Use this URL to cite or link to this record in EThOS:
Title: Performance comparison of functional and effective brain connectivity methods
Author: Spangler, Robert
ISNI:       0000 0004 9356 3506
Awarding Body: University of Glasgow
Current Institution: University of Glasgow
Date of Award: 2020
Availability of Full Text:
Access from EThOS:
Access from Institution:
Functional and effective connectivity estimates based on electroencephalography (EEG) and magnetoencephalography (MEG) are widely used to understand and reveal new insight into the dynamic behaviour of the brain. However, with a large number of different connectivity methods that are currently available, there is a lack of systematic comparative studies including a statistical evaluation of their performance to understand the strengths and shortcomings of competing methods. Here, we present a simulation framework to evaluate and compare the performance of connectivity estimators on simulated, yet realistic electromagnetic recordings. We assess the ability of various methods to reconstruct cortical networks, while systematically varying specific parameters which are of significant importance during the simulation, preprocessing or inverse source reconstruction of realistic EEG recordings. A decisive advantage of this simulation framework, when compared with models utilised in other studies, is the integration of volume conduction artifacts. This is achieved by modelling the propagation of electric or magnetic fields from an electric primary current source through biological tissue towards measurement sensors. Subsequently, inverse source reconstruction approaches are applied to estimate the temporal activity patterns of underlying network nodes. The implementation of these concepts enabled the analysis of parameters involved during forward modelling and source reconstruction which may affect the estimation of connectivity on the source level. The experiments carried out in this work unfold the behaviour of estimators regarding the effect of signal-to-noise ratio (SNR), length of data sets, various phase shifts between correlated signals, the impact of regularization used in inverse source reconstruction, errors in the localization and varying network sizes. For each simulation, strengths and weaknesses of methods are pointed out. Furthermore, pitfalls and obstacles researchers might come across when applying particular estimators on EEG recordings are discussed. Building on the insight gained from simulation studies, the final part of the thesis analyses the performance of connectivity estimators when applied to resting-state EEG recordings. Network reconstructions with priority on the alpha frequency band reveal a default-mode-network (DMN) with dominant posterior-to-anterior information flow. We detected no significant variations in the amount of correctly identified network links between connectivity methods. However, we discuss differences in connectivity spectra that emerged, which affect the interpretability and applicability of methods.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available
Keywords: RC0321 Neuroscience. Biological psychiatry. Neuropsychiatry