| dc.description.abstract | Spatial spread of a particular brain signal can be defined as the area of the cortical tissue around
the recording electrode that contributes to the electrical activity recorded by the electrode. More
specifically, assuming brain signals to be a weighted sum of electrical activity of a pool of
neurons, spatial spread represents the spatial weighting function that primarily depends on the
properties of the recording electrode such as its size, impedance and location as well as some
properties of the brain tissue such as its conductance and filtering characteristics. Different
signals, depending on the frequency content, represent different types of neuronal activity. For
example, multi-unit activity (MUA), obtained by band-pass filtering the signal recorded from
a microelelctrode (tip diameter of a few microns) primarily represents the weighted sum of
action potentials, while the local field potential (LFP), obtained by low-pass filtering the same
signal, primarily represents summed synaptic activity. Another signal is Electrocorticogram
(ECoG), obtained by low-pass filtering the signal obtained from a macro-electrode (diameter
of 2.3 mm) placed subdurally on the surface of the cortex of epileptic patients for localization
of the seizure focus. These ECoG signals are used to determine the brain area that is responsible
for seizures, which is subsequently surgically removed. Accurate estimation of the spatial
spread of ECoG is therefore extremely important from a clinical perspective. Similarly,
accurate estimation of the spatial spread of LFP is important from a basic science perspective,
since these signals are now routinely used to study cognition and behavior, and also in braincomputer
interfacing applications. However, the spatial spread of ECoG is unknown, and that
of LFP is highly controversial. In the first two studies in this thesis, we investigate the spatial
spreads of LFP and ECoG.
Brain signals are often analyzed in the spectral domain where the slope of the power spectral
density (PSD), as well as oscillations that are observed as peaks in the spectra, can reveal
important information about the neural network. For example, gamma oscillations observed in
the 30-70 Hz frequency range has been associated with several high-level cognitive functions
such as attention, memory, perception etc. Further, the high-gamma activity observed as a
broadband in 60-250 Hz frequency range has shown to be correlated with the spiking activity.
These different signatures provide a robust measure to understand the brain dynamics at
different recording levels. In the third study, we compare the tuning properties of gamma
oscillations and high-gamma activity for different stimulus properties in LFP and ECoG.
In the first study, we examined whether different frequencies of LFP spread differently.
Recording from a microelectrode array implanted in the primary visual cortex (V1) of two
macaques, we estimated the LFP spread as a function of frequency. We found that LFP spread
is neither “low-pass” nor ‘all-pass” as suggested by previous studies but “band-pass” with
frequencies in the high-gamma (60-150 Hz) range spreading more than both lower (20-40 Hz)
and higher (>250 Hz) frequencies. Further, we found that this increase in high-gamma range is
mirrored by an increase in the phase coherency across neighboring sites in the same frequency
range.
Spatial spreads can be estimated by measuring the receptive field (RF) and multiplying it with
the cortical magnification factor, but this method overestimates the spatial spread because RF
size gets inflated due to several factors such as eye jitter, stimulus size and RF scatter. This
issue can be partially addressed by comparing the RFs of two measures (such as LFP and multiunit
activity). Therefore, in the second study, we estimated the spatial spread of ECoG by
simultaneously recording LFP and ECoG from the primary visual cortex (V1) of three
behaving monkeys using a specialized hybrid grid which consists of both ECoG electrodes and
a microelectrode array. We simultaneously mapped the RF responses of MUA, LFP, and ECoG
at several cortical sites and found that spatial spread of ECoG is surprisingly local (standard
deviation of ~1.5 millimeters, or a diameter of ~3 mm), only ~3 times the spread of the LFP,
even though the size of the electrode is several hundred times larger than the microelectrode.
Further, using a completely different approach, we estimated the spatial spread of ECoG by
comparing the slope of the PSD of LFP and ECoG for spontaneous activity (no stimulus
condition). We found that the slope of the ECoG was much steeper than LFP in the 20-100 Hz
frequency range. Next, using a simple model based on linear superposition, we simulated the
ECoG signal by averaging LFP signals over a progressively larger set of electrodes. We found
that around ~50 LFP electrodes that correspond to a 7x7 grid when averaged had the similar
slope as ECoG. The estimated spread in millimeters- 7 x 400 μm (inter-microelectrode
distance) = 2.8 mm was remarkably similar to the first approach.
Finally, in the last part, as an indirect measure, we investigated the spatial extent of ECoG by
comparing the high-gamma activity observed in LFP and ECoG signals. We simultaneously
recorded the LFP and ECoG signals for six different stimulus radii and computed the change
in high-gamma power as a function of stimulus size. We found that tuning curve for LFP and
ECoG were similar, with a maximum ECoG high-gamma power for the stimulus of 0.3° radii,
suggesting local origins of ECoG. Further, we compared the orientation preference of LFP and
ECoG for gamma oscillations. The preferred orientation for gamma oscillations in ECoG was
similar to the gamma oscillation in LFP even though the ECoG electrodes were widely
distributed in the cortex (center- to-center distance of 10 mm). This suggests that orientation
tuning of gamma is not location specific but monkey specific.
Overall, our results suggest that ECoG is a local signal which can provide a useful tool for
clinical purposes, cognitive neuroscience and brain-machine-interface applications | en_US |
