Investigating information processing within the brain using multi-electrode array (MEA) electrophysiology data

Horton, Paul Michael (2011) Investigating information processing within the brain using multi-electrode array (MEA) electrophysiology data. Doctoral thesis (DPhil), University of Sussex.

PDF - Published Version
Download (9MB) | Preview


How a stimulus, such as an odour, is represented in the brain is one of the main
questions in neuroscience. It is becoming clearer that information is encoded by
a population of neurons, but, how the spiking activity of a population of neurons
conveys this information is unknown. Several population coding hypotheses have
formulated over the years, and therefore, to obtain a more definitive answer as to
how a population of neurons represents stimulus information we need to test, i.e.
support or falsify, each of the hypotheses. One way of addressing these hypotheses
is to record and analyse the activity of multiple individual neurons from the brain
of a test subject when a stimulus is, and is not, presented. With the advent of multi
electrode arrays (MEA) we can now record such activity. However, before we can
investigate/test the population coding hypotheses using such recordings, we need to
determine the number of neurons recorded by the MEA and their spiking activity,
after spike detection, using an automatic spike sorting algorithm (we refer to the
spiking activity of the neurons extracted from the MEA recordings as MEA sorted
data). While there are many automatic spike sorting methods available, they have
limitations. In addition, we are lacking methods to test/investigate the population
coding hypotheses in detail using the MEA sorted data. That is, methods that
show whether neurons respond in a hypothesised way and, if they do, shows how
the stimulus is represented within the recorded area. Thus, in this thesis, we were
motivated to, firstly, develop a new automatic spike sorting method, which avoids
the limitations of other methods. We validated our method using simulated and
biological data. In addition, we found our method can perform better than other
standard methods. We next focused on the population rate coding hypothesis (i.e.
the hypothesis that information is conveyed in the number of spikes fired by a pop-
ulation of neurons within a relevant time period). More specifically, we developed
a method for testing/investigating the population rate coding hypothesis using the
MEA sorted data. That is, a method that uses the multi variate analysis of variance
(MANOVA) test, where we modified its output, to show the most responsive subar-
eas within the recorded area. We validated this using simulated and biological data.
Finally, we investigated whether noise correlation between neurons (i.e. correlations
in the trial to trial variability of the response of neurons to the same stimulus) in
a rat's olfactory bulb can affect the amount of information a population rate code
conveys about a set of stimuli. We found that noise correlation between neurons
was predominately positive, which, ultimately, reduced the amount of information
a population containing >45 neurons could convey about the stimuli by ~30%.

Item Type: Thesis (Doctoral)
Schools and Departments: School of Engineering and Informatics > Informatics
Subjects: Q Science > QM Human anatomy
Q Science > QZ Psychology
Depositing User: Library Cataloguing
Date Deposited: 13 Jun 2011 13:41
Last Modified: 14 Aug 2015 13:42

View download statistics for this item

📧 Request an update