Limits...
Tracking slow modulations in synaptic gain using dynamic causal modelling: validation in epilepsy.

Papadopoulou M, Leite M, van Mierlo P, Vonck K, Lemieux L, Friston K, Marinazzo D - Neuroimage (2014)

Bottom Line: Bayesian model selection was used to identify the intrinsic (within-source) and extrinsic (between-source) connectivity.Having established the underlying architecture, we were able to track the evolution of key connectivity parameters (e.g., inhibitory connections to superficial pyramidal cells) and test specific hypotheses about the synaptic mechanisms involved in ictogenesis.Crucially, these changes spoke to an increase in the sensitivity of principal cells to intrinsic inhibitory afferents and a transient loss of excitatory-inhibitory balance.

View Article: PubMed Central - PubMed

Affiliation: Department of Data-analysis, University of Ghent, B9000 Ghent, Belgium.

Show MeSH

Related in: MedlinePlus

Left panels: Response characteristics of a single source within a dynamic causal model of the sort used in subsequent analyses (a canonical microcircuit neural mass model). The upper panels show the first and second order impulse response functions of time in terms of their impulse responses (Volterra kernels). These reflect the impact of inputs on observed responses and are a function of the model's parameters. The equivalent formulation of the impulse response in frequency space is shown in the lower panels graphically (on the lower left) and in image format for different values of the inhibitory connection (on the lower right). These are called (modulation) transfer functions and represent the frequencies in the inputs that are expressed in the output. In this example, we have shown the responses as a function of (the log scaling of) recurrent inhibitory connectivity to one of four neuronal populations comprising the source (see Fig. 3). These response functions can be used to compute the expected cross spectral density for any values of the parameters. Right panels: these illustrate changes in neuronal activity when increasing recurrent inhibition. The top panel shows strength of recurrent inhibition as a function of time in seconds, while the second panel shows a simulated response obtained by integrating the neural mass model with random fluctuating inputs, with the value of inhibitory connection set to 1.5. The simulated time frequency response is shown below in terms of the spectral power over 4 to 96 Hz. The lowest panel shows the predicted power based upon the transfer functions shown on the left.
© Copyright Policy - CC BY
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4306529&req=5

f0010: Left panels: Response characteristics of a single source within a dynamic causal model of the sort used in subsequent analyses (a canonical microcircuit neural mass model). The upper panels show the first and second order impulse response functions of time in terms of their impulse responses (Volterra kernels). These reflect the impact of inputs on observed responses and are a function of the model's parameters. The equivalent formulation of the impulse response in frequency space is shown in the lower panels graphically (on the lower left) and in image format for different values of the inhibitory connection (on the lower right). These are called (modulation) transfer functions and represent the frequencies in the inputs that are expressed in the output. In this example, we have shown the responses as a function of (the log scaling of) recurrent inhibitory connectivity to one of four neuronal populations comprising the source (see Fig. 3). These response functions can be used to compute the expected cross spectral density for any values of the parameters. Right panels: these illustrate changes in neuronal activity when increasing recurrent inhibition. The top panel shows strength of recurrent inhibition as a function of time in seconds, while the second panel shows a simulated response obtained by integrating the neural mass model with random fluctuating inputs, with the value of inhibitory connection set to 1.5. The simulated time frequency response is shown below in terms of the spectral power over 4 to 96 Hz. The lowest panel shows the predicted power based upon the transfer functions shown on the left.

Mentions: In summary, DCM solves the inverse problem of recovering plausible parameters (of both neuronal dynamics and noise) that explain observed cross spectra. It uses standard variational Bayesian procedures (Friston et al., 2007) to fit time-series or cross spectra – under model complexity constraints – to provide maximum a posteriori estimates of the underlying model parameters and the evidence for any particular model; see Friston et al. (2012) for more details in this particular setting. Fig. 2 illustrates the basic idea behind the application of dynamic causal modelling to cross spectral responses. The key point made by this figure is that changes in connectivity can have profound effects on spectral behaviour responses to endogenous input. It is these effects that are used to estimate (changes in) the underlying connectivity (Friston, 2014). If we take the modifications in the amplitude and frequencies produced by changes in model parameters as a simple model of seizure onset, one can use the predicted spectral responses as a likelihood model of empirical responses and thereby estimate the time-dependent changes in parameters. The simulations reported in Fig. 2 can be reproduced using the seizure onset demonstration in the neuronal modelling toolbox of the academic SPM freeware (http://www.fil.ion.ucl.ac.uk/spm). These simulation results use standard parameter values (prior expectations: see Table 1).


Tracking slow modulations in synaptic gain using dynamic causal modelling: validation in epilepsy.

Papadopoulou M, Leite M, van Mierlo P, Vonck K, Lemieux L, Friston K, Marinazzo D - Neuroimage (2014)

Left panels: Response characteristics of a single source within a dynamic causal model of the sort used in subsequent analyses (a canonical microcircuit neural mass model). The upper panels show the first and second order impulse response functions of time in terms of their impulse responses (Volterra kernels). These reflect the impact of inputs on observed responses and are a function of the model's parameters. The equivalent formulation of the impulse response in frequency space is shown in the lower panels graphically (on the lower left) and in image format for different values of the inhibitory connection (on the lower right). These are called (modulation) transfer functions and represent the frequencies in the inputs that are expressed in the output. In this example, we have shown the responses as a function of (the log scaling of) recurrent inhibitory connectivity to one of four neuronal populations comprising the source (see Fig. 3). These response functions can be used to compute the expected cross spectral density for any values of the parameters. Right panels: these illustrate changes in neuronal activity when increasing recurrent inhibition. The top panel shows strength of recurrent inhibition as a function of time in seconds, while the second panel shows a simulated response obtained by integrating the neural mass model with random fluctuating inputs, with the value of inhibitory connection set to 1.5. The simulated time frequency response is shown below in terms of the spectral power over 4 to 96 Hz. The lowest panel shows the predicted power based upon the transfer functions shown on the left.
© Copyright Policy - CC BY
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4306529&req=5

f0010: Left panels: Response characteristics of a single source within a dynamic causal model of the sort used in subsequent analyses (a canonical microcircuit neural mass model). The upper panels show the first and second order impulse response functions of time in terms of their impulse responses (Volterra kernels). These reflect the impact of inputs on observed responses and are a function of the model's parameters. The equivalent formulation of the impulse response in frequency space is shown in the lower panels graphically (on the lower left) and in image format for different values of the inhibitory connection (on the lower right). These are called (modulation) transfer functions and represent the frequencies in the inputs that are expressed in the output. In this example, we have shown the responses as a function of (the log scaling of) recurrent inhibitory connectivity to one of four neuronal populations comprising the source (see Fig. 3). These response functions can be used to compute the expected cross spectral density for any values of the parameters. Right panels: these illustrate changes in neuronal activity when increasing recurrent inhibition. The top panel shows strength of recurrent inhibition as a function of time in seconds, while the second panel shows a simulated response obtained by integrating the neural mass model with random fluctuating inputs, with the value of inhibitory connection set to 1.5. The simulated time frequency response is shown below in terms of the spectral power over 4 to 96 Hz. The lowest panel shows the predicted power based upon the transfer functions shown on the left.
Mentions: In summary, DCM solves the inverse problem of recovering plausible parameters (of both neuronal dynamics and noise) that explain observed cross spectra. It uses standard variational Bayesian procedures (Friston et al., 2007) to fit time-series or cross spectra – under model complexity constraints – to provide maximum a posteriori estimates of the underlying model parameters and the evidence for any particular model; see Friston et al. (2012) for more details in this particular setting. Fig. 2 illustrates the basic idea behind the application of dynamic causal modelling to cross spectral responses. The key point made by this figure is that changes in connectivity can have profound effects on spectral behaviour responses to endogenous input. It is these effects that are used to estimate (changes in) the underlying connectivity (Friston, 2014). If we take the modifications in the amplitude and frequencies produced by changes in model parameters as a simple model of seizure onset, one can use the predicted spectral responses as a likelihood model of empirical responses and thereby estimate the time-dependent changes in parameters. The simulations reported in Fig. 2 can be reproduced using the seizure onset demonstration in the neuronal modelling toolbox of the academic SPM freeware (http://www.fil.ion.ucl.ac.uk/spm). These simulation results use standard parameter values (prior expectations: see Table 1).

Bottom Line: Bayesian model selection was used to identify the intrinsic (within-source) and extrinsic (between-source) connectivity.Having established the underlying architecture, we were able to track the evolution of key connectivity parameters (e.g., inhibitory connections to superficial pyramidal cells) and test specific hypotheses about the synaptic mechanisms involved in ictogenesis.Crucially, these changes spoke to an increase in the sensitivity of principal cells to intrinsic inhibitory afferents and a transient loss of excitatory-inhibitory balance.

View Article: PubMed Central - PubMed

Affiliation: Department of Data-analysis, University of Ghent, B9000 Ghent, Belgium.

Show MeSH
Related in: MedlinePlus