Limits...
Pattern segmentation with activity dependent natural frequency shift and sub-threshold resonance.

Shtrahman E, Zochowski M - Sci Rep (2015)

Bottom Line: Understanding the mechanisms underlying distributed pattern formation in brain networks and its content driven dynamical segmentation is an area of intense study.We investigate a theoretical mechanism for selective activation of diverse neural populations that is based on dynamically shifting cellular resonances in functionally or structurally coupled networks.We find that this mechanism is robust and suggest it as a general coding strategy that can be applied to any network with oscillatory nodes.

View Article: PubMed Central - PubMed

Affiliation: Applied Physics Program, University of Michigan - Ann Arbor 48109, USA.

ABSTRACT
Understanding the mechanisms underlying distributed pattern formation in brain networks and its content driven dynamical segmentation is an area of intense study. We investigate a theoretical mechanism for selective activation of diverse neural populations that is based on dynamically shifting cellular resonances in functionally or structurally coupled networks. We specifically show that sub-threshold neuronal depolarization from synaptic coupling or external input can shift neurons into and out of resonance with specific bands of existing extracellular oscillations, and this can act as a dynamic readout mechanism during information storage and retrieval. We find that this mechanism is robust and suggest it as a general coding strategy that can be applied to any network with oscillatory nodes.

Show MeSH

Related in: MedlinePlus

Response time to current induced resonance shift.(a) Example resonance curves for a network after additional current input begins 10 seconds into the simulation. The current input shifts the resonance frequency of the neurons as marked by the arrows. (b–d) response times of the network depicted on (a) 16 Hz (b), 18.5 Hz (c), and 21 Hz (d). Solid black line marks the onset of additional sub-threshold current. The average instantaneous firing rate over time is plotted for neurons within and outside of the network heterogeneity (σ = 4 and σ = 1, respectively) and also for non-oscillatory current input (A = 0). The inset in (c) shows the magnified timescale of firing onset. Black dashed lines mark 5 standard deviations above the baseline firing rate. All firing rate curves are averaged over 4 simulations.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4352860&req=5

f5: Response time to current induced resonance shift.(a) Example resonance curves for a network after additional current input begins 10 seconds into the simulation. The current input shifts the resonance frequency of the neurons as marked by the arrows. (b–d) response times of the network depicted on (a) 16 Hz (b), 18.5 Hz (c), and 21 Hz (d). Solid black line marks the onset of additional sub-threshold current. The average instantaneous firing rate over time is plotted for neurons within and outside of the network heterogeneity (σ = 4 and σ = 1, respectively) and also for non-oscillatory current input (A = 0). The inset in (c) shows the magnified timescale of firing onset. Black dashed lines mark 5 standard deviations above the baseline firing rate. All firing rate curves are averaged over 4 simulations.

Mentions: We have shown that current dependent resonance shifts can separate out regions of a network with higher coupling in both their firing frequency and phase locking to sub-threshold oscillatory input. We then investigated the response time of the network to new patterns of input in order to understand how long the network takes to form separated representations in response to shifting resonances. Here the timescale is set by the frequency of the oscillatory input and can potentially be directly compared with experimental findings measuring response time to the incoming stimulus293031. We investigated the timescale of pattern formation by observing the response time required to reach the significantly elevated firing rates observed in the resonance curves. Specifically, to probe this onset time of enhanced resonant firing, we applied an additional sub-threshold current input during the simulation and measured the time delay between this input and the significant change (see methods) in neuronal firing rate (solid black lines in Fig. 5). As before, we simulated the sub-networks having two strengths of coupling constants σ. We found that the response time critically depends on the relative position (in frequency space) of driving frequency with respect to network resonant curve (Fig. 5a). The network response time to the external input on the rising portion of the resonance curve (Fig. 5b) is significantly faster than on the falling portions (Fig. 5c,d). The shifted resonance curve of the heterogeneity (high σ) relative to the rest of the network (low σ) results in the rise and decline portions of their resonance curves to occur for different driving frequencies. Therefore, we observe that for driving frequencies where the resonance curves overlap (Fig. 5c) and the neurons in the heterogeneity are on the upward sloping portion of their resonance and the neurons outside of the heterogeneity are on the downward portion, the response times are widely different. (Fig. 5c inset).


Pattern segmentation with activity dependent natural frequency shift and sub-threshold resonance.

Shtrahman E, Zochowski M - Sci Rep (2015)

Response time to current induced resonance shift.(a) Example resonance curves for a network after additional current input begins 10 seconds into the simulation. The current input shifts the resonance frequency of the neurons as marked by the arrows. (b–d) response times of the network depicted on (a) 16 Hz (b), 18.5 Hz (c), and 21 Hz (d). Solid black line marks the onset of additional sub-threshold current. The average instantaneous firing rate over time is plotted for neurons within and outside of the network heterogeneity (σ = 4 and σ = 1, respectively) and also for non-oscillatory current input (A = 0). The inset in (c) shows the magnified timescale of firing onset. Black dashed lines mark 5 standard deviations above the baseline firing rate. All firing rate curves are averaged over 4 simulations.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4352860&req=5

f5: Response time to current induced resonance shift.(a) Example resonance curves for a network after additional current input begins 10 seconds into the simulation. The current input shifts the resonance frequency of the neurons as marked by the arrows. (b–d) response times of the network depicted on (a) 16 Hz (b), 18.5 Hz (c), and 21 Hz (d). Solid black line marks the onset of additional sub-threshold current. The average instantaneous firing rate over time is plotted for neurons within and outside of the network heterogeneity (σ = 4 and σ = 1, respectively) and also for non-oscillatory current input (A = 0). The inset in (c) shows the magnified timescale of firing onset. Black dashed lines mark 5 standard deviations above the baseline firing rate. All firing rate curves are averaged over 4 simulations.
Mentions: We have shown that current dependent resonance shifts can separate out regions of a network with higher coupling in both their firing frequency and phase locking to sub-threshold oscillatory input. We then investigated the response time of the network to new patterns of input in order to understand how long the network takes to form separated representations in response to shifting resonances. Here the timescale is set by the frequency of the oscillatory input and can potentially be directly compared with experimental findings measuring response time to the incoming stimulus293031. We investigated the timescale of pattern formation by observing the response time required to reach the significantly elevated firing rates observed in the resonance curves. Specifically, to probe this onset time of enhanced resonant firing, we applied an additional sub-threshold current input during the simulation and measured the time delay between this input and the significant change (see methods) in neuronal firing rate (solid black lines in Fig. 5). As before, we simulated the sub-networks having two strengths of coupling constants σ. We found that the response time critically depends on the relative position (in frequency space) of driving frequency with respect to network resonant curve (Fig. 5a). The network response time to the external input on the rising portion of the resonance curve (Fig. 5b) is significantly faster than on the falling portions (Fig. 5c,d). The shifted resonance curve of the heterogeneity (high σ) relative to the rest of the network (low σ) results in the rise and decline portions of their resonance curves to occur for different driving frequencies. Therefore, we observe that for driving frequencies where the resonance curves overlap (Fig. 5c) and the neurons in the heterogeneity are on the upward sloping portion of their resonance and the neurons outside of the heterogeneity are on the downward portion, the response times are widely different. (Fig. 5c inset).

Bottom Line: Understanding the mechanisms underlying distributed pattern formation in brain networks and its content driven dynamical segmentation is an area of intense study.We investigate a theoretical mechanism for selective activation of diverse neural populations that is based on dynamically shifting cellular resonances in functionally or structurally coupled networks.We find that this mechanism is robust and suggest it as a general coding strategy that can be applied to any network with oscillatory nodes.

View Article: PubMed Central - PubMed

Affiliation: Applied Physics Program, University of Michigan - Ann Arbor 48109, USA.

ABSTRACT
Understanding the mechanisms underlying distributed pattern formation in brain networks and its content driven dynamical segmentation is an area of intense study. We investigate a theoretical mechanism for selective activation of diverse neural populations that is based on dynamically shifting cellular resonances in functionally or structurally coupled networks. We specifically show that sub-threshold neuronal depolarization from synaptic coupling or external input can shift neurons into and out of resonance with specific bands of existing extracellular oscillations, and this can act as a dynamic readout mechanism during information storage and retrieval. We find that this mechanism is robust and suggest it as a general coding strategy that can be applied to any network with oscillatory nodes.

Show MeSH
Related in: MedlinePlus