Presentation on theme: "DCM demo André Bastos and Martin Dietz Wellcome Trust Centre for Neuroimaging."— Presentation transcript:
DCM demo André Bastos and Martin Dietz Wellcome Trust Centre for Neuroimaging
pseudo-random auditory sequence 80% standard tones – 500 Hz 20% deviant tones – 550 Hz time standardsdeviants Mismatch negativity (MMN) paradigm and hypothesis time (ms) Paradigm: amplitude (μV) 0 200 - + 0 Deviant ERP Standard ERP Hypothesis: MMN is caused by recurrent dynamics enabled by backward connections Garrido et al., Neuroimage 2007
Mismatch Negativity scalp topography of ERPs time (ms) sensors standard deviant Deviant ERP Standard ERP time (ms) amplitude (μV) 0 200 - + 0 Hypothesis: MMN is caused by recurrent dynamics enabled by backward connections 100200 0
The generative model Source dynamics f states x parameters θ Input u Evoked response data y Observation model g David et al., Neuroimage 2006; Kiebel et al., Neuroimage 2006
DCM specification A1 STG input STG IFG several plausible models… modulation of effective connectivity Forward - F Backward - B Both - FB 1 2 3 4 5 Garrido et al., Neuroimage 2007 Opitz et al., 2002 lSTG rSTG rIFG Deviant response Standard response time (ms) amplitude (μV) 0 200 - + 0 What set of areas and interconnections caused the MMN?
A1 STG Forward Backward Lateral STG input A1 STG Forward Backward Lateral input A1 STG Forward Backward Lateral input Forward-FBackward-B Forward and Backward-FB STG IFG DCM specification of different models modulation of effective connectivity Garrido et al., Neuroimage 2007
Analysis steps 0. Have a HYPOTHESIS! 1.Preprocessing and SVD decomposition 2.Model specification: specify cortical areas and inter- areal connections for various competing models that you think might explain your data 3.Model inversion: find the parameters that minimize differences between observed measurements and model predictions for each of the competing models 4.Bayesian model comparison: make a statistical inferences about which model best describes the data