Limits...
LFP and oscillations-what do they tell us?

Friston KJ, Bastos AM, Pinotsis D, Litvak V - Curr. Opin. Neurobiol. (2014)

Bottom Line: In particular, we treat oscillations as the (observable) signature of context-sensitive changes in synaptic efficacy that underlie coordinated dynamics and message-passing in the brain.This rich source of information is now being exploited by various procedures-like dynamic causal modelling-to test hypotheses about neuronal circuits in health and disease.Furthermore, the roles played by neuromodulatory mechanisms can be addressed directly through their effects on oscillatory phenomena.

View Article: PubMed Central - PubMed

Affiliation: The Wellcome Trust Centre for Neuroimaging, University College London, Queen Square, London WC1N 3BG, UK. Electronic address: k.friston@ucl.ac.uk.

Show MeSH

Related in: MedlinePlus

This schematic illustrates the link between the parameters of a dynamic causal model—such as effective connectivity or synaptic efficacy—and the spectral signatures of these coupling parameters. Left panel: state space or dynamic causal model of neuronal states x generating observed data y. The equations at the top represent the equations of motion and (static) observer function generating data. These dynamics are driven by random fluctuations v, where w represents measurement noise. The example shown here is perhaps the simplest; with recurrently and reciprocally (and linearly) coupled excitatory (black) and inhibitory (red) neuronal populations. Right panel: this illustrates the corresponding spectral behaviours expressed in terms of spectral densities. The top equation shows that the observed spectral density g(ω) is a mixture of signal generated by applying transfer functions K(ω) to the spectral density of the random fluctuations (assumed to be the identity matrix here for simplicity) plus a component due to measurement noise. Crucially, the transfer functions and ensuing spectral density are determined by the eigenvalues of the model's connectivity (shown on the lower left). In turn, the eigenvalues are relatively simple functions of the connectivity. The resulting (Lorentzian) spectral density is centred on the imaginary part of the eigenvalue and corresponds to the connection strength of reciprocal connections. The dispersion (full width half maximum) of the spectral peak is determined by the recurrent connectivity. This example shows how connectivity parameters can be expressed directly and intuitively in measured spectra. Furthermore, peristimulus time-dependent changes in the spectral peak disclose stimulus-induced changes in the strength of reciprocal connectivity (i.e. short-term changes in synaptic efficacy of the sort that could be mediated by NMDA receptors)—as illustrated on the lower right. In practice, dynamic causal models are much more complicated than the above example; they usually consider distributed networks of sources with multiple populations within each source and multiple states within each population—with non-linear coupling.
© Copyright Policy - CC BY
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4376394&req=5

fig0005: This schematic illustrates the link between the parameters of a dynamic causal model—such as effective connectivity or synaptic efficacy—and the spectral signatures of these coupling parameters. Left panel: state space or dynamic causal model of neuronal states x generating observed data y. The equations at the top represent the equations of motion and (static) observer function generating data. These dynamics are driven by random fluctuations v, where w represents measurement noise. The example shown here is perhaps the simplest; with recurrently and reciprocally (and linearly) coupled excitatory (black) and inhibitory (red) neuronal populations. Right panel: this illustrates the corresponding spectral behaviours expressed in terms of spectral densities. The top equation shows that the observed spectral density g(ω) is a mixture of signal generated by applying transfer functions K(ω) to the spectral density of the random fluctuations (assumed to be the identity matrix here for simplicity) plus a component due to measurement noise. Crucially, the transfer functions and ensuing spectral density are determined by the eigenvalues of the model's connectivity (shown on the lower left). In turn, the eigenvalues are relatively simple functions of the connectivity. The resulting (Lorentzian) spectral density is centred on the imaginary part of the eigenvalue and corresponds to the connection strength of reciprocal connections. The dispersion (full width half maximum) of the spectral peak is determined by the recurrent connectivity. This example shows how connectivity parameters can be expressed directly and intuitively in measured spectra. Furthermore, peristimulus time-dependent changes in the spectral peak disclose stimulus-induced changes in the strength of reciprocal connectivity (i.e. short-term changes in synaptic efficacy of the sort that could be mediated by NMDA receptors)—as illustrated on the lower right. In practice, dynamic causal models are much more complicated than the above example; they usually consider distributed networks of sources with multiple populations within each source and multiple states within each population—with non-linear coupling.

Mentions: The circumstantial evidence for predictive coding is substantial—both in terms of the anatomy of extrinsic (between-areas) and intrinsic (within-area) connections and the physiology of synaptic interactions [3]. In these schemes, top-down predictions are used to form prediction errors at each level of cortical and subcortical hierarchies. The prediction errors are then returned to the level above to update predictions in a Bayesian sense. In brief, the prediction errors report the ‘newsworthy’ information from a lower hierarchical level that was not predicted by the higher level. A crucial aspect of this message passing is the selection of ascending information by adjusting the ‘volume’ or gain of prediction errors that compete for influence over higher levels of processing. Functionally, this gain corresponds to the expected precision (inverse variance or signal-to-noise ratio) that sets the confidence afforded to prediction errors. Psychologically this has been proposed as the basis of attentional gain [6]. Physiologically, precision corresponds to the postsynaptic gain or sensitivity of cells reporting prediction errors (currently thought to be large principal cells that send extrinsic efferents of a forward type, such as superficial pyramidal cells in cortex). This is important because the synaptic gain or efficacy of coupled neuronal populations determines the form of their spectral (oscillatory) behaviour. See Figure 1. Because, synchronous activity determines synaptic gain [7], oscillations have a mechanistic impact on neuronal processing—rather than being epiphenomenal—which completes the circular causality between synchrony and synaptic efficacy.


LFP and oscillations-what do they tell us?

Friston KJ, Bastos AM, Pinotsis D, Litvak V - Curr. Opin. Neurobiol. (2014)

This schematic illustrates the link between the parameters of a dynamic causal model—such as effective connectivity or synaptic efficacy—and the spectral signatures of these coupling parameters. Left panel: state space or dynamic causal model of neuronal states x generating observed data y. The equations at the top represent the equations of motion and (static) observer function generating data. These dynamics are driven by random fluctuations v, where w represents measurement noise. The example shown here is perhaps the simplest; with recurrently and reciprocally (and linearly) coupled excitatory (black) and inhibitory (red) neuronal populations. Right panel: this illustrates the corresponding spectral behaviours expressed in terms of spectral densities. The top equation shows that the observed spectral density g(ω) is a mixture of signal generated by applying transfer functions K(ω) to the spectral density of the random fluctuations (assumed to be the identity matrix here for simplicity) plus a component due to measurement noise. Crucially, the transfer functions and ensuing spectral density are determined by the eigenvalues of the model's connectivity (shown on the lower left). In turn, the eigenvalues are relatively simple functions of the connectivity. The resulting (Lorentzian) spectral density is centred on the imaginary part of the eigenvalue and corresponds to the connection strength of reciprocal connections. The dispersion (full width half maximum) of the spectral peak is determined by the recurrent connectivity. This example shows how connectivity parameters can be expressed directly and intuitively in measured spectra. Furthermore, peristimulus time-dependent changes in the spectral peak disclose stimulus-induced changes in the strength of reciprocal connectivity (i.e. short-term changes in synaptic efficacy of the sort that could be mediated by NMDA receptors)—as illustrated on the lower right. In practice, dynamic causal models are much more complicated than the above example; they usually consider distributed networks of sources with multiple populations within each source and multiple states within each population—with non-linear coupling.
© Copyright Policy - CC BY
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4376394&req=5

fig0005: This schematic illustrates the link between the parameters of a dynamic causal model—such as effective connectivity or synaptic efficacy—and the spectral signatures of these coupling parameters. Left panel: state space or dynamic causal model of neuronal states x generating observed data y. The equations at the top represent the equations of motion and (static) observer function generating data. These dynamics are driven by random fluctuations v, where w represents measurement noise. The example shown here is perhaps the simplest; with recurrently and reciprocally (and linearly) coupled excitatory (black) and inhibitory (red) neuronal populations. Right panel: this illustrates the corresponding spectral behaviours expressed in terms of spectral densities. The top equation shows that the observed spectral density g(ω) is a mixture of signal generated by applying transfer functions K(ω) to the spectral density of the random fluctuations (assumed to be the identity matrix here for simplicity) plus a component due to measurement noise. Crucially, the transfer functions and ensuing spectral density are determined by the eigenvalues of the model's connectivity (shown on the lower left). In turn, the eigenvalues are relatively simple functions of the connectivity. The resulting (Lorentzian) spectral density is centred on the imaginary part of the eigenvalue and corresponds to the connection strength of reciprocal connections. The dispersion (full width half maximum) of the spectral peak is determined by the recurrent connectivity. This example shows how connectivity parameters can be expressed directly and intuitively in measured spectra. Furthermore, peristimulus time-dependent changes in the spectral peak disclose stimulus-induced changes in the strength of reciprocal connectivity (i.e. short-term changes in synaptic efficacy of the sort that could be mediated by NMDA receptors)—as illustrated on the lower right. In practice, dynamic causal models are much more complicated than the above example; they usually consider distributed networks of sources with multiple populations within each source and multiple states within each population—with non-linear coupling.
Mentions: The circumstantial evidence for predictive coding is substantial—both in terms of the anatomy of extrinsic (between-areas) and intrinsic (within-area) connections and the physiology of synaptic interactions [3]. In these schemes, top-down predictions are used to form prediction errors at each level of cortical and subcortical hierarchies. The prediction errors are then returned to the level above to update predictions in a Bayesian sense. In brief, the prediction errors report the ‘newsworthy’ information from a lower hierarchical level that was not predicted by the higher level. A crucial aspect of this message passing is the selection of ascending information by adjusting the ‘volume’ or gain of prediction errors that compete for influence over higher levels of processing. Functionally, this gain corresponds to the expected precision (inverse variance or signal-to-noise ratio) that sets the confidence afforded to prediction errors. Psychologically this has been proposed as the basis of attentional gain [6]. Physiologically, precision corresponds to the postsynaptic gain or sensitivity of cells reporting prediction errors (currently thought to be large principal cells that send extrinsic efferents of a forward type, such as superficial pyramidal cells in cortex). This is important because the synaptic gain or efficacy of coupled neuronal populations determines the form of their spectral (oscillatory) behaviour. See Figure 1. Because, synchronous activity determines synaptic gain [7], oscillations have a mechanistic impact on neuronal processing—rather than being epiphenomenal—which completes the circular causality between synchrony and synaptic efficacy.

Bottom Line: In particular, we treat oscillations as the (observable) signature of context-sensitive changes in synaptic efficacy that underlie coordinated dynamics and message-passing in the brain.This rich source of information is now being exploited by various procedures-like dynamic causal modelling-to test hypotheses about neuronal circuits in health and disease.Furthermore, the roles played by neuromodulatory mechanisms can be addressed directly through their effects on oscillatory phenomena.

View Article: PubMed Central - PubMed

Affiliation: The Wellcome Trust Centre for Neuroimaging, University College London, Queen Square, London WC1N 3BG, UK. Electronic address: k.friston@ucl.ac.uk.

Show MeSH
Related in: MedlinePlus