Bilinear Dynamical Systems

Slides:



Advertisements
Similar presentations
Dynamic Causal Modelling (DCM) for fMRI
Advertisements

Bayesian fMRI models with Spatial Priors Will Penny (1), Nelson Trujillo-Barreto (2) Guillaume Flandin (1) Stefan Kiebel(1), Karl Friston (1) (1) Wellcome.
Bayesian inference Lee Harrison York Neuroimaging Centre 01 / 05 / 2009.
Hierarchical Models and
DCM: Dynamic Causal Modelling for fMRI
GUIDE to The… D U M M I E S’ DCM Velia Cardin. Functional Specialization is a question of Where? Where in the brain is a certain cognitive/perceptual.
Dynamic Causal Modelling THEORY SPM Course FIL, London October 2009 Hanneke den Ouden Donders Centre for Cognitive Neuroimaging Radboud University.
Neural Decoding: Model and Algorithm for Evidence Accumulator Inference Thomas Desautels University College London Gatsby Computational Neuroscience Group.
Dynamic Causal Modelling
Measuring Functional Integration: Connectivity Analyses
DCM Advanced, Part II Will Penny (Klaas Stephan) Wellcome Trust Centre for Neuroimaging Institute of Neurology University College London SPM Course 2014.
Dynamic Causal Modelling (DCM) for fMRI
Dynamic Causal Modelling (DCM): Theory Demis Hassabis & Hanneke den Ouden Thanks to Klaas Enno Stephan Functional Imaging Lab Wellcome Dept. of Imaging.
Dynamic Causal Modelling (DCM) for fMRI
18 th February 2009 Stephanie Burnett Christian Lambert Methods for Dummies 2009 Dynamic Causal Modelling Part I: Theory.
Dynamic Causal Modelling for fMRI Friday 22 nd Oct SPM fMRI course Wellcome Trust Centre for Neuroimaging London André Marreiros.
Dynamic Causal Modelling for fMRI Justin Grace Marie-Hélène Boudrias Methods for Dummies 2010.
Bayesian Modelling of Functional Imaging Data Will Penny The Wellcome Department of Imaging Neuroscience, UCL http//:
Bayesian Inference and Posterior Probability Maps Guillaume Flandin Wellcome Department of Imaging Neuroscience, University College London, UK SPM Course,
Dynamic Causal Modelling (DCM) Functional Imaging Lab Wellcome Dept. of Imaging Neuroscience Institute of Neurology University College London Presented.
Dynamic Causal Modelling Will Penny Wellcome Department of Imaging Neuroscience, University College London, UK FMRIB, Oxford, May
J. Daunizeau ICM, Paris, France ETH, Zurich, Switzerland Dynamic Causal Modelling of fMRI timeseries.
DCM – the theory. Bayseian inference DCM examples Choosing the best model Group analysis.
1 Design Matrix intrinsic correlations and non-sphericity correction intrinsic correlations and non-sphericity correction filter Data SPM{T} The General.
Dynamic Causal Modelling Advanced Topics SPM Course (fMRI), May 2015 Peter Zeidman Wellcome Trust Centre for Neuroimaging University College London.
SPM short course Functional integration and connectivity Christian Büchel Karl Friston The Wellcome Department of Cognitive Neurology, UCL London UK http//:
Dynamic Causal Modelling (DCM) Marta I. Garrido Thanks to: Karl J. Friston, Klaas E. Stephan, Andre C. Marreiros, Stefan J. Kiebel,
Dynamic Causal Modelling Introduction SPM Course (fMRI), October 2015 Peter Zeidman Wellcome Trust Centre for Neuroimaging University College London.
The world before DCM. Linear regression models of connectivity Structural equation modelling (SEM) y1y1 y3y3 y2y2 b 12 b 32 b 13 z1z1 z2z2 z3z3 0 b 12.
Ch. 5 Bayesian Treatment of Neuroimaging Data Will Penny and Karl Friston Ch. 5 Bayesian Treatment of Neuroimaging Data Will Penny and Karl Friston 18.
Dynamic Causal Model for evoked responses in MEG/EEG Rosalyn Moran.
Multimodal Brain Imaging Wellcome Trust Centre for Neuroimaging, University College, London Guillaume Flandin, CEA, Paris Nelson Trujillo-Barreto, CNC,
Bayesian Methods Will Penny and Guillaume Flandin Wellcome Department of Imaging Neuroscience, University College London, UK SPM Course, London, May 12.
Bayesian inference Lee Harrison York Neuroimaging Centre 23 / 10 / 2009.
Dynamic Causal Models Will Penny Olivier David, Karl Friston, Lee Harrison, Andrea Mechelli, Klaas Stephan Mathematics in Brain Imaging, IPAM, UCLA, USA,
Bayesian Inference in fMRI Will Penny Bayesian Approaches in Neuroscience Karolinska Institutet, Stockholm February 2016.
Bayesian Inference in SPM2 Will Penny K. Friston, J. Ashburner, J.-B. Poline, R. Henson, S. Kiebel, D. Glaser Wellcome Department of Imaging Neuroscience,
Bayesian selection of dynamic causal models for fMRI Will Penny Olivier David, Karl Friston, Lee Harrison, Andrea Mechelli, Klaas Stephan The brain as.
Dynamic Causal Models Will Penny Olivier David, Karl Friston, Lee Harrison, Stefan Kiebel, Andrea Mechelli, Klaas Stephan MultiModal Brain Imaging, Copenhagen,
Bayesian Model Comparison Will Penny Wellcome Department of Imaging Neuroscience, University College London, UK UCL, June 7, 2003 Large-Scale Neural Network.
Tijl De Bie John Shawe-Taylor ECS, ISIS, University of Southampton
5th March 2008 Andreina Mendez Stephanie Burnett
Dynamic Causal Modeling of Endogenous Fluctuations
Variational filtering in generated coordinates of motion
Variational Bayesian Inference for fMRI time series
Effective Connectivity: Basics
Bayesian Inference Will Penny
Effective Connectivity
Dynamic Causal Modelling (DCM): Theory
Neuroscience Research Institute University of Manchester
Dynamic Causal Model for evoked responses in M/EEG Rosalyn Moran.
DCM: Advanced issues Klaas Enno Stephan Laboratory for Social & Neural Systems Research Institute for Empirical Research in Economics University of.
The General Linear Model
Dynamic Causal Modelling
Dynamic Causal Modelling
SPM2: Modelling and Inference
Dynamic Causal Modelling for M/EEG
Dynamic Causal Modelling
Bayesian Methods in Brain Imaging
The General Linear Model
CRIS Workshop: Computational Neuroscience and Bayesian Modelling
Hierarchical Models and
Effective Connectivity
Wellcome Centre for Neuroimaging at UCL
Bayesian Inference in SPM2
Wellcome Centre for Neuroimaging, UCL, UK.
The General Linear Model
The General Linear Model
Group DCM analysis for cognitive & clinical studies
The General Linear Model
Presentation transcript:

Bilinear Dynamical Systems A unified framework for fMRI deconvolution, system identification and connectivity analysis Will Penny, Zoubin Ghahramani, Karl Friston Brain Connectivity Workshop, April 2004, Havana, Cuba

BDS Linear Hemodynamics Bilinear Stochastic Neurodynamics (from GLM) basis functions BDS Lagged Neuronal Activity Region-dependent basis coefficients Observation Noise Linear Hemodynamics (from GLM) Bilinear Stochastic Neurodynamics (from DCM) Driving inputs Intrinsic connections State Noise Modulatory connections Deconvolution: Estimation of st System Identification: Estimation of q={b,A,Bm,D} Connectivity Analysis: Estimation of A, Bm

f1 f2 f3 Linear Hemodynamics – via basis functions Canonical, Temporal Derivative, f2 Dispersion Derivative, f3 Seconds

Data from generative model for a single region ut1 ut2 st yt Seconds

Embedding Neuronal Activity st-3 st-2 st st-1 ut-3 ut-2 ut ut-1 yt Xt=[st,st-1,st-1,…,st-L] Deconvolution: Estimation of st Kalman Filtering, p(st|y1,..,yt) Kalman Smoothing, p(st|y1,..,yT) System Identification: Estimation of b,A,Bm,D Connectivity Analysis: Estimation of A, Bm E-Step M-Step Xt-3 Xt-2 Xt Xt-1 ut-3 ut-2 ut ut-1 yt EM for LDS (Ghahramani,1996) EM for BDS (this work) faster than Pseudo-Newton/Simplex methods Priors over model parameters lead to Variational EM (Ghahramani, 2001) Extension to MAR neurodynamics

Example: System Identification True BDS parameters; a=0.72, d=0.88 BDS parameters as estimated by EM; a=0.68, d=0.83 Assumption of deterministic dynamics (wt=0), ML estimates; a=0.45, d=1.13 Single Region

Example: Deconvolution fMRI Gets intrinsic dynamics. Misses evoked responses. Wiener Misses intrinsic dynamics. Gets ‘average’ evoked response. BDS Kalman Filtering Trial-to-trial variability in evoked response due to intrinsic dynamics. BDS Kalman Smoothing

Example: Connectivity (DCMs) SPC Motion Photic Attention 0.86 0.56 -0.02 1.42 0.55 0.75 0.89 V1 V5 SPC Motion Photic Attention 0.96 0.39 0.06 0.58 m=3 V1 V5 SPC Motion Photic Attention 0.85 0.57 -0.02 1.36 0.70 0.84 0.23 Evidence: Bayes factors: