Bayesian Inference in SPM2 Will Penny K. Friston, J. Ashburner, J.-B. Poline, R. Henson, S. Kiebel, D. Glaser Wellcome Department of Imaging Neuroscience, University College London, UK
SPM99 RealignmentSmoothing Normalisation General linear model Statistical parametric map (SPM) fMRI time-series Parameter estimates Design matrix Template Kernel p <0.05 Inference with Gaussian field theory Adjusted regional data spatial modes and effective connectivity
What’s new in SPM2 ? §Spatial transformation of images §Batch Mode §Modelling and Inference Expectation-Maximisation (EM) Restricted Maximum Likelihood (ReML) Parametric Empirical Bayes (PEB)
Hierarchical model Single-level model Parametric Empirical Bayes (PEB) Restricted Maximimum Likelihood (ReML) Hierarchical models
Bayes Rule
Example 2:Univariate model Likelihood and Prior Posterior Relative Precision Weighting
Example 2:Multivariate two-level model Likelihood and Prior Assume diagonal precisions PosteriorPrecisions Data-determined parameters Assume Shrinkage Prior
General Case: Arbitrary Error Covariances E-Step yCXC XCXC T yy T y M-Step y Xyr for i and j { }{ }{}{ CQCQtrJ XCQCXC rCQCrCQ g ijij i T y i T ii } kk QCC gJ 1 Friston, K. et al. (2002), Neuroimage EM algorithm
Pooling assumption Decompose error covariance at each voxel, i, into a voxel specific term, r(i), and voxel-wide terms.
What’s new in SPM2 ? §Corrections for Non-Sphericity §Posterior Probability Maps (PPMs) §Haemodynamic modelling §Dynamic Causal Modelling (DCM)
Non-sphericity §Relax assumption that errors are Independent and Identically Distributed (IID) §Non-independent errors eg. repeated measures within subject §Non-identical errors eg. unequal condition/subject error variances §Correlation in fMRI time series §Allows multiple parameters at 2 nd level ie. RFX
Single-subject contrasts from Group FFX PET Verbal Fluency SPMs,p<0.001 uncorrected SphericityNon-sphericity Non-identical error variances
Correlation in fMRI time series Model errors for each subject as AR(1) + white noise.
The Interface OLS Parameters, REML Hyperparameters PEB Parameters and Hyperparameters No Priors Shrinkage priors
Bayesian estimation: Two-level model 1 st level = within-voxel 2nd level = between-voxels Likelihood Shrinkage Prior In the absence of evidence to the contrary parameters will shrink to zero
LikelihoodPrior Posterior SPMs PPMs Bayesian Inference: Posterior Probability Maps
SPMs and PPMs PPMs: Show activations of a given size SPMs: show voxels with non-zero activations
PPMs AdvantagesDisadvantages One can infer a cause DID NOT elicit a response SPMs conflate effect-size and effect-variability No multiple comparisons problem (hence no smoothing) P-values don’t change with search volume Use of Shrinkage priors over voxels is computationally demanding Utility of Bayesian approach is yet to be established
The Interface Hemodynamic Modelling
The hemodynamic model
Hemodynamics
Inference with MISO models FUNCTIONAL SEGREGATION: This voxel IS NOT responsive to attention
The Interface Dynamic Causal Modelling
hemodynamics response y(t)= (X) response y(t)= (X) hemodynamics response y(t)= (X) hemodynamics Hemodynamic model Extension to a MIMO system Input u(t) activity x 1 (t) activity x 3 (t) activity x 2 (t) The bilinear model neuronal changes intrinsic connectivity induced response induced connectivity
V1 V4 BA37 STG BA39 Cognitive set - u 2 (t) {e.g. semantic processing} Stimuli - u 1 (t) {e.g. visual words} Dynamical Causal Models Functional integration and the modulation of specific pathways
Summary §SPM2 – modelling and inference §Hierarchical Models and EM §Corrections for Non-Sphericity §Posterior Probability Maps (PPMs) §Haemodynamic modelling §Dynamic Causal Modelling (DCM)