Limits...
Conversion of phase information into a spike-count code by bursting neurons.

Samengo I, Montemurro MA - PLoS ONE (2010)

Bottom Line: We show that the number of spikes per burst varies systematically with the phase of the fluctuating input at the time of burst onset.The mapping between input phase and number of spikes per burst is a robust response feature for a broad range of stimulus statistics.Our results suggest that cortical bursting neurons could play a crucial role in translating LFP phase information into an easily decodable spike count code.

View Article: PubMed Central - PubMed

Affiliation: Centro Atómico Bariloche and Instituto Balseiro, San Carlos de Bariloche, Argentina.

ABSTRACT
Single neurons in the cerebral cortex are immersed in a fluctuating electric field, the local field potential (LFP), which mainly originates from synchronous synaptic input into the local neural neighborhood. As shown by recent studies in visual and auditory cortices, the angular phase of the LFP at the time of spike generation adds significant extra information about the external world, beyond the one contained in the firing rate alone. However, no biologically plausible mechanism has yet been suggested that allows downstream neurons to infer the phase of the LFP at the soma of their pre-synaptic afferents. Therefore, so far there is no evidence that the nervous system can process phase information. Here we study a model of a bursting pyramidal neuron, driven by a time-dependent stimulus. We show that the number of spikes per burst varies systematically with the phase of the fluctuating input at the time of burst onset. The mapping between input phase and number of spikes per burst is a robust response feature for a broad range of stimulus statistics. Our results suggest that cortical bursting neurons could play a crucial role in translating LFP phase information into an easily decodable spike count code.

Show MeSH

Related in: MedlinePlus

Comparison of different candidate burst-codes.(A, B, C) Normalized histograms of the stimulus amplitude (A), slope (B) and phase (C) at burst initiation, for different burst lengths n. For each vale of n, the color plot represents an estimation of the probability of each stimulus feature. The highest discriminability is obtained for the stimulus phase, given that the phase distributions show minimal overlap for different n values. The plots pool together all the phases obtained with the collection of stimulus amplitudes and periods used in Figure 3 B. (D, E, F) Same distributions as in A, B, C, but when the stimulus consists of a low-pass filtered Gaussian signal, with 30 Hz cut-off frequency. As for the sinusoidal case, the phase code has the maximal discriminability.
© Copyright Policy
Related In: Results  -  Collection


getmorefigures.php?uid=PMC2837377&req=5

pone-0009669-g005: Comparison of different candidate burst-codes.(A, B, C) Normalized histograms of the stimulus amplitude (A), slope (B) and phase (C) at burst initiation, for different burst lengths n. For each vale of n, the color plot represents an estimation of the probability of each stimulus feature. The highest discriminability is obtained for the stimulus phase, given that the phase distributions show minimal overlap for different n values. The plots pool together all the phases obtained with the collection of stimulus amplitudes and periods used in Figure 3 B. (D, E, F) Same distributions as in A, B, C, but when the stimulus consists of a low-pass filtered Gaussian signal, with 30 Hz cut-off frequency. As for the sinusoidal case, the phase code has the maximal discriminability.

Mentions: In order for a neural code based on the number of spikes per burst to be useful, the mapping between n and the encoded stimulus feature must not depend strongly on the properties of the input signal. For instance, changing the stimulus frequency should not lead to a significantly different mapping. We therefore assessed whether the three neural codes discussed in Figure 4 were stable with respect to changes in the stimulus amplitude and frequency. In Figure 5A–C we show the probability distribution of each candidate stimulus feature, given that the cell generated a burst of n spikes. Each distribution pools together the results obtained with stimuli covering a broad range of periods and amplitudes (the range is shown in Figures 4B and 4D–F). Light colors represent high probability. When considering the distribution of stimulus amplitudes (A) and slopes (B), the probability densities corresponding to different n values overlap significantly. Hence, by reading out the number of spikes per burst, it is not possible to guess the value of the stimulus amplitude or slope. In contrast, the correspondence between stimulus phase and n is remarkably narrow (Figure 5C). Therefore, all stimuli seem to induce the same mapping between phase and n, irrespective of the amplitude or frequency of the input signal.


Conversion of phase information into a spike-count code by bursting neurons.

Samengo I, Montemurro MA - PLoS ONE (2010)

Comparison of different candidate burst-codes.(A, B, C) Normalized histograms of the stimulus amplitude (A), slope (B) and phase (C) at burst initiation, for different burst lengths n. For each vale of n, the color plot represents an estimation of the probability of each stimulus feature. The highest discriminability is obtained for the stimulus phase, given that the phase distributions show minimal overlap for different n values. The plots pool together all the phases obtained with the collection of stimulus amplitudes and periods used in Figure 3 B. (D, E, F) Same distributions as in A, B, C, but when the stimulus consists of a low-pass filtered Gaussian signal, with 30 Hz cut-off frequency. As for the sinusoidal case, the phase code has the maximal discriminability.
© Copyright Policy
Related In: Results  -  Collection

Show All Figures
getmorefigures.php?uid=PMC2837377&req=5

pone-0009669-g005: Comparison of different candidate burst-codes.(A, B, C) Normalized histograms of the stimulus amplitude (A), slope (B) and phase (C) at burst initiation, for different burst lengths n. For each vale of n, the color plot represents an estimation of the probability of each stimulus feature. The highest discriminability is obtained for the stimulus phase, given that the phase distributions show minimal overlap for different n values. The plots pool together all the phases obtained with the collection of stimulus amplitudes and periods used in Figure 3 B. (D, E, F) Same distributions as in A, B, C, but when the stimulus consists of a low-pass filtered Gaussian signal, with 30 Hz cut-off frequency. As for the sinusoidal case, the phase code has the maximal discriminability.
Mentions: In order for a neural code based on the number of spikes per burst to be useful, the mapping between n and the encoded stimulus feature must not depend strongly on the properties of the input signal. For instance, changing the stimulus frequency should not lead to a significantly different mapping. We therefore assessed whether the three neural codes discussed in Figure 4 were stable with respect to changes in the stimulus amplitude and frequency. In Figure 5A–C we show the probability distribution of each candidate stimulus feature, given that the cell generated a burst of n spikes. Each distribution pools together the results obtained with stimuli covering a broad range of periods and amplitudes (the range is shown in Figures 4B and 4D–F). Light colors represent high probability. When considering the distribution of stimulus amplitudes (A) and slopes (B), the probability densities corresponding to different n values overlap significantly. Hence, by reading out the number of spikes per burst, it is not possible to guess the value of the stimulus amplitude or slope. In contrast, the correspondence between stimulus phase and n is remarkably narrow (Figure 5C). Therefore, all stimuli seem to induce the same mapping between phase and n, irrespective of the amplitude or frequency of the input signal.

Bottom Line: We show that the number of spikes per burst varies systematically with the phase of the fluctuating input at the time of burst onset.The mapping between input phase and number of spikes per burst is a robust response feature for a broad range of stimulus statistics.Our results suggest that cortical bursting neurons could play a crucial role in translating LFP phase information into an easily decodable spike count code.

View Article: PubMed Central - PubMed

Affiliation: Centro Atómico Bariloche and Instituto Balseiro, San Carlos de Bariloche, Argentina.

ABSTRACT
Single neurons in the cerebral cortex are immersed in a fluctuating electric field, the local field potential (LFP), which mainly originates from synchronous synaptic input into the local neural neighborhood. As shown by recent studies in visual and auditory cortices, the angular phase of the LFP at the time of spike generation adds significant extra information about the external world, beyond the one contained in the firing rate alone. However, no biologically plausible mechanism has yet been suggested that allows downstream neurons to infer the phase of the LFP at the soma of their pre-synaptic afferents. Therefore, so far there is no evidence that the nervous system can process phase information. Here we study a model of a bursting pyramidal neuron, driven by a time-dependent stimulus. We show that the number of spikes per burst varies systematically with the phase of the fluctuating input at the time of burst onset. The mapping between input phase and number of spikes per burst is a robust response feature for a broad range of stimulus statistics. Our results suggest that cortical bursting neurons could play a crucial role in translating LFP phase information into an easily decodable spike count code.

Show MeSH
Related in: MedlinePlus