Biomedical Engineering Reference
In-Depth Information
over time. For example, the sequence of firings in the right side of Fig. 9.5 have been separated into four
bins.
9.2.5 AGraphical Representation
As extracellular potentials are relatively simple to record, it is often the case that many hundreds of
electrodes are used. The result is hundreds of φ e recordings from potentially thousands of neurons. After
sorting and binning, information is often represented graphically in one of two ways. First, the individual
spikes may be shown in a matrix, or raster , where each row represents a single neuron (Fig. 9.5, right panel).
A raster plot shows all of the raw data but can be difficult to interpret. As information is encoded as a
firing frequency, the timing of an individual spike is less important. The second graphical representation,
therefore displays firing rate and as a color (Fig. 9.5, right panel). Again, each row corresponds to a single
neuron. The advantage of the color coded map is that the firing frequency of different neurons may be
correlated more easily. It is often the case that many trials may be recorded of the same phenomenon and
the frequency results averaged before displaying the data graphically.
Spikes
Frequency Matrix
bin1
bin2
bin3
bin4
Figure 9.5: Graphical binning and frequency matrix.
9.2.6 Windowing
There are a number of potential problems with time binning. For example, consider the impact of slightly
offsetting the four bins in Fig. 9.5. If the firing frequency changed suddenly, the change may be averaged
out over two neighboring bins. A simple solution would be to add more time bins, each with a shorter
duration. As the bin size is made smaller, however, there will be less possible spikes within each bin to
count and the range of possible frequencies will be limited. On the other hand, if the bin size is large, the
frequency curve will be much more discrete. Amore clever method of binning is to wait for some number
of spikes (e.g., 10 or 20) and count the time that elapses. In this way, the bin size adapts to the rate of
firing. For fast firing the bin is small and for slow firing the bin expands. The problem with adjustable
bin sizes is that it becomes difficult to correlate the firing patterns of different neurons.
A more dynamic method of computing firing frequency is to define a single bin of some duration
( t ) and slide it along the spike train. This technique is known as windowing . For each slide, the number
of spikes can be counted and used to compute a frequency for that particular slide time. The simplest
window may be defined as
Search WWH ::




Custom Search