Biomedical Engineering Reference
In-Depth Information
in part due to the complications induced by neuroanatomy. In the
cerebral cortex, neurons form thousands of synapses with other
cortical neurons in neighboring and distant brain areas and have
descending and recurrent connections to subcortical systems
(6)
.
Moreover, cortical neurons are typically embedded in a processing
unit such as a column or local microcircuit
(7)
. As such, cortical
neurons can have distinct non-synaptic functional relationships
with neurons within and outside of their local processing units.
There have been mixed opinions in the literature about
whether functional interactions between neurons underlie the
encoding of behaviorally relevant information. Some studies have
argued that interactions have little impact on information pro-
cessing by neurons
(8-12)
. Other studies have presented strong
evidence that interactions between neurons may be a central
feature of neural coding
(3, 13-19)
. To establish the relevance
of functional interactions to understanding how populations
of neurons control behavior, neuronal population data must
be studied with a variety of techniques and must be studied in
awake, behaving animals using behavioral tasks that challenge the
brain areas of interest (see Refs.
(20, 21)
for an example of this
approach).
Here we describe seven distinct techniques for inves-
tigating functional interactions between neurons: Functional
grouping by principal component analysis, cross-correlation,
joint-peristimulus time histograms, rate correlations, trial-by-trial
rate correlations, predictive interactions, and network interac-
tions. We also establish the significance of interaction based on
time-shuffling or trial-shuffling spike trains. We then apply each
of these techniques to a population of simultaneously recorded
neurons from the rodent dorsomedial prefrontal cortex (dmPFC)
and motor cortex. We use each of these techniques to generate
insights into how neurons interact; and then we combine these
techniques to derive a portrait of neuronal interactions across each
cortical area.
2. Methods
2.1.
Neurophysiological
Recordings
The multi-electrode data used here were collected from the
rodent dorsomedial prefrontal cortex (dmPFC; comprising pre-
limbic and anterior cingulate cortex) and motor cortex of one
rat during performance of a delayed-response task with a fixed
1.0 s delay
(20)
. In this task, animals had to initiate and main-
tain a lever press for 1.0 s, and then release the lever promptly
to get a liquid reward (
Fig. 7.1A
). Twenty-one neurons (11
in motor cortex, 10 in dmPFC) were isolated from one animal
during delayed-response performance. Methods used to acquire
