Synopsis

Keywords: Pulse Sequence Design, Contrast Mechanisms, Neuro

We demonstrate sub-millimeter generalized Slice Dithered Enhanced Resolution (gSLIDER) for high-resolution spin-echo (SE) fMRI at 3T. Activations were significantly greater than standard SE fMRI, demonstrating the suitability of this method for high-resolution, mesoscale fMRI.

Synopsis

We demonstrate sub-millimeter generalized Slice Dithered Enhanced Resolution (gSLIDER) for high-resolution spin-echo (SE) fMRI at 3T. Activations were significantly greater than standard SE fMRI, demonstrating the suitability of this method for high-resolution, mesoscale fMRI.

Introduction

Submillimeter spin-echo based functional MRI (fMRI) at ultra-high-field can resolve neural architecture at the level of columns and layers^1,2. Such scanners are limited in number so there is strong incentive to develop techniques for sub-millimeter functional imaging at lower field strengths, despite limitations in sensitivity and specificity to microvascular sources of BOLD activations. Slice Dithered Enhanced Resolution (SLIDER) methods can increase SNR efficiency using thicker slices that overlap spatially. Initial approaches used thick slices with a dithered spatial offset to achieve twice the SNR for fMRI^3,4. In contrast, generalized SLIDER (gSLIDER) uses RF encoding instead of spatial offsets for an orthogonal basis set which allows for maintaining higher SNR after deblurring/thin-slice reconstruction^5,6. Although originally designed for sub-millimeter diffusion MRI, gSLIDER’s SE-based contrast should also reduce large vein bias to facilitate mesoscale fMRI. To evaluate the suitability of gSLIDER for fMRI, we compared gSLIDER against standard SE using a visual stimulation paradigm at 1mm and 0.8mm isotropic resolutions.

Methods

Data was acquired on 2 healthy volunteers in a Siemens 3T Prisma using a 64ch head/neck coil. Subjects viewed a visual hemifield localizer stimulus (35s blocks and 9 repetitions per hemifield) presented with PsychoPy v2021. Trigger pulses were added per slice to the pulse sequence to facilitate syncing between sequence and stimulus delivery. Data collection: 1mm iso whole brain gSLIDER/SE scans: In plane FOV = 220x220 mm; PF = 6/8; GRAPPA 3; MB=2, TE = 69 ms; TR = 17.5s (3.5s per dithered volume); Slice Thickness = 1mm (Slab Thickness = 5mm); PE direction = AP. 0.8mm isotropic whole brain gSLIDER/SE scans were also collected: Matrix = 270x270; PF = 6/8; GRAPPA 3; TE = 79 ms; TR = 17.5s (3.5s per dithered volume); Slice Thickness = 0.8mm (Slab Thickness = 4mm); PE direction = AP. gSLIDER reconstruction: 5 thick-slab volumes were acquired, each with slabs 5x the thickness of the final slice resolution and with a different slice phase (Figure 1). Custom MATLAB code incorporating magnitude, phase data and B1 maps processed the reconstruction into time series for analysis. The slab combination used Tikhonov regularization, wherein a larger regularization parameter (λ) results in greater residual blurring but reduced noise⁷ and λ = 0.1 was used in the current study. fMRI analyses: gSLIDER and SE functional analyses used identical GLM-based AFNI pipelines. A T-test was performed for a right versus left hemifield contrast. Results were cluster thresholded at p<0.1, k=30. Extent of activation (voxel counts after thresholding) were calculated within skull-stripped brain masks.

Results

Figure 2 demonstrates in a representative slice ~2x high tSNR and greater activation extent with gSLIDER vs traditional SE whole brain data at 1mm isotropic resolution. This slice demonstrates an almost 300% increase in activation extent from gSLIDER over standard SE. Figure 3 demonstrates the same pattern with 0.8mm isotropic whole brain data. This slice demonstrates an 82% increase in activation extent from gSLIDER over standard SE. Head motion was comparable between subjects.

Conclusion

This work demonstrates that gSLIDER is a strong alternative to traditional SE in ways that make mesoscale functional imaging possible at 3T. With ~2x the tSNR of SE fMRI, gSLIDER at 3T approaches a specificity and sensitivity combination typically seen at higher field strengths like 7T. A limitation of the spin-echo-based gSLIDER sequence is the slow temporal resolution due to the need for T1 relaxation between each RF encoding. Future work will include more tissue-based quantitative comparisons and will additionally be aimed at improving the temporal resolution of gSLIDER with sliding window reconstruction approaches or Stimulus Locked approaches⁸.

Acknowledgements

Grants R01EB028670, UCSF RAP REAC Award