Introduction to Neural Interfaces - Brain-Machine Interface Engineering - page 10

being collected by the scalp electrode, the signal is attenuated and spatially filtered by the pia, dura,
the corticospinal fluid, bone, and scalp. It is also contaminated by many electrical artifacts of bio-
logical or nonbiological origins.
Alternatively, if the electrodes are invasively placed at the surface of the cortex as in ECoG,
the SNR is much better and the useable bandwidth increases up to at least 1 kHz (as shown in
Figure 1.5 c) using electrodes of 4 mm in diameter [ 47 ]. Theoretical analysis outlined by Freeman
[ 32 , 48-52 ] and Nunez [ 53-55 ] has identified the utility of ECoG potentials and attempts to
explain how to extract the relevant modulation of neural assemblies. Closely spaced subdural elec-
trodes have been reported to measure the spatially averaged bioelectrical activity of an area much
smaller than several square centimeters [ 56 ]. The production of potentials is because of the super-
position of many aligned and synchronous dipole sources [ 53 ]. The coactivation of sources is related
to neural “synchrony” and is used to describe the amplitude modulations in extracellular recordings
that occur during state changes [ 36 ]. The fluctuating cortical potentials have been associated with
traveling waves of local dendritic and axonal potentials [ 57 ]. The results in Reference [ 57 ] indicate
that at least two separate sources of signal coherence are produced either through the action of
short-length axonal connections or the action of long distance connections. Synchronous activity
can occur at different spatial scales and over time lags [ 53 ], which requires wide spatial sampling of
the cortex and analysis over fine temporal scales.
Being a multiscale system, brain activity can also be collected at an intermediate level, which
targets the neural assembly and is named the LFP as shown in Figure 1.5 b. LFPs are continuous
amplitude signals also in the frequency range of 0.1-200 Hz created by the same principles as the
EEG, but they are normally collected with higher impedance microelectrodes with cross-sectional
recording diameters of 20-100 µm. Therefore, the spatial averaging operated by the larger EEG/
ECoG electrodes is not present. The LFPs differ in structure depending where the microelectrode
tip is placed with respect to the dendritic or axonal processes of the neural assembly and therefore
better describe the electrical field fluctuations in a more localized volume of tissue. In the frequency
domain, LFP tuning has been shown to be manifested also, and differentially, in oscillatory activi-
ties in different frequency ranges. Oscillations and spiking in sensorimotor cortex in relation to mo-
tor behavior in both humans [ 36 , 58 ] and nonhuman primates [ 59-61 ] in the 15- to 50-Hz range
were found to increase in relation to movement preparation and decrease during movement execu-
tion [ 62 ]. Other researchers have also described the temporal structure in LFPs and spikes where
negative deflections in LFPs were proposed to reflect excitatory, spike-causing inputs to neurons in
the neighborhood of the electrode [ 63 ].
At the microscopic scale, the electrical activity captured by microelectrode arrays placed
closer to the neuron cell body displays a pulse (called a spike) whenever the neuron fires an ac-
tion potential shown in Figure 1.5 b. Spikes are approximately 1 msec in duration, 30-200 µV in