Autocorrelation and Fourier Analysis for Detecting Periodic Cell Potentials in a Simulated Inhibitory Neural Network
Loading...
Authors
Rohrkemper, Robert R., Jr.
Issue Date
2003-12-23
Type
Thesis
Language
en_US
Keywords
theta , hippocampus , MS-GABA , periodicity , interneurons , Com- putational Neuroscience , GABAergic , cholinergic
Alternative Title
Abstract
The field of Computational Neuroscience is best described as a mathematical approach to the study of neural systems reducing neural systems to a set of computational tasks. Computer models are an important part of this analysis for their
insight into the phenomenon.
In order to begin a meaningful study, one has to find a pattern that exists
throughout the brain. Experimental studies have shown that isolated neurons
can be naturally periodic repeating almost "clock-like". However, due to inter-neuron interactions, signals from cells in networks are neither perfectly periodic
nor completely random. The goal of my analysis was to characterize periodic
behavior of inhibitory neurons.
I created a network model of a hundred neurons, in which I could influence
the percentage of periodic neurons by changing the number of connections or their
strength. During the simulations, membrane potential traces were recorded from
each of the one-hundred cells. Then, I developed three methods for the detection
of periodic neurons.
The first method was the Autocorrelation Function (AC), which I computed
using Matlab. I improved this method by modifying several functions, allowing for
the removal of initial transients in each simulation. AC's tells us how well a signal
correlates with itself when one copy is shifted by a time ¢t. Because an animal's
behavior has correlates with activity in the µ frequency (4¡10Hz), we filtered out
other frequencies. For example, REM sleep correlates with µ activity. I then fitted
an exponential function to the local maxima of the AC. Decay constants of the
exponentials were recorded for each cell and the cell was identified as periodic when
the decay was greater than a threshold. Despite this well thought out method,
some cells were identified incorrectly, based on our visual inspection of the raw
data.
Therefore, I implemented a Fast Fourier Transform (FFT) algorithm. A Fourier
plot gives the activity of a signal for each frequency. I created a method to take
an integral of the FFT over the µ range. As with the AC method, cells with values
over a threshold were labelled periodic. My data show that the FFT method identifies periodic cells more accurately than the AC method. After a few initial tests,
this model was assumed to be biologically accurate for the purposes of my project.
It is my hope that the creation and implementation of more accurate methods will
allow for a further understanding of periodicity in the Medial Septum.
Description
ix, 72 p.
Citation
Publisher
License
U.S. copyright laws protect this material. Commercial use or distribution of this material is not permitted without prior written permission of the copyright holder.