Sue-Jin Lin1,2, Aiping Liu3, Alex MacKay4,5, Brenda Kosaka6, Samantha Beveridge7, Irene Vavasour5, Anthony Traboulsee8, and Martin J McKeown1,2,8
1Graduate Program in Neuroscience, University of British Columbia, Vancouver, BC, Canada, 2Pacific Parkinson’s Research Centre, University of British Columbia Hospital, Vancouver, BC, Canada, 3Department of Electrical and Computer Engineering Program, University of British Columbia, Vancouver, BC, Canada, 4Department of Physics and Astronomy, University of British Columbia, Vancouver, BC, Canada, 5Department of Radiology, University of British Columbia Hospital, Vancouver, BC, Canada, 6Department of Psychiatry, University of British Columbia Hospital, Vancouver, BC, Canada, 7Graduate Program in Counselling Psychology, University of British Columbia, Vancouver, BC, Canada, 8Neurology, Faculty of Medicine, University of British Columbia, Vancouver, BC, Canada
Synopsis
Brain connectivity networks are usually estimated with the
assumption that neural networks do not change over time. However, functional
connectivity is inherently non-stationary, changing across time from seconds to
minutes. In healthy subjects, dynamic reconfiguration of functional
connectivity assessed by fMRI has been estimated and has been linked to cognitive tests, indicating that
flexibility of connectivity normally contributes to cognitive performance. In
this study, we applied a novel time-varying analysis to study network dynamics
in healthy controls and subjects with Multiple Sclerosis (MS). Purpose
To investigate whether there are significant differences in network
dynamics between Multiple Sclerosis (MS) subjects and healthy controls, and
whether network dynamics correlate with behavioural data
Introduction
Brain connectivity networks are usually estimated with the
assumption that neural networks do not change over time (i.e. only include
average time courses for connectivity analysis). However, functional
connectivity is inherently non-stationary, changing across time from seconds to
minutes. Lately dynamic reconfiguration of functional connectivity assessed by
fMRI has been estimated
1,2,3 in healthy subjects and has been linked
to cognitive tests such as working memory
4, indicating that
flexibility of connectivity normally contributes to cognitive performance. There
are some studies investigating network dynamics in disease groups
5,
but little is known about the relations between dynamic features and cognitive
performance. In this study, we applied a novel time-varying analysis to study
network dynamics in healthy controls and subjects with Multiple Sclerosis (MS)
as well as the correlations with cognition.
Methods
In this study, resting state fMRI scans of fifteen healthy
subjects (28.93±5.0 y/o) and eighteen relapsing-remitting MS subjects (32.00±4.93
y/o) were acquired in a Philips Achieva 3.0 Tesla MRI scanner with echo-planar
imaging sequence and the following parameters: 3×3×3 mm
3 resolution,
36 slices, 2000 ms TR, 30 ms TE, 90 degree flip angel, and 240 volumes/dynamics.
3D T1 weighted images were acquired with 1×1×1 mm
3 resolution, 60
slices, 28 ms TR, 4 ms TE and 27 degree flip angle. MS subjects underwent a
comprehensive neuropsychological battery which evaluated executive skills
including mental flexibility, concept formation, attentional switching,
information generation, and working memory as well as processing speed
abilities including attentional and visual scanning. Together with clinical
data, there were 31 variables for every MS subject.
Image preprocessing steps were performed in each subject’s
native space with functions from SPM (UCL, London) and FSL (FMRIB, Oxford). Thirty-six
cognition-associated regions-of-interest (ROIs) were extracted using the open
source Freesurfer program (MGH, Boston). The ROIs acted as masks to determine
the appropriate voxels making up the ROI timecourses. A sliding window approach
was used (Matlab) with window length 60 time points. In total there were 181
windows for each subject. Five network features were acquired based on learned
dynamic connectivity: network variation (NV), network power (NP), flexibility
of homologous connections (FOHC), flexibility of cross-hemispheric connections
(non-homologues regions, FOCC), and flexibility of intrahemispheric connections
(within hemisphere, FOWC). Correlation analysis was performed between network
features and behavioural variables in MS subjects only.
Results
MS subjects demonstrated significantly lower network power
(p=0.02) compared to healthy subjects, which is the summed correlation values across windows, and higher
flexibility of homologous connections (p=0.007, which is the difference of homologous
regions between two windows) [figure 1&2]. MS subjects also demonstrated
higher network variation (summed differences between two windows), higher
flexibility of intrahemispheric connectivity (summed differences of left/right
regions between two windows) and cross-hemisphere connections (summed
differences except homologues between two windows). Nevertheless, these results
did not reach statistical significance (p=0.06, 0.064, and 0.068,
respectively).
In addition, correlation analysis revealed that Verbal
Fluency Test performance was negatively correlated with NV (p=0.03), FOWC
(p=0.03), FOCC (p=0.033), and FOHC (p=0.012) [figure 3].
Discussion
Our results imply that MS subjects demonstrate weak dynamic
connectivity globally and potentially larger fluctuations across time, which
may reflect network instability. The increased network flexibility in
interhemispheric connectivity (based on homologous regions) in MS subjects may
be a compensatory mechanism to deal with possible damage in transcallosal
information transfer.
There was a significant negative correlation between dynamic
features and the Verbal Fluency Test which requires executive skills to
spontaneously generate verbal information according to rules. Presumably
generating many words requires not only participation of distributed brain
regions, but also network flexibility.
Conclusion
With
time-varying approaches, we revealed that network dynamics (i.e. temporal
changes of functional connectivity) are altered in MS subjects and several
network features are correlated with aspects of cognitive performance. This
suggests that connectivity in MS subjects becomes “overly flexible” to
compensate for decreased network power. In addition,
connectivity flexibility may play a role in
cognitive flexibility in MS. Taken together, learning network
dynamics may provide another dimension to probe cognitive dysfunction in MS.
Acknowledgements
We acknowledge the MS
Society of Canada for funding this study.References
1, Hutchison RM, Womelsdorf T, Allen EA, et al. Dynamic
functional connectivity: promise, issues, and interpretations. Neuroimage. 2013;
80: 360-78.
2, Jones DT, Vemuri P, Murphy MC, et al. Non-stationarity in
the "resting brain's" modular architecture. PLoS One. 2012; 7(6):e39731.
3, Allen EA, Damaraju E, Plis SM, et al. Tracking
whole-brain connectivity dynamics in the resting state. Cereb Cortex. 2014; 24(3):
663-76.
4, Braun U, Schäfer A, Walter H, et al. Dynamic
reconfiguration of frontal brain networks during executive cognition in humans.
Proc Natl Acad Sci U S A. 2015; 112(37):11678-83.
5, Damaraju E, Allen EA, Belger A, et al. Dynamic functional
connectivity analysis reveals transient states of dysconnectivity in
schizophrenia. Neuroimage Clin. 2014; 24(5):298-308.