3983

Higher functional activity in the somatomotor network is associated with functional reorganization in the early stage of multiple sclerosis
Ceren Tozlu1, Sophie Card2, Keith Jamison1, Susan Gauthier3, and Amy Kuceyeski1
1Department of Radiology, Weill Cornell Medicine, New York, NY, United States, 2Horace Greeley High School, Chappaqua, NY, United States, 3Department of Neurology, Weill Cornell Medicine, New York, NY, United States

Synopsis

Multiple Sclerosis (MS) is a disease that causes neuroinflammation and neurodegeneration in central nervous system. Neuroimaging techniques may enable us to better understand the neuropathological mechanisms of MS and how the brain may compensate the pathological changes. We identified the recurring dynamic brain states using fMRI data and investigated the energy required for the transition between those states using the network control theory approach. Our main findings are the dynamic states were visual, somatomotor, and frontoparietal states and the transition energy averaged over all transitions was significantly greater in the MS patients with disability compared to those without disability.

Background

In multiple sclerosis (MS), the structural damage due to neural loss, demyelination, and inflammation is not always proportional to disability, which makes prognostication very challenging1. Advanced neuroimaging techniques may enable us to better understand the neuropathological mechanisms of disability in MS and how the brain function may compensate for the structural damage. Brain dynamics can be used to investigate the time-varying alterations in the brain activity. Brain dynamics method was previously applied in MS2,3,4, however the structural basis of the brain activity measured when transitioning between those dynamic states remains unclear in MS.

Aims

First, we aimed to identify the dynamic brain states using the blood-oxygenation level-dependent (BOLD) time series in healthy controls (HC) and in people with MS (pwMS). Second, the minimum energy required for the transition between those states was computed using structural connectivity applied with the network control theory (NCT) approach. Third, temporal dynamics, transition probability, and energy were compared between HC vs pwMS as well as between pwMS who had no disability vs evidence of disability.

Material and Methods

One hundred pwMS and 19 HC were included in our study. Expanded Disability Status Scale (EDSS) was used to assess disability, where 67 pwMS who had EDSS<2 were considered as not having disability. Dynamic brain states were computed using k-means clustering applied on the regional brain activity measured with BOLD signals from fMRI (See Figure 1). The transition probability between states and temporal dynamics such as dwell time, fractional occupancy, and appearance rate were computed for each subject. The NCT approach was used to compute the minimum energy required for the transition between states. Wilcoxon rank sum test was performed to compare temporal dynamics, transition probability and energy between HC and pwMS as well as pwMS who had no disability vs evidence of disability.

Results

The most recurrent dynamic brain states were visual and somatomotor with high and low amplitude of activity as well as frontoparietal with high amplitude of activity (See Figure 2). The pwMS who had no disability had significantly greater dwell time and fractional occupancy in the somatomotor with high amplitude activity compared to HC (p-value<0.05), while there was no significant difference between HC and pwMS who had evidence of disability (See Figure 3). The persistence energy in all dynamic states were significantly greater in pwMS who had evidence of disability compared to those without disability (p-value0.05) (See Figure 4). The transition energy averaged over all transition possible was significantly greater in the pwMS who had evidence of disability compared to those without disability (p-value<0.05) (See Figure 4).

Discussion

PwMS in the early stage of the disease showed functional compensation and resilience in motor-related regions to limit the disability, while the functional activity in those regions decreases in the later stage of the disease. The pwMS with evidence of disability needed greater energy for the transition between dynamic brain states compared to those without disability, which might be related to less structural damage in pwMS without disability.

Acknowledgements

This work was supported by postdoctoral fellowship FG-2008-36976 (C.T.) from the National Multiple Sclerosis Societythe NIH R21 NS104634-01 (A.K.), NIH R01 NS102646-01A1 (A.K.), and grant UL1 TR000456-06 (S.G.) from the Weill Cornell Clinical and Translational Science Center (CTSC).

References

1. F. Barkhof, “The clinico-radiological paradox in multiple sclerosis revisited.,Current opinion in neurology, vol. 15, pp. 239–45, 6 2002.

2. A. Rocca, M. Hidalgo de La Cruz, P. Valsasina, S. Mesaros, V. Martinovic, J. Ivanovic, J. Drulovic, and M. Filippi, “Two-year dynamic functional network connectivity in clinically isolated syndrome.” Multiple Sclerosis Journal, 2019

3. A.d’Ambrosio, P. Valsasina, A. Gallo, N. De Stefano, D. Pareto, F. Barkhof, O. Ciccarelli, C. Enzinger, G. Tedeschi, M. L. Stromillo, M. J. Ar ́evalo, H. E. Hulst, N. Muhlert, M. Koini, M. Filippi, and M. A. Rocca, “Reduced dynamics of functional connectivity and cognitive impairment in multiple sclerosis.” Multiple Sclerosis Journal, p. 135245851983770, 3 2019

4.Q. van Geest, H. E. Hulst, K. A. Meijer, L. Hoyng, J. J. Geurts, and L. Douw, “The importance of hippocampal dynamic connectivity in explaining memory function in multiple sclerosis.” Brain and Behavior, vol. 8, 5 2018

Figures

The workflow of the study. (A) The region-wise BOLD time series were concatenated for all subjects (19 HC and 100 pwMS). k-means was applied to find the dynamic brain states. This approach allows us to identify which state was assigned to each TR. The assignment information was used then to compute the temporal dynamic metrics such as dwell time, fractional occupancy, appearance rate, and transition probability. (B) The NCT approach was applied with the structural connectivity networks and brain states to compute the minimum energy required for the transition between states.

Recurrent dynamic brain states for k=5. Visualization of dynamic brain states using (A) glass brain and (B) radial plots. The radial plots summarize the coactivation by representing the mean of the positive and negative differences in the same functional networks used for the glass brain plot. The absolute value of the mean negative connections were presented in the radial plots. DAN= Dorsal Attention, VAN= Ventral Attention, LIM= Limbic,FP= Fronto-Parietal, DMN= Default-Mode Network, SUB= Subcortex, CER= Cerebellum, VIS=Visual, and SOM= Somatomotor.

Dynamic state metrics. Dwell time, fractional occupancy, and appearance rate were presented for each dynamic state obtained for each group (HC, pwMS, pwMS who had no disability and those who had evidence of disability).

Comparison of (A) transition probability, (B) transition energy required for each transition or persistence energy, and (C) minimum control energy overall transitions or persistence energy between HC vs pwMS and between two MS subgroups. The group differences are visualized via -log(p)*sign(difference in median), where the differences were computed as pwMS - HC and pwMS with evidence of disability - no disability. *: p-value<0.05 before multiple comparison correction. **: p-value<0.05 after multiple comparison correction.

Proc. Intl. Soc. Mag. Reson. Med. 30 (2022)
3983
DOI: https://doi.org/10.58530/2022/3983