Blood oxygenation level dependent (BOLD) functional magnetic resonance imaging (fMRI) is widely used to assess local neuronal activity. However, the absolute metabolic demand related to BOLD-signal change is still largely unknown. In the current study we investigated the feasibility of multi-parametric quantitative BOLD (mq-BOLD) measurements to detect changes in tissue oxygenation in healthy, cortical gray matter during different states of visual stimulation. Our results show that mq-BOLD measurements are well feasible for standard activation paradigms. However, subject motion and transient states of brain activation might impact on quantitative parameter mapping due to longer acquisition times.
Introduction
BOLD-signal changes in fMRI are relative with respect to an unknown metabolic baseline, and scale non-linearly with energy metabolism.1,2 Thus, absolute metabolic demands, i.e. the cerebral metabolic rate of oxygen (CMRO2), across brain regions or task load are still largely unknown.
Established MRI-methods to assess absolute CMRO2 changes are based on gas-calibration,3,4 but gas-free approaches have gained traction in recent years due to their ease of application.5-9 Gas-free multi-parametric quantitative BOLD (mq-BOLD) imaging measures oxygenation by integrating susceptibility-related transverse relaxation rate R2’ and cerebral blood volume10,11. CMRO2 is obtained when adding cerebral blood flow (CBF).12 This method was successfully applied in patients,12-15 but its reliability to assess metabolic variability across cognitive states in healthy subjects is less clear.
In the current study, we tested the sensitivity of mq-BOLD for task-related
changes in tissue oxygenation in healthy, cortical gray matter during different
states of visual stimulation by performing quantitative multi-echo gradient
echo (multi-GE) and multi-echo spin echo (multi-SE) measurements of intrinsic,
effective and susceptibility-related transverse relaxation rates R2, R2* and
R2’. To assess plausibility of observed quantitative condition-related differences,
changes in effective transverse relaxation rate ΔR2*(BOLD-TC)
were also derived from BOLD-EPI time courses of a functional localizer
experiment (task-fMRI).
Methods:
We performed whole-brain functional imaging of twelve healthy subjects (age 30.8±6.8y; five men) during randomized conditions of eyes closed (closed), eyes open (open), and during visual stimulation with flickering (8Hz) checkerboard (checker) using a 3T Elition MR scanner (Philips Healthcare, Best, Netherlands) and a 32-channel head-coil. We acquired T2-weighted multi-spin-echo (multi-SE), T2*-weighted multi-gradient-echo (multi-GE), and pseudo-continuous arterial spin labeling (pCASL) data averaged across prolonged condition blocks (avg.duration: 7min). Additionally, we acquired classical task-fMRI data (30sec block-design) employing T2*-weighted multiband-EPI and structural T1-weighted data (MPRAGE; voxel size 1x1x1mm3) (see Fig.1 for experimental details).
Data analysis used SPM1216 and in-house Matlab17 programs. We applied motion correction and spatial coregistration, omitting spatial normalization and smoothing of task data to reduce voxel blurring. Task-fMRI: We identified voxels with significant BOLD-signal changes from EPI-time series using a general linear model and created individual volumes-of-interest (VOIs) for each condition based on thresholded t-maps (t>10 for checker>open, and open>closed, respectively). To minimize influence of CSF partial volume and susceptibility artifacts, voxels were removed from VOIs using thresholds applied to T2 and R2’ maps (excluding thresholds: T2>100ms, R2’>10sec-1). mq-BOLD: We obtained intrinsic R2 (=1/T2), effective R2* (=1/T2*) and susceptibility related R2’ (=R2*-R2) from multi-SE and multi-GE data by mono-exponential fits11,18 and corrected T2* for macroscopic background gradients19,20 and motion.21,22 ΔR2*(BOLD-TC): We calculated changes in effective R2* (ΔR2*=ΔS/S0/TE) from dynamic EPI-time series signal changes (ΔS=S-S0, baseline S0). CBF maps were calculated from pCASL data following Alsop et.al.23
Results:
Task-fMRI-based functional localizer experiments yielded robust activation in each individual subject (Fig.2a). Contrasts ‘open>closed’ and ‘checker>open’ identified largely distinct areas of visual cortex, where strong responses in central areas were related to ‘checker>open’, while ‘open>closed’ elicited weaker more lateral activation. VOI-average ΔR2*(BOLD-TC) time courses demonstrate reliable changes across all conditions for each subject (Fig.2b-d).
Fig.3 shows an exemplary slice of all imaging modalities with comparable
data quality across conditions. Visual inspection already suggests higher CBF in
visual cortex for ‘open’ and ‘checker’. VOI analysis of mq-BOLD parameters
(Fig.4, Table1) yielded a significant increase of CBF in both contrasts, though
weaker for ‘open>closed’. We also found significantly decreased R2’ (driven
mainly by decreased R2*) upon checkerboard stimulation, but not for
‘open>closed’. Please note consistent CBF and R2* changes for each
individual (Fig.4a,c) for visual stimulation yet higher variability for
‘open>closed’.
Discussion:
In regions of significant visual cortex activation
(task-fMRI), we found significant differences for CBF, R2’ and R2*
(checker>open) and for CBF (open>closed) (Fig.4, Table1), according with literature.5,24-26 We compared these findings with ΔR2*(BOLD-TC) from the task-fMRI localizer experiments. On the
individual subject level (Fig.4), we found consistent changes for CBF and R2’
(diverging for 0/12 and 3/12) after visual stimulation. The more subtle
condition of ‘open>closed’ did not induce reliable changes of R2’ and R2*,
yet significant changes for CBF and ΔR2*(BOLD-TC) from
EPI-BOLD data. This suggests, that either a) transient activation is only
captured by shorter blocks of task-fMRI, but not during extended blocks
necessary for mq-BOLD, or b) that GE-based T2* measurement are more prone to
motion artefacts, which cannot always sufficiently be corrected.21,22
Conclusion:
Our results demonstrate that mq-BOLD measurements of activation-related changes are well feasible for standard visual stimulation experiments. However, subject motion and transient states of brain activation might impact on image quality due to longer measurement times. These challenges need to be addressed by study design and improved head immobilization.