EEG-fMRI: novel methods for gradient artefact correction

The general aim of the work detailed in this thesis is to improve the quality of electroencepholography (EEG) recordings acquired simultaneously with functional magnetic resonance imaging (fMRI) data. Simultaneous EEG-fMRI recordings offer significant advantages over the isolated use of each modality for measuring brain function. The high temporal resolution associated with EEG complements the high spatial resolution provided by fMRI. However, combining the two modalities can have significant effects on the overall data quality. The gradient artefact (GA), which is induced on the EEG cables by the time varying magnetic fields associated with fMRI sequences, can be particularly problematic to correct for in experiments containing any subject movement. In this thesis, two novel, movement-invariant methods are introduced for correcting the GA. The first method is named the gradient model fit (GMF) and relies upon the assumption that the GA can be modelled as a linear combination of basis components, where the relative weighting of each component varies dependent upon subject position. By modelling these underlying components, it is possible to characterise and remove the GA, which is particularly beneficial in the presence of subject movement. The second method named the difference model subtraction (DMS) relies on the assumption that the GA varies linearly for small changes in subject position. By modelling the change in GA for a basis set of likely head movements, it was shown to be possible to combine DMS with standard GA correction methods to improve the attenuation of the GA for data acquired during subject movement. Both methods showed a significant improvement over the existing GA correction techniques, particularly for experiments containing subject movement. These methods are therefore relevant to any experimenter interested in working with subject groups such as children or patients where movement is likely to occur