Acquisition and evaluation of sub-millimetre resolution GE-BOLD laminar fMRI datasets at 3T and 7 T

1881 

Abstract Submission 

Sriranga Kashyap¹, Prathamesh Chopade², Seong-Gi Kim³, Kamil Uludag^1,4,5

¹Krembil Brain Institute, University Health Network, Toronto, Ontario, Canada, ²Center for Neuroscience Imaging Research, Institute for Basic Science, Suwon, Korea, Republic of, ³Center for Neuroscience Imaging Research, Institute for Basic Science, Sungkyunkwan University, Suwon, Korea, Republic of, ⁴Department of Medical Biophysics, University of Toronto, Toronto, Ontario, Canada, ⁵Center for Neuroscience Imaging Research, Institute for Basic Science & Department of Biomedical Engineering, Sungkyunkwan University, Suwon, Korea, Republic of

Sriranga Kashyap, Ph.D.  
Krembil Brain Institute, University Health Network
Toronto, Ontario, Canada

Prathamesh Chopade  
Center for Neuroscience Imaging Research, Institute for Basic Science
Suwon, Korea, Republic of

Seong-Gi Kim  
Center for Neuroscience Imaging Research, Institute for Basic Science
Sungkyunkwan University, Suwon, Korea, Republic of

Kamil Uludag  
Krembil Brain Institute, University Health Network|Department of Medical Biophysics, University of Toronto|Center for Neuroscience Imaging Research, Institute for Basic Science & Department of Biomedical Engineering
Toronto, Ontario, Canada|Toronto, Ontario, Canada|Sungkyunkwan University, Suwon, Korea, Republic of

The study aims to, first, investigate the feasibility of sub-millimetre functional magnetic resonance imaging (fMRI) in human subjects using gradient-echo BOLD (GE-BOLD) contrast at standard clinical field strengths of 3T, and, second, to characterise the patterns of activation and temporal signal-to-noise ratio (tSNR) values in the same subjects scanned on both 3 and 7T field strengths. The feasibility of sub-millimetre resolution GE-BOLD fMRI enables the investigation of layer-specific differences in wider cohorts than is currently possible at ultra-high field strengths due to the ubiquity of 3T MRI scanners as well as the GE-BOLD fMRI sequence available on all vendor platforms.

12 healthy participants were scanned on a Siemens 3T Prisma and 7T Terra using a 2D-EPI sequence at nominal resolutions of 0.6, 0.7, 0.8 and 0.9 mm isotropic, respectively. Data were acquired oblique coronal covering at least V1 and V2 (60 slices). Other sequence parameters were adapted to enable scanning within SAR limits with the TE/TR being 44/3000-3600 ms and 28-35/2500-3000 ms at 3 and 7T, respectively. A 0.6 mm isotropic MP2RAGE acquired at 7T served as the anatomical reference. All participants were shown eight blocks of 20 (on) - 40 (off) flickering checkerboard per run to elicit visual activation (Figure 1). Data from both scanners were corrected for gradient nonlinearities using the HCP's grad_unwarp tool and denoised using NORDIC [1]. A high-resolution slab-optimised pipeline using ANTs [2] was used for data pre-processing and AFNI [3] for statistical analyses.

Fig1b shows that NORDIC improves the activation maps both at 3 and 7T for all resolutions, but has a more significant effect for 3T data demonstrated by activation volume for 3T 0.9 mm being similar to 7T 0.6 mm. Fig1c illustrates on a single subject that tSNR maps for resolutions 0.6-0.9 mm isotropic are highly comparable between 3 and 7T and is indicative of the robust performance of product 2D-EPI sequence. Fig2a shows that robust activation maps can be obtained in a single run in less than 10 min, even at 0.6 mm at 3T. Consistent with expectations, 7T yields increased contrast-to-noise (CNR) of the BOLD signal (z-statistics) compared to 3T as is evident from the histogram difference plots. Laminar profiles in Fig2b show the characteristic GE-BOLD signal increase to the pial surface for all data and consistent with activation maps in Fig2b, 7T has a higher dynamic range across depths than at 3T.

Numerous fMRI studies have shown that spatial patterns of neuronal activation can be resolved at a mesoscale (< 1 mm) using the GE-BOLD signal. However, mesoscale fMRI at 3T (and 7T) has been fraught with challenges particularly due to low SNR at sub-millimetre resolutions. While there exists previous work on attempting laminar fMRI at 3T, they often require bespoke coils, offline reconstruction, have limited spatial coverage and/or temporal resolutions [4-9]. We demonstrate the feasibility of acquisition of highest spatial resolution fMRI at 3T (0.6 mm isotropic) using the vendor sequence. We show that novel denoising methods such as NORDIC-PCA enables robust high-resolution fMRI at 3T for resolutions 0.6-0.9 mm isotropic with activation maps that are comparable to those acquired at 7T. The present work successfully pushes the boundary of what has been thus far considered infeasible or difficult to achieve at standard clinical field strengths, showcasing a systematic characterisation of sub-millimetre resolution fMRI data. Our work broadens accessibility to high-resolution fMRI at 3T. Thus, encouraging new directions in neuroimaging and enabling a deeper understanding of human brain function in health and disease, impacting not only neuroscientists but also the broader scientific and medical communities.

Activation (eg. BOLD task-fMRI)

Methods Development ¹

BOLD fMRI ²

Acquisition

Cortical Layers

FUNCTIONAL MRI

HIGH FIELD MR

NORMAL HUMAN