Interactive Data Visualization for Functional Brain Connectivity PDF Free Download
/1
100%
Interactive Data Visualization for Functional Brain Connectivity
Nathalie Henry Riche
Microsoft Research, Redmond, WA
Benjamin Bach
Microsoft Research-Inria Joint Research
Center, Paris/Seattle/Redmond
Roland Fernandez
Microsoft Research, Redmond, WA
Bongshin Lee
Microsoft Research, Redmond, WA
Tara Madhyastha
Radiology, University of Washington, WA
Thomas Grabowski
Radiology and Neurology,
University of Washington, WA
Network science has been instrumental in help-
ing to analyze brain connectivity [6]. More recently,
there has been a growing understanding that net-
work structure changes dynamically throughout the course of an fMRI scan [2,3,4].
However, visualization is not typically used to generate or investigate hypotheses
about brain networks, even though advanced visual interfaces may yield insights into
the underlying data. Research in Information Visualization has yielded such tools for
the interactive exploration of (network) data and to look at the data early in the analysis
phase and to help generate hypotheses. Such visualizations can then be combined
with analytical modeling and statistical analysis to accept or reject hypotheses. Many
current tools represent the network within the space of a 3D brain (e.g., [7]), but many
connectivity patterns, especially changes over time, are better recognized using ab-
stract and purposeful visual representations [1].
Motivation
ConnectoScope is an exploratory vis-
ualization that allows for the interactive
exploration of brain connectivity over time.
Users upload their data in 4D NIFTI le format. Data is uploaded to our
server and stored securily. You can access your session with a pass-
word. ConnectoScope extracts brain connectivity graphs from the 4D
NIFTI images in MNI space. You can provide locations of your spheres in
MNI coordinates.
Dynamic connectivity graphs (correlation between ROIs) are calculated
on sliding windows. Window size is also dened by the user.
Connectoscope consists of 3 visualizations, each of which shows a dif-
ferent aspect of the data: 3D positions of regions (Glass Brain), topolo-
gy of correlation network (Correlation Matrix), and changes in correla-
tion strength over time (LinkWave).
Highlight and color similar elements across visualizations and explore
your data with each visualization individually.
Visualization
ConnectoScope
This research was supported by grants from the National
Institutes of Health
1RC4NS073008-01 and P50NS062684.
www.visualizingbrainconnectivity.org/connectoscope
Examples
• ConnectoScope is an extensible visualization platform for visualiz-
ing dynamic functional brain connectity.
• We currently provide 4 visualizations as well as a client server infra-
structure to load NIFTI les
• Using the LinkWave visualization we observed that periodically,
across multiple networks and individuals, connectivity decreases
synchronously.
• We found that the range of the mean signal in the ROIs in the win-
dow on which correlations are calculated predicts a signicant pro-
portion of the variance in the sum of the correlations.
Conclusions
www.visualizingbrainconnectivity.org/connectoscope
Try Online
[1] Bach, B. (2014), ‘Visualizing Dynamic Networks with Matrix Cubes’, Proceedings of the 32nd
annual ACM conference on Human factors in computing systems, pp. 877 886
[2] Calhoun, V.D. (2014), ‘The chronnectome: time varying connectivity networks as the next
frontier in fMRI data discovery’, Neuron, vol. 84, no. 2, pp. 262 274
[3] Hutchison, R. M. (2013), ‘Dynamic functional connectivity: promise, issues, and interpreta-
tions’, Neuroimage, vol. 80, pp. 360 378 [4] Jones, D. T. (2012), ‘Non stationarity in the “resting
brain’s” modular architecture’, PLoS One, vol. 7, no. 6, e39731
[5] Madhyastha, T. (2014), ‘Dynamic Connectivity at Rest Predicts Attention Task Performance’,
Brain Connectivity.
[6] Sporns, O. (2011), ‘Networks of the Brain’, MIT Press
[7] Xia, M. (2013), ‘BrainNet Viewer: a Network Visualization Tool for Human Brain Connectom-
ics’, PlosOne, vol. 8, no. 7, e68910
References
1
2
3
4
Glass Brain LinkWave Each row (“timeline”) in this visualization shows the correlation of activity be-
tween one pair of ROIs. On each timeline, thickness encodes the corre-
lation strength at a given time point. Solid color indicates positive cor-
relation while outlines indicates negative correlation. Timelines are aligned temporally and sortable by a viewer,
making periods of synchronized activity immediately visible. Interactions such as selecting an element in one view
or changing its color, are reected in other views, showing the data from dierent perspectives.
A glass brain shows ROIs at their
actual positions in the brain. Corre-
lations above the user set thresh-
old are shown as links. Regions
and links can be selected to be
highlighted in the other views.
ROIs
The legend shows the
loaded regions, orgnized
hierarchical. Colores are
applied to regions and
help visually linking them
across all views.
Correlation Matrix The corre-
lation matrix
shows the
correlation for the currently selected time period. Dark cells in-
dicate high correlation, bright cells indicate low corre-
lation (correlation values are averaged over the respective time
period). Some cells are colored, indicating correlations within the
same subnetwork (e.g. bule highlights intra-correlation for re-
gions in the default mode network (DMN)).
The examples on the left show correlation matrices, at dierent
points in time, showing how LinkWave and matrices comple-
ment each other in exploring data and to compare non-adjacent
time points.
ROIs
Correlation
Time
ROI Pairs
Time
Correlation strenght
In the detail on the right, all correlations
that involve regions from the FTPC) are
aggregated, dierentiated by alternating
red/dark ows.
Correlations ordered by mean, while colors indicate higher level regions
(subnetworks).
Ordered by mean correlation Users can select a time period on the correlation between two regions
and ask ConnectoScope to nd similar patterns in other correlations.
Rows in LinkWave are ordered according to their smililarity fo the selected
one (for the selected time period)
Pattern Search
High to Low Low to High
Time Slider
38 39 173
These example show some views created interactively as users explore their
data with ConnectoScope.
