Date of Award

Spring 2019

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Biomedical Engineering

First Advisor

Beardsley, Scott

Second Advisor

Schmit, Brian

Third Advisor

Scheidt, Robert

Abstract

Electroencephalography (EEG) is a non-invasive technique used to measure brain activity. Despite its near ubiquitous presence in neuroscience, very little research has gone into connecting the electrical potentials it measures on the scalp to the underlying network activity which generates those signals. This results in most EEG analyses being more macroscopically focused (e.g. coherence and correlation analyses). Despite the many uses of macroscopically focuses analyses, limiting research to only these analyses neglects the insights which can be gained from studying network and microcircuit architecture. The ability to study these things through non-invasive techniques like EEG depends upon the ability to understand how the activity of individual neurons affect the electrical potentials recorded by EEG electrodes on the scalp. The research presented here is designed to take the first steps towards providing that link.Current dipole moments generated by multiple multi-compartment, morphologically accurate, three-dimensional neuron models were characterized into a single time series called a dipole response function (DRF). We found that when the soma of a neuron is directly stimulated to threshold, the resulting action potential caused an excess of current which backpropagated up the dendritic tree activating voltage gated ion channels along the way. This backpropigation created a dipole which had a magnitude and duration greater than the current dipoles created by neurons that were synaptically activated to near threshold. Additionally, we presented a novel technique, where, through the combination of the DRFs with point source network activity via convolution, dipoles generated by populations of neurons can be simulated. We validated this technique at multiple spatial scales using data from both animal models and human subjects. Our results show that this technique can provide a reasonable representation of the extracellular fields and EEG signals generated in their physiological counterparts. Finally, analysis of a simulated evoked potential generated via the convolutional methodology proposed showed that ∼ 98% of the variability of simulated signal could be accounted for by the dipoles originating from DRFs of spiking pyramidal cells.

COinS