Synopsis

Keywords: fMRI, Magnetization transfer, laminar fMRI

The GRE-BOLD contrast used in laminar fMRI suffers from suboptimal specificity due to unwanted extravascular effects. Recently, efforts were made to increase specificity of GRE-BOLD by means of CBV-weighting using the magnetization transfer contrast. The specificity improvement depends on the amount of selective GM signal reduction. We investigate whether PUSH pTx pulses, designed to maximize MT contrast in a given area, can achieve a higher functional contrast. Our results suggest that when opting for high-power, off-resonant MT pulses, both PUSH and a CP²⁺ mode perform equally well in increasing the functional contrast at short TE compared to GRE-BOLD in V1.

Introduction

Recently, laminar fMRI experiments using GRE-BOLD in combination with MT-preparation¹⁠ indicate an increase in sensitivity toward the microvascular blood compartment by effectively performing a CBV-weighting and reducing unwanted extravascular effects from large unspecific venous vessels²⁠. The specificity improvement depends on the amount of GM signal reduction without blood signal loss achieved with MT, which is limited by SAR and RF power. In this preliminary work, we capitalize on existing parallel transmission (pTx) technology and modify the recently proposed PUSH pulses³⁠ to maximize the MT effect in a given ROI and investigate how this affect laminar fMRI profiles.

Methods

The sequence consists of an off-resonant MT-module followed by a center-out, multi-echo 3D FLASH readout.
Experiments were performed on 1 male subject (33 years old) at a 7T (MAGNETOM Terra, Siemens Healthcare, Erlangen, Germany) scanner used in pTx mode with an 8Tx/32Rx head coil (Nova Medical, Wilmington, USA). A 252x252x32 matrix were scanned with a monopolar readout at 4 different TEs ranging from 1.86 ms to 9.12 ms. The echo spacing was 2.42 ms, TR was 11.6 ms and FA was 6°. Both PE dimensions in the 3D readout were sampled center-out and the primary PE dimension was 3-fold undersampled and 7-fold segmented. The segment TR, i.e. time between two MT pulses, was 179 ms and volume TR was 40.1 s. We performed a flickering checkerboard fMRI experiment split into 6 pseudo-random runs, 2 without MT-preparation, 2 with MT-preparation using a 4 ms long Gaussian MT-pulse played out with CP²⁺ excitation mode to improve power efficiency in V1 (referred to as MT-prep in the following) and 2 runs using a 4 ms PUSH-1 pTx pulse (referred to as MT-PUSH). Both MT-pulses were played out at -1.5 kHz off-resonance and 98 dummy shots preceded each run to reach a steady state. The PUSH-1 pulse was designed to maximize B_1,rms in an ROI covering V1 within a SAR constraint of 90 % 1st level mode (Fig. 1). The amplitude of the Gaussian MT-pulse was scaled to match the SAR achieved with the PUSH pulse.
Structural data were acquired with an MP2RAGE sequence and processes with FreeSurfer to estimate WM and pial surface boundaries. Functional data were processed using SPM12 for motion correction and linear detrending. Coregistration was performed with ITK-SNAP. The processed functional data were equidistantly sub-sampled between WM and pial boundary to create laminar profiles for each echo. For each volume, the multi-echo profiles were extrapolated toward TE = 0 ms using nonlinear fitting in MATLAB before GLM fitting.

Results

Fig. 1A,B shows the predicted B_1,rms in V1 using the CP²⁺ (A) and the PUSH‑1 pulse (B) (mean ± std within black optimization volume below each panel) indicating a gain in maximum B_1,rms in the occipital pole. This translates into a more favorable MTR (Fig. 1D compared to 1C, MTR of GM in V1 below each panel).
In Fig. 2, smoothed maps of functional signal change (anatomical informed smoothing using Laynii^4⁠) for all measured echoes and conditions are shown. While at the 1^st measured TE, signal change for BOLD is visually absent, both MT preparations show signal change in GM. At longer TE, the BOLD response comes primarily from areas in and around large veins while both MT-prep and MT-PUSH are less biased toward those veins.
Profiles at rest in V1 are shown in Fig. 3 for BOLD (A), MT-prep (B) and MT-PUSH (C). The activation profiles (D-F) indicate less extravascular (EV) effects with both MT preparation methods compared to BOLD and a higher signal change in the parenchyma at the 0 ms TE and at the 1st measured TE.

Discussion

Similar to previous work^1,2⁠, the presented results indicate that MT preparation increases specificity by reducing EV contributions in GRE-BOLD (profiles in Fig. 2 are closer together) and increasing signal change in the parenchyma at very short TE. In this work, we achieved 25 – 30 % GM signal reduction with pulses played out at a larger frequency offset than in our previous work²⁠ and therefore reducing the amount of unwanted suppression of blood signals. This increased the functional contrast. The preliminary data in Fig. 3 suggests that a power matched CP²⁺ transmission mode performs on average equally well compared to a more dedicated PUSH pulse if V1 is the area of interest. However, PUSH adds the flexibility to increase the MT contrast in any given ROI resulting in a more favorable functional contrast. Future work on more subjects will be performed to investigate the generalizability of MT-PUSH for laminar fMRI.
In summary, PUSH pulses can potentially be used to achieve a higher MT contrast in a given ROI within SAR limits increasing the more specific functional contrast in MT-weighted laminar fMRI.

Acknowledgements

This work was funded by Mercur grant Ko-2021-0010 and DFG grant 432657511.