Use this URL to cite or link to this record in EThOS:
Title: Predictability of epileptic seizures by fusion of scalp EEG and fMRI
Author: Jing, Min
ISNI:       0000 0004 2751 5733
Awarding Body: Cardiff University
Current Institution: Cardiff University
Date of Award: 2008
Availability of Full Text:
Access from EThOS:
Access from Institution:
The systems for prediction of epileptic seizure investigated in recent years mainly rely on the traditional nonlinear analysis of the brain signals from intracranial electroencephalograph (EEG) recordings. The overall objective of this work focuses on investigation of the predictability of seizure from the scalp signals by applying effective blind source separation (BSS) techniques to scalp EEGs, in which the epileptic seizures are considered as independent components of the scalp EEGs. The ultimate goal of the work is to pave the way for epileptic seizure prediction from the scalp EEG. The main contributions of this research are summarized as follows. Firstly, a novel constrained topographic independent component analysis (CTICA) algorithm is developed for the improved separation of the epileptic seizure signals. The related CTICA model is more suitable for brain signal separation due to the relaxation of the independence assumption, as the source signals geometrically close to each other are assumed to have some dependencies. By incorporating the spatial and frequency information of seizure signals as the constraint, CTICA achieves a better performance in separating the seizure signals in comparison with other conventional ICA methods. Secondly, the predictability of seizure is investigated. The traditional method for quantification of the nonlinear dynamics of time series is employed to quantify the level of chaos of the estimated sources. The simultaneously recorded intracranial and scalp EEGs are used for the comparison of the results. The experiment results demonstrate that the separated seizure sources have a similar transition trend as those achieved from the intracranial EEGs. Thirdly, simultaneously recorded EEG and functional Magnetic Resonance Imaging (fMRI) is studied in order to validate the activated area of the brain related to the seizure sources. An effective method to remove the fMRI scanner artifacts from the scalp EEG is established by applying the blind source extraction (BSE) algorithm. The results show that the effect of fMRI scanner artifacts has been reduced in scalp EEG recordings. Finally, a data driven model, spatial ICA (SICA) subject to EEG as the temporal constraint is proposed in order to detect the Blood Oxygen-Level Dependence (BOLD) from the seizure fMRI. In contrast to the popular model driven method General Linear Model (GLM), SICA does not rely on any predefined hemodynamic response function. It is based on the fact that brain areas executing different tasks are spatially independent. Therefore SICA works perfectly for non-event-related fMRI analysis such as seizure fMRI. By incorporating the temporal information existing within the EEG as the constraint, the superiority of the proposed constrained SICA is validated in terms of better algorithm convergence and a higher correlation between the time courses of the component and the seizure EEG signals as compared to SICA.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available