Sources of signal fluctuations in fMRI at 7 Tesla

The development of high-field magnetic resonance imaging (MRI) systems has allowed for critical improvements in image signal-to-noise ratio (SNR), potentially leading to higher sensitivity and spatial resolution for functional MRI (fMRI) techniques. However, recent studies have shown that these potential advantages become significantly compromised by increased signal fluctuations arising from correlated noise sources, namely physiological processes. This work comprises the analysis of blood oxygenation level-dependent (BOLD) fMRI data acquired at 7 Tesla, utilizing a standard 2D echo planar imaging (EPI) and a segmented 3D echo volumar imaging (EVI) technique, from healthy subjects at rest or submitted to a visual localizer (Loc) paradigm. A principal component analysis (PCA)-based approach and a physiological regressor (PR)-based approach were investigated for correlated noise correction. As expected, 3D data displayed significantly higher spatial SNR (sSNR) than 2D data, but comparable temporal SNR (tSNR); the information explained by general linear models was similar in both cases, and 3D analyses exhibited lower sensitivity to neural activation. In general, both correction approaches produced significant increases in tSNR, explained information, and detection sensitivity for both acquisition techniques, but especially for segmented EVI. The PCA-based approach was the most effective correction method; however, the PR-based approach allowed the characterization of the relative contributions from different noise sources, confirming significant relative increases in physiological noise from 2D to 3D acquisitions. In summary, although greater physiological noise contributions were found in segmented EVI data acquired at 7 Tesla, these were shown to be adequately compensated for by appropriate noise correction methods.