The human brain networks mediating the vestibular sensation of self-motion

Vestibular Agnosia - where peripheral vestibular activation triggers the usual reflex nystagmus response but with attenuated or no self-motion perception - is found in brain disease with disrupted cortical network functioning, e.g. traumatic brain injury (TBI) or neurodegeneration (Parkinson’s Disease). Patients with acute focal hemispheric lesions (e.g. stroke) do not manifest vestibular agnosia. Thus brain network mapping techniques, e.g. resting state functional MRI (rsfMRI), are needed to interrogate functional brain networks mediating vestibular agnosia. Whole-brain rsfMRI was acquired from 39 prospectively recruited acute TBI patients with preserved peripheral vestibular function, along with self-motion perceptual thresholds during passive yaw rotations in the dark. Following quality-control checks, 25 patient scans were analyzed. TBI patients were classified as having vestibular agnosia (n = 11) or not (n = 14) via laboratory testing of self-motion perception. Using independent component analysis, we found altered functional connectivity in the right superior longitudinal fasciculus and left rostral prefrontal cortex in vestibular agnosia. Moreover, regions of interest analyses showed both inter-hemispheric and intra-hemispheric network disruption in vestibular agnosia. In conclusion, our results show that vestibular agnosia is mediated by bilateral anterior and posterior network dysfunction and reveal the distributed brain mechanisms mediating vestibular self-motion perception.

Sciences, Imperial College London 7 8   9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26 *Correspondence: 27 Barry M Seemungal 28 b.seemungal@imperial.ac.uk 29 Brain and Vestibular Group 30 Neuro-otology Unit 31 Department of Brain Sciences, 32 Imperial College London. There is evidence to support the notion that self-motion perception is not localizable. We 1 previously showed that the duration of motion-perception sensation in the dark (elicited by a 2 rapid stop from constant angular rotation) in healthy individuals correlated with a widespread, 3 bilateral white matter network (Nigmatullina et al., 2015). Subsequent tractography studies also 4 showed that vestibular networks are bihemispheric linked via corpus callosum (Kirsch et al., Currently no study has shown the brain regions specifically associated with impaired self-24 motion perception with intact peripheral functioning (i.e. VA). Our general objective was to 25 identify brain regions associated with VA. Moreover, to test our postulate of an anterior-26 posterior cortical network mediating VA, we hypothesize that VA in acute TBI is mediated by 27 multi-network dysfunction, i.e. (i) by impaired functional connectivity within anterior 28 networks, (ii) impaired functional connectivity within posterior networks, and (iii) inter-29 network impaired functional connectivity. We further hypothesize that posterior networks 30 containing putative vestibular regions (parieto-insular vestibular cortex, mid temporal regions) 31 will have disrupted functional projections. Finally, since we previously showed that VA was 32 linked to imbalance via damage to the right inferior longitudinal fasciculus (ILF) (Calzolari et 33 individual participants were intersected with functional masks to remove regions with missing 23 functional data. These individual masks were binarized (GM: p > 0.2; WM: p > 0.8) and then 24 averaged across subjects. Voxels identified in >70% subjects were used for GM specific mask 25 creation whereas voxels in >90% were selected for WM specific mask creation. (8) Lesion 26 masks of all subjects were then combined and subtracted from the two group analysis space 27 masks. (9) Smoothing (6 mm FWHM) was performed within the grey-and white-matter using 28 GM and WM masks separately. (10) Subsequently, denoising was performed using aCompCor 29 (Behzadi et al., 2007) approach as implemented in CONN, in which six motion regressors (3 30 translational and 3 angular motion), their temporal derivatives, cerebrospinal fluid (10 principal 31 components), and outlier scans identified by ART toolbox were regressed out. One subject with 32 >50% (199 of 300) outlier scans was removed from the analysis. (11) Data was then band-pass 33 filtered with frequency range of 0.008-0.1 Hz. White matter activity and the global signal were 1 not regressed out since white-matter activity was the signal of interest whereas the global signal 2 regression is known to introduce negative correlations (Murphy et al., 2009). 3 4 Group level analysis was performed to determine the differences between VA+ (n = 11) TBI 5 patients and VA-(n = 14) TBI patients. RMS sway and lesion volume were added as covariates 6 in group-analysis to remove the effects explained by balance and extent of injury. In addition, 7 volume regressors were also added to control for tissue atrophy as a result of injury. Grey 8 matter volume was added as a covariate in GM specific whereas white matter volume was 9 added in WM specific analysis. 10

Independent Component Analysis 11
Group ICA was performed to assess the intra-network resting state differences in VA+ and 12 VA-groups using Fast ICA algorithm in CONN toolbox (Whitfield-Gabrieli & Nieto-13 Castanon, 2012). The optimal number of the independent components (ICs) were estimated in 14 the GIFT toolbox (https://trendscenter.org/software/) using a modified minimum description 15 length algorithm (MDL). The optimal number of ICs for GM specific analysis were found to 16 be 38 and 10 for WM specific analysis. 17 In GM specific ICA, independent components containing putative vestibular regions (i.e. 18 parietal operculum (OP2), temporo-parietal, mid temporal, thalamus, and insular regions) were 19 considered components of interest and were evaluated for group comparisons (VA+ vs VA-). 20 All 10 independent components were evaluated for group comparisons in WM specific ICA. 21

Region of Interest Analysis 22
ROI Analysis was used to assess the inter-network resting state differences between VA+ and 23 VA-groups. To select the network ROIs, we used the dice coefficient, which is a measure of 24 spatial overlap between two images. The dice coefficient was measured for overlap between 25 standard atlases and the independent components from our GM-and WM-specific ICA. with the dice-coefficient ≥ 0.01 (29 ROIs) were included in WM specific ROI analysis. As only 3 10 independent components were estimated in WM specific ICA, the components were not 4 sufficient to resolve all white-matter regions available in the ICBM-DTI-81 atlas with a 5 considerable value of the dice-coefficient. Thus, we used a lenient cut-off for the dice-6 coefficient in WM specific ROI analysis to include more white-matter regions in the analysis. 7

Seed-Based Analysis 8
For testing the hypothesis that posterior networks containing putative vestibular regions will 9 have disrupted functional projections, we selected 3 resting-state networks determined via GM 10 specific ICA as seeds (supplementary Figure S1). All three networks contained regions from 11 PIVC (parietal operculum), mid temporal, and insular regions. Furthermore, we probed the 12 brain regions mapped by ILF using bilateral lingual gyri as seeds using default atlas of CONN, 13 which is a combination of the Harvard-Oxford atlas and the AAL atlas. 14 4 Statistical Analysis 15 The assumptions of gaussian random field theory (RFT) are required for whole brain analysis 16  brain differences in a group of 7 vs 7 TBI patients with impaired or normal balance, which is 13 also mediated by vestibular networks, have also recently been identified (Woytowicz et al., 14 2018). We thus think that our analysis is adequately powered and will allow future sample size 15 calculations for studies evaluating vestibular agnosia. 16   regions, were selected as seeds for seed-based analysis (supplementary Figure S1). We found 9 that left thalamus (MNI: -12, -27, 12) had increased functional connectivity with the seed 10 networks (F(3,18) = 24.72, pFDR < 0.05) in VA+ group compared to VA-group ( Figure 8A). 11

Seed-Based Analysis
The region was localised to be a part of antero-dorsal thalamus (medial and lateral) using 12 Melbourne Subcortex Atlas (Tian et al., 2020). 13

Disconnection Between Temporal and Higher Order Visual Cortices 14
Using bilateral lingual gyrus (LG) as seed regions, we found a cluster composed of superior 15 and mid temporal regions with increased functional connectivity in the VA+ compared to the 16 VA-group (F(2,19) = 23.47, pFDR < 0.05). MNI coordinates of cluster centre were (60, -27, 1 00) and the cluster was composed of superior and mid temporal regions ( Figure 8B). Using DTI imaging, we previously identified that VA was linked to imbalance via damage to 9 right ILF (Calzolari et al., 2021), however, we were unable to identify brain regions specifically 10 explaining VA. Given our a priori hypothesis that vestibular-motion perception is mediated by 11 multiple brain networks, we used resting-state fMRI to identify the functional brain networks 12 linked to vestibular agnosia. We found that vestibular agnosia is linked to: (i) increased 13 functional connectivity in a bilateral white-matter network; (ii) increased functional 14 connectivity in SLF; (iii) a disrupted functional link between regions mapped by ILF (Lingual 15 gyrus (V3v/V4) and mid/superior temporal regions); (iv) a disrupted functional link between 16 the left thalamus and resting state networks containing putative vestibular regions (PIVC, 1 insular, and superior/mid temporal regions). 2  3 Using ICA we found that a decreased activity, of the right Intra Calcarine Cortex (ICC) in a 4 bilateral mid-temporal resting-state network, and an overactivation of left frontal pole in rostral 5 prefrontal cortical network, were linked to vestibular agnosia. While looking at the inter-network differences using ROI approach, we found bilateral 20 functional connections between the superior temporal gyri and cerebellar resting-state 21 networks. There was a relative increase in activation of left hemispheric connections and a 22 decrease in right hemisphere connections, for VA+ patients when compared to VA-patients. 23

Vestibular Agnosia and Grey-Matter fMRI
Since resting-state networks do not provide any localization of regions per se, we localized 24 cerebellar regions using a seed-based analysis. There were no statistically significant 25 differences, however, uncorrected findings (for multiple comparisons and hence require a 26 conservative interpretation) showed bilateral functional connections from temporal gyri The lingual gyrus appears to be an important vestibular processing hub that is also involved in to impact elderly people and neurodegeneration patients, however an advantage of our patient 13 group was their relative young age and good premorbid health excluding incipient 14 neurodegenerative disease (given our stringent exclusion criteria), hence the findings reported 15 herein were overwhelmingly related to the acute TBI and not to other chronic underlying 16 disease. 17

18
Our data provide the first evidence linking resting-state functional networks to vestibular 19 agnosiaimpaired self-motion perception. More specifically, we show that self-motion 20 perception is mediated via bihemispheric brain networks, composed of posterior vestibular 21 regions involved in sensory processing and anterior regions involved in sensory integration and 22 perceptual ignition. 23

Declaration of Competing Interests 24
The authors declare that they have no competing interests. 25

Acknowledgements 26
We would like to thank our patient and healthy volunteers for their participation and for helping 27 us improve the traumatic brain injury patients' care. We are also grateful to the major trauma 28 ward teams at St Mary's Hospital and King's College Hospital London for their help with 29 recruitment and assessment. We are also very grateful to The Imperial Health Charity who 30 provided important kickstarter funding that enabled us to obtain research council funding for 1 this project. Raw data that support the findings of this study are available from the corresponding author, 9 upon reasonable request. The request would require a formal data sharing agreement, approval 10 from the requesting researcher's local ethics committee, a formal project outline, and 11 discussion about authorship on any research output from the shared data.