Introduction

The human cerebellum forms an important part of the clinically relevant sensory and motor networks^1,2. Cerebellar damage across multiple neurological diseases such as multiple sclerosis (MS) often leads to motor impairment³. Specifically, the cerebellum is a prevalent lesion site MS patients³. Research suggests that cerebellar pathology can be an early biomarker for the severity of MS⁴. Therefore, understanding the underlying mechanisms through studying cerebellar function can be of high clinical interest. While functional MRI (fMRI) can noninvasively measure brain function, fMRI in the cerebellum remains underused in patients as damage is difficult to quantify due to the thin and highly-foliated cerebellar cortex. In the neocortex, fMRI methods have revealed altered network connectivity in MS patients both in cognitive and motor domains^{5, 6, 7}. Task-based fMRI studies show higher activity in MS patients compared to healthy controls (HCs)^8,9. It has been hypothesised that increased functional activity compensates for structural damage when performing a task^12,13. However, the direction of change depends on the stage of the disease as a decrease in activation is reported in patients with severe disability^8,10. Studies specifically looking at cerebellar connectivity are rare, with preliminary studies showing that compared to HCs, MS patients show reduced connectivity between the sensory motor cortex and the contralateral cerebellar hemisphere^7,11. Recent 7T-fMRI results show reduced activation in the cerebellum during a lower limb task¹⁴. While HCs show somatomotor network hubs in the anterior and posterior lobe of the cerebellum^2,15, it remains unclear whether this activation pattern is altered in MS^16,17. This study compares functional changes within the cerebellum including motor task responses and network connectivity in MS patients compared to HCs using 7T fMRI and submillimetre resolution anatomical images.

Methods

6 MS patients (2F) and 3 HCs (1F), ages (M=55,SD=9.29), were scanned using a 7T-Philips MRI-scanner (8Tx/32Rx whole-head coil). Selected patients showed signs of cerebellar neurodegeneration. The following was acquired for all participants (Fig-1A): A 3D-EPI slab covering the cerebellum (1mm-isotropic,TR/TE=3288ms/21ms, SENSE=2.6/3.27-AP/RL FOV=192x60x192mm³, α=20°) while participants performed a 5min motor task (right hand flex 10s-ON 10s-OFF) and a 7min resting state task (fixation on cross) (Fig-1B). A 0.4mm isotropic MP2RAGE covering the cerebellum (; 5.65/1.88; TI1/TI2, 1000ms/2900ms ; T_I1/T_I2, α =7°/5, FOV=210×120×60mm³ , sensitivity encoding y/z, 1.5/1) with prospective motion correction (realigning reconstructed fat navigators in real time to update the FOV, 2mm isotropic fat navigator (3D EPI with fat- selective binomial excitation pulse;; 5.65/1.88; T_vol=550ms; α=1°; sensitivity encoding y/z, 4/2; T_acq=0.55s, FOV=240×240×120mm³). A whole-head 1.0 mm-isotropic MP2RAGE-sequence (TR/TE=2.3ms/6.2ms, TI₁/TI₂=800/2700,TR_volume=5500ms, α=7°/5°,FOV=230x230x185)¹⁸. Previously calculated group B1+ phase offsets were used to optimise B1+ over the cerebellum²⁰. FMRI data were motion/distortion-corrected. 0.4mm isotropic MP2RAGE images were denoised using a spatially adapted filter^19,21. A first level GLM (Flex>rest, Z>3.1 P<0.05) was fitted to the motor task using FSL²⁰. A cerebellar motor function mask¹⁷ was projected into the anatomical space of each participant. The mask was manually divided into the left/right and anterior/posterior cerebellar lobes to identify relevant clusters (Fig2-A). A spherical seed mask (radius=4px) was created using the voxel with the maximum Z-score resulting from the motor task (Fig-3-A). Mean time-courses were extracted from this seed and used as input for a first level GLM (Motor-seed>rest, Z>3.1, P<0.05) investigating cerebellar connectivity.

Results

The flexing task resulted in significant (Z>3.1,P<0.05) bilateral activation in the cerebellum for all HCs, contrary to less than 50% of MS patients (Fig-2B/C). Activation levels were lower in MS: maximum Z-scores were higher in HCs (Right Anterior: MED=7.85, SD=0.5, Posterior: MED=5.37, SD=0.62) compared to MS patients (Right Anterior: MED=4.28,SD=2.07, Posterior: MED=4.22, SD=1.08). The resting state seed based GLM analysis revealed all HCs and MS patients had significant (Z>3.1,P<0.05) connectivity within the cerebellar parts of the motor network (Fig3-B/C). Mean connectivity Z-scores were higher for HCs in all four cerebellar motor ROIs compared to MS patients (Fig-3B).

Discussion and Conclusion

The motor task consistently engaged cerebellar motor areas in HCs whereas MS patients showed lower and inconsistent activation in both the anterior and posterior lobes. Conversely, some comparisons in the cerebral cortex show higher fMRI activation for MS patients during flexing tasks^12,14. The resting state analysis indicates decreased connectivity in MS patients but a larger cohort and a more in depth analysis are necessary to identify specific differences in motor network connectivity.
In conclusion, we successfully compared fMRI activation during a motor task in MS patients and HCs in the human cerebellum using 7T fMRI. Our high-resolution anatomical data enabled us to employ subject specific ROIs for each participant in our resting state connectivity analysis.

Acknowledgements

This study was supported by an NWO TTW VIDI grant (VI.Vidi.198.016).