Skip to main page content
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2010;2010:393019.
doi: 10.1155/2010/393019. Epub 2010 Feb 3.

Decoupling Action Potential Bias From Cortical Local Field Potentials

Affiliations
Free PMC article

Decoupling Action Potential Bias From Cortical Local Field Potentials

Stephen V David et al. Comput Intell Neurosci. .
Free PMC article

Abstract

Neurophysiologists have recently become interested in studying neuronal population activity through local field potential (LFP) recordings during experiments that also record the activity of single neurons. This experimental approach differs from early LFP studies because it uses high impedence electrodes that can also isolate single neuron activity. A possible complication for such studies is that the synaptic potentials and action potentials of the small subset of isolated neurons may contribute disproportionately to the LFP signal, biasing activity in the larger nearby neuronal population to appear synchronous and cotuned with these neurons. To address this problem, we used linear filtering techniques to remove features correlated with spike events from LFP recordings. This filtering procedure can be applied for well-isolated single units or multiunit activity. We illustrate the effects of this correction in simulation and on spike data recorded from primary auditory cortex. We find that local spiking activity can explain a significant portion of LFP power at most recording sites and demonstrate that removing the spike-correlated component can affect measurements of auditory tuning of the LFP.

Figures

Figure 1
Figure 1
Removal of spike-coupled LFP signal in simulated data. (a) Average waveform of simulated frequency-tuned spiking activity introduced to a random (1/f noise) LFP signal. (b) Spike-LFP filter estimated by (5). (c) (Top panel) High-frequency spike signal, r h(t), with SUA4 events marked by green circles. (Bottom panel) Raw LFP signal (L 0, black) and “clean” LFP signal with SUA4 events removed using filter in (b) (L SUA4, green). The clean LFP closely matches the underlying LFP signal (L actual, black dotted line) before spiking activity was added. (d) Frequency tuning of SUA4 activity. (e) Frequency tuning measured for raw LFP shows tuning similar to spiking activity (black). After SUA4 activity was removed, frequency tuning disappears (green). When only the mean spike-LFP correlation is subtracted for each spike event, the tuning is only partially removed from the LFP signal (red).
Figure 2
Figure 2
Example spike and LFP responses for an electrophysiological recording from primary auditory cortex (A1). (a) Brief segment of a raw high-pass filtered signal (black curve, top) and spike events identified by sudden changes in the signal (SUAn, circles, and “x”s). Green and blue dashed lines indicate, respectively, the thresholds for SUA4 and SUA3 events. Subpanels at right show 100 examples of SUA4 spikes events (a) and the impulse response function that best predicts the LFP (b) from SUA4 (green) or SUA3 events (blue), with standard error indicated by the shading. The dashed blue line shows the SUA3 filter estimate for spike events on a second electrode 0.4 mm from the LFP electrode. The simultaneously recorded raw LFP (black curve, middle) was substantially modified when the component predicted by SUA4 or SUA3 events was removed. The difference between the cleaned and raw LFP signals was the greatest during periods of elevated spiking activity (dashed curves, bottom). (b) The same procedure for removing coupled spike information from the LFP but using multiunit activity (MUA). The MUA signal for the same data segment as in A, defined by (2), captured the elevated firing at 0.3 second (red curve, top). The LFP signal with MUA removed (red curve, middle; difference in dashed line, bottom) roughly followed the same pattern as the LFP with SUA removed. The subpanel at right shows the impulse response that best predicted the LFP from the MUA.
Figure 3
Figure 3
Example spike and LFP responses for a second recording site. Data are plotted as in Figure 2. (a) The impulse responses for SUA4 and SUA3 (subpanel at lower right) are smaller than the impulse responses in Figure 2, and using this function to remove LFP components that could be predicted by SUA events had very little effect. (b) Similarly, removing LFP components that could be predicted by MUA had very little effect on the LFP for this site.
Figure 4
Figure 4
Effect of removing coupled spike activity on total LFP power. (a) Histogram of the ratio of power in the LFP after removing components explained by SUA4 and power in the raw LFP (n = 127 recording sites). For a small number of sites, LFP power increased slightly, reflecting the introduction of a small amount of noise by the cross-validation procedure used for filter estimation. (b) Histogram of change in power after removing the SUA3 component. The average power was significantly lower than for SUA4 (jackknifed t-test, P = .0008). (c) Histogram of change in power after removing the MUA component. The average power was significantly lower than for SUA3 (jackknifed t-test, P = .0007).
Figure 5
Figure 5
Frequency specificity of the spike-coupled LFP signal. (a) Relative power spectrum of the LFP signal from Figure 2 after removing SUA4, SUA3, and MUA components (colors as in previous figures). This site showed a large decrease at low frequencies (1–25 Hz). (b) Relative power spectrum of the LFP signal from the site in Figure 3 after the removal of spiking components (plotted as in A). This site showed a decrease near 25 Hz and at frequencies above 75 Hz. (c) Average relative power spectrum of LFP signal averaged across n = 127 recording sites. Removing coupled spike activity from the LFP signal reduced power at all frequencies. The effect was the strongest at low frequencies (1–10 Hz) but also showed a tendency to grow larger at high frequencies. Consistent with the overall changes in power reported in Figure 4, the LFP spectrum was reduced more for the more permissive definitions of spiking (MUA < SUA3 < SUA4). The small features around 60 Hz (most prominent for the signal with MUA components removed) reflect artifacts of line noise.
Figure 6
Figure 6
Effect of removing spike-correlated activity on the frequency tuning of LFP in A1. (a) Frequency tuning curves for site shown in Figure 2. Gaussian fits are plotted with dashed lines and best frequency (peak of the Gaussian fits) is indicated by arrows. The raw LFP tuning curve was centered at a higher best frequency than the SUA4 curve (0.92 octaves above SUA4, P = .007, jackknifed t-test). After the SUA4-coupled component was removed, the LFP tuning curve was shifted to even higher frequencies (1.63 octaves above SUA4, P = .01, jackknifed t-test). Similar curves to this last case are observed for the LFP signals with SUA3 and MUA components removed. (b) Tuning curves for site shown in Figure 3 (plotted as in A). For this site, there was no significant difference between the SUA and LFP tuning curves (<0.1 octave difference), even after the spike-coupled component was removed from the LFP.

Similar articles

See all similar articles

Cited by 10 articles

See all "Cited by" articles

References

    1. Mitzdorf U. Properties of the evoked potential generators: current source-density analysis of visually evoked potentials in the cat cortex. International Journal of Neuroscience. 1987;33(1-2):33–59. - PubMed
    1. Logothetis NK. The underpinnings of the BOLD functional magnetic resonance imaging signal. The Journal of Neuroscience. 2003;23(10):3963–3971. - PMC - PubMed
    1. Lopes da Silva F. Functional localization of brain sources using EEG and/or MEG data: volume conductor and source models. Magnetic Resonance Imaging. 2004;22(10):1533–1538. - PubMed
    1. Canolty RT, Edwards E, Dalal SS, et al. High gamma power is phase-locked to theta oscillations in human neocortex. Science. 2006;313(5793):1626–1628. - PMC - PubMed
    1. Gray CM, Konig P, Engel AK, Singer W. Oscillatory responses in cat visual cortex exhibit inter-columnar synchronization which reflects global stimulus properties. Nature. 1989;338(6213):334–337. - PubMed

Publication types

LinkOut - more resources

Feedback