Limits...
Time-Dependent Increase in Network Response to Stimulation.

Hamilton F, Graham R, Luu L, Peixoto N - PLoS ONE (2015)

Bottom Line: Here we demonstrate the effects of a high frequency electrical stimulation signal in training cultured networks of cortical neurons.This increase was found to be statistically significant as compared to control networks that did not receive training.This method was used to identify and track changes in network connectivity strength.

View Article: PubMed Central - PubMed

Affiliation: Department of Electrical and Computer Engineering, George Mason University, Fairfax, VA, United States of America.

ABSTRACT
In vitro neuronal cultures have become a popular method with which to probe network-level neuronal dynamics and phenomena in controlled laboratory settings. One of the key dynamics of interest in these in vitro studies has been the extent to which cultured networks display properties indicative of learning. Here we demonstrate the effects of a high frequency electrical stimulation signal in training cultured networks of cortical neurons. Networks receiving this training signal displayed a time-dependent increase in the response to a low frequency probing stimulation, particularly in the time window of 20-50 ms after stimulation. This increase was found to be statistically significant as compared to control networks that did not receive training. The timing of this increase suggests potentiation of synaptic mechanisms. To further investigate this possibility, we leveraged the powerful Cox statistical connectivity method as previously investigated by our group. This method was used to identify and track changes in network connectivity strength.

No MeSH data available.


Connectivity comparison in networks of Izhikevich neurons.Sensitivity, or number of correctly identified connections, for the Cox method (red curve) and transfer entropy (black curve) in heterogeneous networks of Izhikevich neurons at varying levels of connectivity. Error bars denote standard error over 10 random network realizations. (a) Results for networks of 10 neurons when approximately 500 spikes per neuron are available for analysis. (b) Results for networks of 10 neurons when approximately 1000 spikes per neuron are available. While both methods receive an expected increase in performance going from 500 to 1000 spikes, there is a clear performance advantage in using the Cox method over transfer entropy in both the 500 and 1000 spikes per neuron situation. (c) The discrepancy in performance is even more evident when examining networks of 20 Izhikevich neurons. The Cox method (red curve) with access to only 1000 spikes per neuron outperforms transfer entropy with access to 2000 spikes per neuron (black curve) or 4000 spikes per neuron (dotted black curve).
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4636320&req=5

pone.0142399.g007: Connectivity comparison in networks of Izhikevich neurons.Sensitivity, or number of correctly identified connections, for the Cox method (red curve) and transfer entropy (black curve) in heterogeneous networks of Izhikevich neurons at varying levels of connectivity. Error bars denote standard error over 10 random network realizations. (a) Results for networks of 10 neurons when approximately 500 spikes per neuron are available for analysis. (b) Results for networks of 10 neurons when approximately 1000 spikes per neuron are available. While both methods receive an expected increase in performance going from 500 to 1000 spikes, there is a clear performance advantage in using the Cox method over transfer entropy in both the 500 and 1000 spikes per neuron situation. (c) The discrepancy in performance is even more evident when examining networks of 20 Izhikevich neurons. The Cox method (red curve) with access to only 1000 spikes per neuron outperforms transfer entropy with access to 2000 spikes per neuron (black curve) or 4000 spikes per neuron (dotted black curve).

Mentions: Fig 7 shows the sensitivity results of the Cox method (red curves) and transfer entropy (black curves) as a function of network connectivity for various networks of Izhikevich neurons. Error bars denote standard error over 10 random network realizations. Fig 7a and 7b shows the results of both methods in networks of 10 neurons with approximately (a) 500 spikes per neuron and (b) 1000 spikes per neuron. For both methods there is an improvement in connectivity identification as the number of spikes per neuron increases, which is to be expected. However, the Cox method displays a clear performance advantage over transfer entropy in both situations. This advantage becomes even more evident when examining the connectivity problem in networks of 20 neurons. Fig 7c shows the results of the connectivity identification in networks of 20 neurons. The Cox method (red curve) using only 1000 spikes per neuron is able to identify the majority of the network connections. Transfer entropy, using 2000 spikes per neuron (black curve) and 4000 spikes per neuron (black dotted curve), has a difficult time identifying the network connectivity given the limited number of spikes per neuron in these larger networks.


Time-Dependent Increase in Network Response to Stimulation.

Hamilton F, Graham R, Luu L, Peixoto N - PLoS ONE (2015)

Connectivity comparison in networks of Izhikevich neurons.Sensitivity, or number of correctly identified connections, for the Cox method (red curve) and transfer entropy (black curve) in heterogeneous networks of Izhikevich neurons at varying levels of connectivity. Error bars denote standard error over 10 random network realizations. (a) Results for networks of 10 neurons when approximately 500 spikes per neuron are available for analysis. (b) Results for networks of 10 neurons when approximately 1000 spikes per neuron are available. While both methods receive an expected increase in performance going from 500 to 1000 spikes, there is a clear performance advantage in using the Cox method over transfer entropy in both the 500 and 1000 spikes per neuron situation. (c) The discrepancy in performance is even more evident when examining networks of 20 Izhikevich neurons. The Cox method (red curve) with access to only 1000 spikes per neuron outperforms transfer entropy with access to 2000 spikes per neuron (black curve) or 4000 spikes per neuron (dotted black curve).
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4636320&req=5

pone.0142399.g007: Connectivity comparison in networks of Izhikevich neurons.Sensitivity, or number of correctly identified connections, for the Cox method (red curve) and transfer entropy (black curve) in heterogeneous networks of Izhikevich neurons at varying levels of connectivity. Error bars denote standard error over 10 random network realizations. (a) Results for networks of 10 neurons when approximately 500 spikes per neuron are available for analysis. (b) Results for networks of 10 neurons when approximately 1000 spikes per neuron are available. While both methods receive an expected increase in performance going from 500 to 1000 spikes, there is a clear performance advantage in using the Cox method over transfer entropy in both the 500 and 1000 spikes per neuron situation. (c) The discrepancy in performance is even more evident when examining networks of 20 Izhikevich neurons. The Cox method (red curve) with access to only 1000 spikes per neuron outperforms transfer entropy with access to 2000 spikes per neuron (black curve) or 4000 spikes per neuron (dotted black curve).
Mentions: Fig 7 shows the sensitivity results of the Cox method (red curves) and transfer entropy (black curves) as a function of network connectivity for various networks of Izhikevich neurons. Error bars denote standard error over 10 random network realizations. Fig 7a and 7b shows the results of both methods in networks of 10 neurons with approximately (a) 500 spikes per neuron and (b) 1000 spikes per neuron. For both methods there is an improvement in connectivity identification as the number of spikes per neuron increases, which is to be expected. However, the Cox method displays a clear performance advantage over transfer entropy in both situations. This advantage becomes even more evident when examining the connectivity problem in networks of 20 neurons. Fig 7c shows the results of the connectivity identification in networks of 20 neurons. The Cox method (red curve) using only 1000 spikes per neuron is able to identify the majority of the network connections. Transfer entropy, using 2000 spikes per neuron (black curve) and 4000 spikes per neuron (black dotted curve), has a difficult time identifying the network connectivity given the limited number of spikes per neuron in these larger networks.

Bottom Line: Here we demonstrate the effects of a high frequency electrical stimulation signal in training cultured networks of cortical neurons.This increase was found to be statistically significant as compared to control networks that did not receive training.This method was used to identify and track changes in network connectivity strength.

View Article: PubMed Central - PubMed

Affiliation: Department of Electrical and Computer Engineering, George Mason University, Fairfax, VA, United States of America.

ABSTRACT
In vitro neuronal cultures have become a popular method with which to probe network-level neuronal dynamics and phenomena in controlled laboratory settings. One of the key dynamics of interest in these in vitro studies has been the extent to which cultured networks display properties indicative of learning. Here we demonstrate the effects of a high frequency electrical stimulation signal in training cultured networks of cortical neurons. Networks receiving this training signal displayed a time-dependent increase in the response to a low frequency probing stimulation, particularly in the time window of 20-50 ms after stimulation. This increase was found to be statistically significant as compared to control networks that did not receive training. The timing of this increase suggests potentiation of synaptic mechanisms. To further investigate this possibility, we leveraged the powerful Cox statistical connectivity method as previously investigated by our group. This method was used to identify and track changes in network connectivity strength.

No MeSH data available.