Purchase this article with an account.
Maha Adamo, Jarnet Han, Frank Haist; Denoising developmental FMRI data: Removal of structured noise from a passive-viewing task differentially impacts children and adults. Journal of Vision 2011;11(11):417. doi: https://doi.org/10.1167/11.11.417.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
When collecting functional magnetic resonance imaging (FMRI) data across multiple populations, inherent differences in physiological signals can distort comparisons of effects. For example, children differ from adults with respect to breathing and heart rates, which could differentially affect the blood oxygenation level dependent (BOLD) response. Furthermore, known differences in task-negative or default mode processing may affect comparisons of task-positive activity. We collected FMRI data during a blocked-design task involving passive viewing of faces, diverse objects, and scrambled stimuli, administered to over 70 participants ranging in age from 6 to 34 years old. We fine-tuned a processing methodology to meet the specific challenges of comparing data across age-groups by removing structured noise from FMRI data prior to standard individual regression analyses. We first applied independent components analysis (ICA) to decompose BOLD data into spatially independent patterns of activation, each with a variable timecourse. We then entered the timeseries of the resulting independent components (ICs) as multiple regressors against timeseries data extracted from regions of interest (ROIs) placed at known sites of physiological distortion. ICs that were significant predictors of any physiological ROI timeseries were excluded from the subsequent reconstruction of BOLD data, except those that showed a significant correlation (r > 0.2) with a gamma function modeling the task-related signal. Additionally, we identified ICs that were significantly anticorrelated (r < −0.2) with the task gamma function on the assumption that these components reflected default mode activity. Separate reconstruction of only these task-negative components verified that these signal sources involved regions known to be engaged in resting states, such as medial ventral prefrontal cortex, posterior cingulate, and lateral parietal cortex. Interestingly, however, removal of these task-negative components impacted the assessment of denoised whole-brain task-positive activity differently for children relative to adults. This finding has important implications for comparing task effects across development.
This PDF is available to Subscribers Only