|
DD_models
Neurobiological models
Dynamic Causal Modelling of fMRI dataDecomposing the relation existing between cognitive functions and their neurobiological "signature" (the spatio-temporal properties of brain activity) requires an understanding of how information is transmitted through brain networks. The ambition here is to ask questions such as: "what is the nature of the information that region A passes on to region B"? This stems from the notion of functional integration, which views function as an emergent property of brain networks. Dynamic causal modelling –DCM- has been specifically developed to address this question. Dynamic Causal Modelling or DCM embraces a graph-theoretic perspective on brain networks, whereby functionally segregated sources (i.e. brain regions or neuronal populations) correspond to “nodes” and conditional dependencies among the hidden states of each node are mediated by effective connectivity (directed “edges”). DCM generative models are causal in at least two senses:
The inversion of such models given neuroimaging data can then be used to identify the structure of brain networks and their specific modulation by the experimental manipulation (i.e. induced plasticity). For example, showing that a given connection is modulated by the saliency of some stimulus demonstrates that this connection conveys the saliency information. SPM (http://www.fil.ion.ucl.ac.uk/spm/) proposed the seminal software implementations of DCM. The following script allows one to invert an SPM-specified DCM model using VBA, and then eyeball inversion results and diagnostics: [y,u,f_fname,g_fname,dim,options] = dcm2vba(DCM); [posterior,out] = VBA_NLStateSpaceModel(y,u,f_fname,g_fname,dim,options); DCM = vba2dcm(posterior,out,[],TR); spm_dcm_explore(DCM) where DCM is the variable saved in the SPM DCM-file and TR is the fMRI repetition time.
A number of demonstration scripts have been written for DCM:
Neural fieldsThese models are inspired from statistical physics approaches based upon the notion of mean field, i.e. the idea that interactions within micro-scale ensembles of neurons can be captured by summary statistics (i.e., moments of the relevant distribution). They describe the spatio-temporal response of brain networks to experimental manipulations. At this macroscopic scale, neural states like mean membrane depolarisation can be regarded as a continuum or field, which is a function of space and time. The spatiotemporal dynamics of neural fields essentially depend on how local ensembles influence each other, through connectivity kernels. The latter quantify the amount of anatomical connections as a function of physical distance between any two points on the field. We refer the interested reader to the demonstration script demo_2DneuralField.m.
Spiking neuron modelsThis class of models attempt to capture how individual neurons respond to inputs, typically in terms of ion currents that flow through the cell membrane (this occurs when neurotransmitters cause an activation of ion channels in the cell).
| |