluster-level corrected (p<0.001 voxel-level cut-off) p<0.01 uncorrected LD>SD masked incl. with LVF>RVF p<0.05 cluster-level corrected (p<0.001 voxel-level cut-off) Ventral stream & letter decisions Stephan et al. 2007, J. Neurosci.

62. MOG left LG left LG right RVF stim. LVF stim. FG right FG left LD|RVF LD|LVF LD 0.20  0.04 0.06  0.02 0.00  0.01 0.01  0.01 0.27  0.06 0.11  0.03 MOG right 0.00  0.04 0.01  0.03 0.07  0.02 0.01  0.01 Ventral stream & letter decisions LD>SD, p<0.05 cluster-level corrected (p<0.001 voxel-level cut-off) Left MOG -38,-90,-4 mean parameter estimates  SE (n=12) significant modulation (p<0.05, uncorrected) non-significant significant modulation (p<0.05, corrected) Left FG -44,-52,-18 Right MOG -38,-94,0 p<0.01 uncorrected Left LG -12,-70,-6 Left LG -14,-68,-2 LD>SD masked incl. with RVF>LVF p<0.05 cluster-level corrected (p<0.001 voxel-level cut-off) LD>SD masked incl. with LVF>RVF p<0.05 cluster-level corrected (p<0.001 voxel-level cut-off) Right FG 38,-52,-20 Stephan et al. 2007, J. Neurosci.

63. Asymmetric modulation of LG callosal connections is consistent across subjects Stephan et al. 2007, J. Neurosci.

64. Thank you