Skip to main content

EEG-Based Mental States Identification

Witon, Adrien J-B. (2019) EEG-Based Mental States Identification. Doctor of Philosophy (PhD) thesis, University of Kent,.

PDF
Download (6MB) Preview
[img]
Preview

Abstract

In this thesis, we focus on the identification of mental states described according to the definition of awareness and wakefulness. Using algorithmic methods, we show that it is possible to differentiate between two brain states based on the brain electrical activity collected by EEG. We begin by explaining the overall theoretical framework which enabled us to develop the detection of brain states. It starts with data acquisition. Following that, we analyse the pre-processing of the data before the extraction of features. Finally, we go on to statistically evaluate the results. In order to achieve this task, we propose four experiments. We will first focus on exploring different brain states for patients in Intensive Care Unit (ICU) such as coma and quasi-brain-death states. To distinguish these states, we use a signal processing method based on the EEG signal phase. A phase synchrony index based on Shannon entropy was used to separate the two states. Statistical validation revealed a significant difference between the two via delta-alpha and theta-alpha frequency couplings. Next, we studied the neuronal mechanisms which is used to understand consciousness. We did that by using dipole modelling. This method was applied to local-global experiment and the paradigm of auditory deviance with two hierarchical levels. A modulation of this experiment is generated by a sedative Propofol to study the effect on conscious states. This experiment was analysed in greater detail using the Imaging Method to do the source localisation. We analysed three different time-windows. The first window corresponds to the local effect during the initial response of the brain. We assume that this input is related to auditory areas and activates the temporal lobe. The second window is at the interface between the local effect and the global effect. In here we are especially interested in the interaction between these two effects during the second window. Finally, the third window will enable us to study the overall effect. We hypothesize a global activation of neural networks corresponding to consciousness as described by the global workspace theory. The third experiment focused on brain states high-level athletes experience during a cognitive task. Two different groups of cyclists, endurances and sprinters, were asked to do a Stroop task for 30 minutes. We studied the influence of the task and the potential differences in brain activity between the two groups. We found through the frequency analysis that the brain activity between the two groups can be distinguished, but was not modified by the cognitive task. Finally, we studied the influence of the sensorimotor loop on the brain. A physical task was applied, consisting in lifting a weight with two measurements, where the lifting arm can also be in fatigued state. Using sources reconstruction from EEG, we studied the impact of weight-lifting and the physical fatigue upon neuronal activities and the neuronal origins of these effects. We found that only weight has an effect, whereas fatigue effect is not significant. We conclude with a discussion of the mechanisms of consciousness analysed via algorithmic methods and some future work for the possibility to distinguish better between different cognitive states.

Item Type: Thesis (Doctor of Philosophy (PhD))
Thesis advisor: Li, Ling
Thesis advisor: Bowman, Howard
Uncontrolled keywords: EEG connectivity phase-synchrony 'Shannon entropy' 'mental states' cognition
Subjects: Q Science > QA Mathematics (inc Computing science)
Divisions: Faculties > Sciences > School of Computing
SWORD Depositor: System Moodle
Depositing User: System Moodle
Date Deposited: 02 Apr 2019 15:10 UTC
Last Modified: 03 Jun 2019 09:39 UTC
Resource URI: https://kar.kent.ac.uk/id/eprint/73337 (The current URI for this page, for reference purposes)
  • Depositors only (login required):

Downloads

Downloads per month over past year