Computational Neuroscience Methods: Investigating the Relationship between Resting-State EEG Microstate Syntax and fMRI BOLD Signal

Simultaneous recording of Electroencephalography (EEG) and functional Magnetic Resonance Imaging (fMRI) has been used consistently in the past as a means of understanding EEG microstate function. EEG microstates are quasistable states of EEG activity. Here, I investigate existing methodologies that attempt to draw relationships between microstate classes and fMRI signal, shedding light on their limitations and proposing alternative methods which may better utilise the advantages of simultaneously recorded EEG-fMRI. Three distinct studies are presented, each using a novel methodology which compares EEG microstates to the simultaneously recorded fMRI signal in resting state recordings. Each proposed method could be used and developed upon in the future to address gaps in the existing literature. The first study shows how EEG microstate n-grams exhibit varied durations and frequencies in some participants during concurrent fMRI Co-Activation Patterns (CAPs). The second study employs a random forest regressor model, utilising microstate n-gram parameters as features per fMRI time point in a sliding window, attempting to predict patterns in fMRI activity in a low dimensional space. In the third study, the focus shifts to conceptualising the EEG signal as a continuous signal rather than sequence of microstates, with analysis of microstates occurring post-hoc; a novel means of investigating microstates which has not yet been attempted. I also show how existing investigations of microstate syntax may benefit from adjustments to their processing pipelines in order to better retain the information apparent in EEG microstate sequences.