Keilholz - Martinos Center for Biomedical Imaging

Electrophysiological correlates of
functional network dynamics
Shella Keilholz
Emory University/Georgia Tech
September 12, 2014
Resting State Functional Connectivity Workshop
Boston, MA
BOLD dynamics
Time-varying connectivity
• Chang and Glover1: awake
humans, DMN
• Handwerker et al.2: awake
humans, DMN
• Hutchison et al.3: anesthetized
macaques, OCM
• Us: anesthetized rats6
Quasi-periodic patterns
• Majeed et al. 4,5 first
observed in anesthetized
rats
• Also can be seen in awake
humans5, during REM9
• Involves DMN and TPN in
humans
• Reproducible and
quasiperiodic
Individual events
 Petridou et al., awake humans, DMN, etc7
 Liu and Duyn, awake humans, DMN8
NeuroImage 2010; 2NeuroImage 2012; 3Hum Brain Mapp 2012; 4JMRI 2009;
5NeuroImage 2011; 6Brain Connect 2013; 7Hum Brain Mapp 2013; 8PNAS 2014; 9PNAS
2013
1
A CENTRAL QUESTION
Steady state connectivity, QPPs, sliding windows,
individual events…
Are these linked to different
neural/metabolic/hemodynamic processes and can we
separate them?
Neural basis of functional connectivity
• Simultaneous EEG/MRI studies in humans:
networks linked to multiple frequencies (Laufs
et al., Mantini et al.)
• LFP/MRI studies in animals: gamma power
fluctuations most like BOLD fluctuations
(Scholvinck et al., Shmuel and Leopold); delta
and theta power correlation linked to BOLD
correlation (Lu et al., Pan et al.)
Optical imaging
Cerebral Blood Volume
MRI: 9.4 T, dexmedetomidine or isoflurane
anesthesia, single slice gradient echo EPI,
TR 500 ms, TE 15 ms, 64 x64, FOV 2.56, 1000
images
Postprocessing: global signal correction,
detrending, FIR band pass filter 0.01-0.3 Hz
Gradient artifacts
Exponential decay
LFP recording
0-100 Hz
Sliding window correlation
Sliding window correlation
• Range of window lengths used to calculate
pairwise correlation, stepped by one image
each time
• Calculated for simultaneously-recorded BOLD
and BLP from ROIs in left and right SI
Thompson et al., NeuroImage 2013
Sliding window time courses
Sliding Window Series (r)
1
0.8
0.6
BOLD
0.4
Delta
0.2
Theta
0
Alpha
Low Beta
-0.2
High Beta
-0.4
Gamma
-0.6
-0.8
-1
0
50
100
150
200
Time (s)
250
300
Thompson et al, NeuroImage 2013
cor(BOLD sliding window series, BLP sliding
window series) (z)
Correlation between BOLD and LFP sliding
windows
Global Regression
14
14
12
10
8
*
*
*
No Global Regression
12
10
*
*
*
8
6
6
4
4
2
2
0
0
-2
-2
-4
-4
actual
mismatch
ed
Thompson et al, NeuroImage 2013
SLIDING WINDOW CORRELATION
… gamma BLP correlation (also theta, beta) is linked to
BOLD correlation, like Shmuel and Leopold, Scholvinck
et al. for single sites. Previous work by Lu et al., Pan et
al. found delta, theta best predicted BOLD correlation.
Evidence of sensitivity to coordination across areas
rather than large scale changes?
Quasiperiodic patterns
Quasi-periodic patterns (QPPs)
• “Template” generated from rsMRI data
• Has spatial and temporal components
a
b
c
d
Majeed et al, NeuroImage 2011
Linked to infraslow LFPs
• Comparable in frequency to spontaneous BOLD fluctuations
• Seldom recorded due to concerns about drift
• May involve astrocytic activity and/or subcortical input
• BBB permeability??? See Kiviniemi…
• Recorded in 13 rats under isoflurane or dexmedetomidine
with simultaneous imaging
• Coherence between SI BOLD/LFPs calculated
• Cross-correlation with BOLD calculated for range of time lags
Time-lagged BOLD/DC correlation
Pearson correlation at different time shifts between one LFP
electrode and fMRI data at every location
Spatial correlation
Spatial correlation
Spatial correlation
Spatial correlation
Spatial correlation
15
QPPS
…linked to infraslow, sliding window to higher
frequencies. So, are these independent processes?
Isoflurane
Dexmedetomidine
Dead rat
Bench data
5.5
Not significant
Significant
LFP vs. fMRI
Amplitude Frequency (Hz)
Phase amplitude coupling
t
-1 scores
Thompson et al, Front Integ Neurosci 2014
Correlation vs. partial correlation
Dexmedetomidine
0.010.17Hz
amplitude
s
★ fMRI
0.090.29Hz
amplitude
s
★ fMRI
δ/θ power
(1-9Hz)
★ fMRI
fMRI vs.
fast-band
correlation
*******
β power
(15-25Hz)
★ fMRI
fMRI vs.
Infraslow
correlation
Mean Correlation (z)
Isoflurane
Time shift (s)
*********
Standard
correlation
Significantly different
partial vs. standard
correlation
Partial
correlation
Thompson et al, Front
Integ Neurosci 2014
Other contributions?
Vasomotion and vascular effects
• Pulsation from physiological cycles
• Vasomotion
– 0.1 Hz
– Present in isolated arteries
– Signal from periphery correlates with brain in
fingernail
humans brain
** Mayhew, J.E.W. et al., Neuroimage 1996
Vascular contributions
• Simultaneous optical recording of hemodynamic signals in brain and
periphery (5 rats under isofluane, 25 runs)
• ~0.1 Hz hemodynamic signal correlation within bilateral hind paws as well
as correlation between bilateral cortex (S1).
Correlations (r)
1
0.8
0.6
0.4
0.2
0
-0.2
-0.4
-0.6
-0.8
Lpaw
vs.
Rpaw
Lpaw
vs.
LS1
Lpaw
vs.
RS1
Lpaw
vs.
LV1
Rpaw
vs.
LS1
Rpaw
vs.
RS1
Rpaw
vs.
LV1
LS1
vs.
RS1
LS1
vs.
LV1
RS1
vs.
LV1
Hemodynamic peak in BOLD data?
Magnuson et al, NMR in Biomed 2013
HEMODYNAMIC OSCILLATIONS
…may or may not be independent of neural activity but
probably contribute to functional connectivity
Glover, Bandettini
BOLD/nonBOLD
Coherent neural
activity
Buxton
CMRO2/CBF
Infraslow,
BBB permeability
Vascular
oscillations
Summary
• Quasi-periodic BOLD fluctuations appear linked
to infraslow electrical activity, may be tied to glial
cells and/or BBB permeability
• Transient changes in correlation may reflect
changes in coordination in higher frequency
neural activity
• Different types of functional contrast may have
increased/decreased sensitivity to these
processes
• May be able to separate contributions to increase
sensitivity to disease, cognition, etc
Acknowledgments
44C
48B
Other Contributors:
Dieter Jaeger
Chip Epstein
Eric Schumacher
Alessio Medda
Waqas Majeed
Martha Willis
Lukas Hoffmann
Mac Merritt
Matt Magnuson
Biomedical Imaging Technology Center
Center for Advanced Brain Imaging
38C
3B
24A
Funding: NIH 1 R21NS057718-01, NIH 1 R21NS07281001, NIH 1 R01NS078095-01
Air Force Center of Excellence on Bio-nano-enabled
Inorganic/Organic Nanostructures and Improved
Cognition (BIONIC)
NIH Training Grant (Matthew)
DHS Fellowship (Garth)