Effects of MR-Parameters on Relative BOLD-Sensitivity in the Striatum
Poster No:
2330
Submission Type:
Abstract Submission
Authors:
Michael Marxen1, Marco Bottino1, Michael Smolka2
Institutions:
1Technische Universität Dresden, Dresden, Saxony, 2Department of Psychiatry and Psychotherapy, Technische Universität Dresden, Dresden, Germany
First Author:
Co-Author(s):
Michael Smolka
Department of Psychiatry and Psychotherapy, Technische Universität Dresden
Dresden, Germany
Department of Psychiatry and Psychotherapy, Technische Universität Dresden
Dresden, Germany
Introduction:
The striatum is a subcortical brain region involved in many brain functions such as reward processing, cognition, learning or salience detection. Faster fMRI acquisition techniques, such as multi-band (MB) imaging or sensitivity encoding, are often considered "state-of-the-art" nowadays and dominate fMRI research currently. However, Srirangarajan et al. (Srirangarajan et al., 2021) have already pointed out that the detection of mesolimbic reward responses suffers from the reduced BOLD sensitivity of such techniques in particular brain regions, especially the ventral striatum (nucleus accumbens) region, which is both susceptible to B0 field inhomogeneities as well as far from the receiver coils and thus suffers from reconstruction g-factor noise. Both of these effects are not observable in whole-brain average signal-to-fluctuation-noise (SFNR) but require a regional analysis of SFNR and depend on numerous MR acquisition parameters. Here, we will investigate the effect of MB factor, iPAT factor (skip of k-space lines), TE, and resolution on whole brain SFNR and relative regional SFNR for bilateral nucleus accumbens (NA), putamen (PU), pallidum (PA) & caudate (CA) motivated by our own observation of the SFNR drop in NA when using a sequence (UKB sequence) modeled after the UK-Biobank protocol with the primary alteration of reducing the MB factor from 8 to 6 (Fig. 1).
Methods:
Single-subject resting state data (>=150 frames, Siemens Prisma Fit 3T) are compared for 2.4 mm isotropic and 3x3x2 mm3 (gap 1mm) sequences with variable [TR]-[TE]-[MB factor]-[iPAT factor] (Echo spacing: 0.57-0.58ms for 2.4mm except UKB 0.69ms; 0.49-0.5ms for 3mm). While the UKB sequence had a long TE=38ms and iPAT=1, most other sequences had iPAT=2 and TE=25 or 30ms to reduce susceptibility-induced signal drop-out. All images underwent standard preprocessing, including realignment, field map correction, and T1-based normalization, and were resampled to 2.4mm isotropic resolution. Signal time courses were detrended using a 2nd order polynomial. SFNR was computed as the voxel-wise temporal Mean/Std. Median SFNR values were extracted for all regions of the AAL3.1 atlas and the union of all AAL regions (whole-brain). Relative regional SFNR was computed as the regional/whole-brain SFNR.
Results:
Results are presented in Fig. 2. Whole-brain SFNR magnitude a) decreases only marginally with MB when correcting for TR; b) decreases with TE, but BOLD-sensitivity may increase and is optimal at TE=T2*; c) increases only marginally with voxel volume, likely due to physiological noise; d) decreases substantially for iPAT=2 compared to iPAT=1.
Relative regional SFNR e) is particularly low in NA and PA; f) drops substantially for iPAT = 2 compared to 1; g) drops with MB, especially 2->3; h) ranges from 41% (PA-2.4mm-605-30-6-2) to 97% (PU-3x3x2mm-1400-25-2-1).
Relative regional SFNR e) is particularly low in NA and PA; f) drops substantially for iPAT = 2 compared to 1; g) drops with MB, especially 2->3; h) ranges from 41% (PA-2.4mm-605-30-6-2) to 97% (PU-3x3x2mm-1400-25-2-1).
Conclusions:
1) Both multi-band as well as iPAT parallel imaging techniques reduce whole-brain SFNR.
2) Both techniques additionally decrease relative SFNR in subcortical brain regions such as the striatum. PA and NA are especially affected.
3) The cause for this heterogeneity in SFNR and thus BOLD sensitivity is likely multi-fold: reduced signal due to a) susceptibility drop-out (NA) and b) reduced T2* due to iron content (PA) and c) increased g-factor noise far from receiver coils.
4) Assuming that within-subject standard error of BOLD effects scales inversely with SFNR for the same MR sequence, a drop of SFNR by 50% would reduce an effect of d=0.36 to 0.2 assuming 75% within-subject, 25% between-subject noise initially.
5) As reported by Srirangarajan, use of parallel imaging for BOLD fMRI has substantially biased the literature towards cortical rather than subcortical effects.
6) We recommend NOT to use iPAT or MB > 2 for whole-brain or subcortical fMRI.
7) The effect of slice thickness and no MB needs to be explored further.
2) Both techniques additionally decrease relative SFNR in subcortical brain regions such as the striatum. PA and NA are especially affected.
3) The cause for this heterogeneity in SFNR and thus BOLD sensitivity is likely multi-fold: reduced signal due to a) susceptibility drop-out (NA) and b) reduced T2* due to iron content (PA) and c) increased g-factor noise far from receiver coils.
4) Assuming that within-subject standard error of BOLD effects scales inversely with SFNR for the same MR sequence, a drop of SFNR by 50% would reduce an effect of d=0.36 to 0.2 assuming 75% within-subject, 25% between-subject noise initially.
5) As reported by Srirangarajan, use of parallel imaging for BOLD fMRI has substantially biased the literature towards cortical rather than subcortical effects.
6) We recommend NOT to use iPAT or MB > 2 for whole-brain or subcortical fMRI.
7) The effect of slice thickness and no MB needs to be explored further.
Modeling and Analysis Methods:
Activation (eg. BOLD task-fMRI) 2
Novel Imaging Acquisition Methods:
BOLD fMRI 1
Keywords:
Basal Ganglia
Design and Analysis
fMRI CONTRAST MECHANISMS
FUNCTIONAL MRI
Limbic Systems
MRI PHYSICS
Sub-Cortical
1|2Indicates the priority used for review