Biomedical Engineering Reference
In-Depth Information
but you may be able to gain some information about who is a leader, who is a follower, when arguments
occur and when they are resolved. All the while, the noise is still in the background. Now, consider that
you perform this task, day after day, week after week. Eventually, you would begin to recognize patterns
in the language and pick up on subtleties of the conversations.
Eavesdropping on a foreign language class is similar to our current status in decoding neural
patterns. We understand parts of the neural language, and can gain some idea of when a population of
neurons is working on a problem, but on the whole they speak a language that is very foreign to us.
9.2.3 Spike Sorting
The prospect that a recording from one electrode contains information from many neurons is another
step toward decoding neural information.The next step is to separate out the firing of individual neurons.
Typically this separation is called
spike sorting
. The idea is to create a bin for each neuron and then sort
the timing of firings into that bin. While there are many methods for accomplishing binning the most
straightforward method is to define a number of
template deflections
that correspond to specific neurons
firing (Fig. 9.4). This is an important step for two reasons. First, the number of template deflections will
be the same as the number of bins and therefore makes an assumption about how many neurons are
contributing to
φ
e
. If there are two few or two many bins, some spikes will be misclassified. Second, the
templates will serve as the ideal to which other deflections will be compared. Although the templates
may be determined by eye, there do exist some methods of automating the selection of templates using
statistical differences in the signals.
Figure 9.4:
Spike sorting of extracellular deflections.
9.2.4 Time Binning and Firing Frequency
Once the timing of each neural firing has been extracted, the firing rate (frequency) can be found. It is
often the case, however, that a neuron fires in a burst and then remains quiet, followed by another burst.
To detect the change in frequency, the spikes may be binned in time. In
time binning
, all spikes are counted
within a bin that spans a given length of time. By counting the number of spikes that occur within the
time bin, an average firing rate may be computed. Changes in firing frequency can therefore be tracked
