Thesis defense Daniel Gomez (Donders series 396)

Functional magnetic resonance imaging (fMRI) is a non-invasive tool used to study brain activity in humans while they perform tasks or rest. Despite being used routinely within academia, the fMRI signal is hard to interpret because it is confounded by non-neuronally driven fluctuations, such as those arising from the brain physiology or MRI artifacts. Because these fluctuations may bias or preclude a proper statistical analysis of fMRI, different acquisition sequences and denoising strategies have been proposed in the literature, yet it remains unclear which of these techniques and strategies are better suited for task and resting-state fMRI.

Here we implement and compare two promising acquisition strategies, a highly-accelerated Multiband EPI and a contrast-rich Multiband-Multiecho EPI, and compare different denoising strategies. We find evidence that that Multiband EPI with proper denoising generally performs better than Multiband-Multiecho as judged by a comparison metrics that measure fMRI sensitivity, specificity and reproducibility. These comparisons are done in both task and resting-state fMRI, and suggest that optimising for faster sampling (as in the Multiband EPI) may be better than optimising for improved contrast (as in Multiband-Multiecho EPI).

In chapter 4 we use an ultra-fast fMRI sequence to investigate temporally independent functional modes (TFMs) of brain activity at the single-subject level. Previously this was considered unpractical because of the limited number of samples obtained in conventional fMRI. Our results suggest that TFMs are feasible and reproducible at the single-subject level, opening new possibilities for investigating functional networks and their integration.

We anticipate our results to be of interest for researchers designing acquisition protocols for fMRI experiments (Chapters 2 and 3) and to be a starting point for further research on temporal functional modes of brain activity (Chapter 4).