10866 The Journal of Neuroscience, July 29, 2015 35(30):10866 10877

Systems/Circuits

Loss of Consciousness Is Associated with Stabilization of

Cortical Activity
Guillermo Solovey,1,3,4* Leandro M. Alonso,1* Toru Yanagawa,5 Naotaka Fujii,5 Marcelo O. Magnasco,1
Guillermo A. Cecchi,6 and X Alex Proekt2,7
Laboratories of 1Mathematical Physics and 2Neurobiology and Behavior, Rockefeller University, New York, New York 10065, 3Integrative Neuroscience Laboratory,
Buenos Aires, Argentina, 4Instituto de Calculo, Faculty of Natural Sciences, University of Buenos Aires (C1428EGA), Buenos Aires, Argentina, 5Laboratory for
Adaptive Intelligence, Brain Science Institute, RIKEN, Saitama, 351-0198 Japan, 6Computational Biology Center, T. J. Watson IBM Research Laboratory, Yorktown
Heights, New York 10598, and 7Department of Anesthesiology, Weill Medical Center, New York, New York 10065

What aspects of neuronal activity distinguish the conscious from the unconscious brain? This has been a subject of intense interest and
debate since the early days of neurophysiology. However, as any practicing anesthesiologist can attest, it is currently not possible to
reliably distinguish a conscious state from an unconscious one on the basis of brain activity. Here we approach this problem from the
perspective of dynamical systems theory. We argue that the brain, as a dynamical system, is self-regulated at the boundary between stable
and unstable regimes, allowing it in particular to maintain high susceptibility to stimuli. To test this hypothesis, we performed stability
analysis of high-density electrocorticography recordings covering an entire cerebral hemisphere in monkeys during reversible loss of
consciousness. We show that, during loss of consciousness, the number of eigenmodes at the edge of instability decreases smoothly,
independently of the type of anesthetic and specific features of brain activity. The eigenmodes drift back toward the unstable line during
recovery of consciousness. Furthermore, we show that stability is an emergent phenomenon dependent on the correlations among
activity in different cortical regions rather than signals taken in isolation. These findings support the conclusion that dynamics at the edge
of instability are essential for maintaining consciousness and provide a novel and principled measure that distinguishes between the
conscious and the unconscious brain.
Key words: anesthesia; consciousness; dynamical criticality; dynamical systems; ECoG; stability analysis

Significance Statement
What distinguishes brain activity during consciousness from that observed during unconsciousness? Answering this question has
proven difficult because neither consciousness nor lack thereof have universal signatures in terms of most specific features of
brain activity. For instance, different anesthetics induce different patterns of brain activity. We demonstrate that loss of con-
sciousness is universally and reliably associated with stabilization of cortical dynamics regardless of the specific activity charac-
teristics. To give an analogy, our analysis suggests that loss of consciousness is akin to depressing the damper pedal on the piano,
which makes the sounds dissipate quicker regardless of the specific melody being played. This approach may prove useful in
detecting consciousness on the basis of brain activity under anesthesia and other settings.

Introduction phenomenon that arises out of the interactions between compo-

Consciousness is the most prominent feature of the mind, yet its nents of the nervous system but is not reducible to them
neurophysiological underpinnings remain clouded by a web of (OConnor and Wong, 2012). What, then, are the emergent fea-
disconnected, circumstantial observations. One source of con- tures of brain activity that are associated with consciousness?
ceptual obfuscation is that, according to a growing consensus We begin with a simple intuitionthe sine qua non of con-
(Thompson and Varela, 2001), consciousness is an emergent sciousness is responsiveness. Responsiveness to sensory stimuli is the
cornerstone of assessment of consciousness in the settings of brain
Received Nov. 28, 2014; revised June 4, 2015; accepted June 11, 2015.
Author contributions: T.Y., N.F., M.O.M., G.A.C., and A.P. designed research; T.Y., N.F., G.A.C., and A.P. performed Correspondence should be addressed to Dr. Alex Proekt at his present address: Department of Anesthesiology and
research; G.S., L.M.A., M.O.M., G.A.C., and A.P. analyzed data; G.S., M.O.M., G.A.C., and A.P. wrote the paper. Critical Care, Perelman School of Medicine at the University of Pennsylvania, Philadelphia, PA 19104. E-mail:
This work was supported by National Science Foundation Grant EF-0928723 (M.O.M.) and National Institute of proekt@gmail.com.
General Medical Sciences Grant 1K08GM106144-01 (A.P.). DOI:10.1523/JNEUROSCI.4895-14.2015
*G.S. and L.M.A. contributed equally to this work. Copyright 2015 the authors 0270-6474/15/3510866-12$15.00/0
Solovey, Alonso et al. Stabilization of Cortical Dynamics in Unconsciousness J. Neurosci., July 29, 2015 35(30):10866 10877 10867

Materials and Methods

Subjects and data acquisition. Data from four
male monkeys were collected at the Laboratory
for Adaptive Intelligence, Brain Science Insti-
tute, RIKEN. The datasets are shared in the
public server Neurotycho (http://neurotycho.
org/). The ECoG array consists of 128 elec-
trodes covering the visual, auditory, somato-
sensory, and motor areas and the parietal and
frontal association cortices (Fig. 1). The ECoG
Figure 1. Diagram of the array configuration used in each one of the four monkeys. potentials were digitized at 1 kHz. Additional
details of the procedure have been described
injury, neurologic disorders, and anesthesia. However, responsive- previously (Nagasaka et al., 2011; Yanagawa et al., 2013). ECoG record-
ness need not relate specifically to sensory perturbations. It seems ings were obtained during the induction of anesthesia starting from the
likely that, to exhibit consciousness, activity arising in one brain awake state. In this study, we analyze a total of 16 experiments each
network must be able to perturb activity in a different network. This consisting of reversible induction of anesthesia starting from the awake
property of responsiveness is related closely to the concept of stability state. The drug doses used to induce anesthesia are shown in
in dynamical systems theory. Stability governs whether the trajectory Table 1.
of the system in phase space is going to change drastically after a A total of 12 anesthetic inductions were performed using one of the
ketaminemedetomidine doses. Four anesthetic inductions were per-
small perturbation or remain essentially unchanged. If the brain dy-
formed with propofol. Ketaminemedetomidine inductions were per-
namics were too stable, then any perturbation would quickly formed by injecting the drugs intramuscularly, whereas propofol was
dampen and brain activity would remain essentially unchanged. administered intravenously. Each monkey received more than one anes-
Conversely, if the dynamics were unstable, then even infinitesimal thetic induction that was separated by at least 1 d.
amounts of noise would grossly disrupt ongoing activity. Therefore, Data processing. Before the stability analysis (see below), the power
we propose that dynamical criticalitythe behavior exhibited by spectra of the ECoG channels were examined to identify the potential for
systems in the vicinity of a bifurcation between dynamically stable contamination with electrical line noise. One channel in monkey M2 was
and unstable behaviors (Magnasco et al., 2009)is essential for found to have significant noise contamination and was removed from
consciousness. subsequent analysis. All other channels for all other monkeys were in-
General anesthetics present an ideal set of tools to test this cluded in the autoregressive models (see below). All channels were fil-
tered using a noncausal filter to remove harmonics of the 50 Hz electrical
hypothesis. Anesthetics act on a variety of receptors (Franks,
line noise and bandpass filtered between 5 and 500 Hz. Both notch and
2008) distributed among many brain regions (Concas et al., 1990; bandpass filters were implemented using the idealfilter function in
Devor and Zalkind, 2001; Chen et al., 2005; Jia et al., 2005). MATLAB (MathWorks) to avoid phase shifts.
Although some, such as propofol, induce the canonical large- Stability analysis. ECoG is a multivariate time series whose dynamical
amplitude low-frequency oscillations in the EEG, resembling properties can be inferred using autoregressive models fitted indepen-
slow-wave sleep (Brown et al., 2010), others, such as ketamine, dently to short time segments as described previously (Solovey et al.,
produce awake-like EEG characterized by oscillations in the 2012). Note that, for our purposes here, we are not interested in recon-
40 Hz range (Maksimow et al., 2006). In fact, despite routine structing the dynamics per se. Our focus here is linear stability analysis.
use of EEG-based monitors, some patients awaken during sur- This is accomplished by performing a locally linear approximation of the
gery, have postoperative recall (Avidan et al., 2011), and develop system dynamics in short temporal windows using autoregressive mod-
eling. For every non-overlapping 500 ms window centered at time T, a
posttraumatic stress disorder. Thus, it is currently impossible to
first-order autoregressive model,
reliably distinguish conscious from unconscious brain even in the
relatively controlled setting of anesthesia. x t A T x T t , (1)
To determine whether dynamical criticality is associated with
consciousness, we performed stability analysis of high-density was fit independently to the ECoG potentials using a least-squares ap-
electrocorticography (ECoG) in nonhuman primates during re- proach after casting the autoregressive model in the form of an ordinary
versible transitions into unconsciousness induced with different regression (Neumaier and Schneider, 2001). In Equation 1, x (t) is a 128-
anesthetic regimens. An array of ECoG electrodes each recording dimensional (number of electrodes in the ECoG array) vector of random
the field potentials of 10 5 neurons (Miller et al., 2009; Ritaccio variables representing ECoG potentials at time t, A( T) is the 128 128
evolution matrix fit to the 500 ms temporal window centered at time T,
et al., 2010) was constructed to cover an entire cerebral hemi-
1 ms (the sampling period), and (t)is the 128-dimensional error
sphere. This allowed us to address simultaneously the macro- vector at time t (T 250 ms t T 250 ms).
scopic stability of cortical dynamics and the dependence of An assumption implicit in this analysis is that the data are both linear
stability on the interactions among activities in different cortical and stationary over the time window for which the model is estimated.
regions. Autoregressive models have been used previously to fit different neuro-
Thus, our overarching hypothesis that dynamical criticality is nal signals including ECoG (Leuthardt et al., 2004). The usual approach
essential for the responsiveness characteristic of consciousness is to fit a single autoregressive model to the entire time series (usually
suggests more specifically that loss of consciousness would be several seconds or minutes long). In this case, it is common to use an
accompanied by a loss of criticality, an implication first prelimi- order p autoregressive model x i1 p
Ai x t i , in which
narily tested by Alonso et al. (2014). Here we show that dynamical the p is left as a free parameter. An alternative approach is to fit a complex,
biologically inspired nonlinear model to the entire dataset (Boly et al.,
criticality is disrupted universally during loss of consciousness
2012). However, this approach requires many free parameters and mod-
regardless of the anesthetic regimen or specific microscopic fea- els of the mechanisms responsible for the generation of the signal.
tures of neuronal activity. Furthermore, stabilization is an emer- Our approach is purely phenomenological in that it does not make any
gent macroscopic phenomenon dependent on the correlations in assumptions about the underlying neural mechanisms and does not re-
activity among different cortical regions rather than on any spe- quire that the signal remain globally stationary. Rather than fitting a
cific feature of each individual local field potential. single linear model to the entire dataset, we divide the dataset into short
10868 J. Neurosci., July 29, 2015 35(30):10866 10877 Solovey, Alonso et al. Stabilization of Cortical Dynamics in Unconsciousness

Table 1. Detailed information of the experimental protocols: three different doses Table 2. Exploration of the effect of high-pass filtering and windows size used to
of ketaminemedetomidine mixture (only a single dose of ketamine estimate the autoregressive models
medetomidine was used in any given monkey) and a single dose of propofol Pearsons Rank correlation
(propofol was only used in monkeys M1 and M2 (Yanagawa et al., 2013, their Comparison correlation coefficient coefficient
Table S1)
Filtered versus unfiltered (500 ms window) 0.999 0.971
Anesthetic agent (mg/kg)
500 ms window (filtered) versus 350 ms 0.935 0.847
Monkey name Species Electrodes (n) Ketamine Medetomidine Propofol window (unfiltered)
M1 Macaca fuscata 128 4.7 0.019 5.2 500 ms window (filtered) versus 750 ms 0.981 0.874
M2 M. fuscata 128 5.6 0.011 5.5 window (unfiltered)
M3 M. fuscata 128 4.7 0.019 Neither filtering nor changes in the window size between 350 and 750 ms fundamentally affect the conclusions.
M4 M. mulata 128 8.8 0.053

strongly affect the results. This was confirmed with this dataset as well
(data not shown).
non-overlapping time segments and estimate the evolution matrix inde-
Methodology controls: robustness of the conclusions with respect to filter-
pendently for each time segment. This relies on a much weaker assump-
ing and window size used to estimate the autoregressive models. Although
tion that, although the signal is globally nonstationary, it may still, to a
the choice of the window for model fitting (500 ms in this case) is some-
good approximation, be locally stationary over a short time segment.
what arbitrary, there are some constraints on this choice. If the time
One significant advantage of autoregressive models is that the stability of
segment is too long, the system is likely to deviate significantly from the
the system can be understood in terms of the eigenvalues of the evolution
assumed stationary and linear regimes. Conversely, if the window is
matrix A (Neumaier and Schneider, 2001). For a 128-electrode array, using
made too short, then less data are available to fit the model with confi-
an order 1 autoregressive matrix (128 128), we obtain 128 eigenvalues of
dence. This latter constraint depends on the sampling rate of the data
the system in a short temporal window. Complex eigenvalues define a fre-
(1000 Hz in this case). Given a 128-electrode grid, the shortest time
quency of oscillation along the corresponding eigenvector. More impor-
segment that can be used to fit a first-order 128 128 autoregressive
tantly, for our analysis, the stability of the dynamics (i.e., exponential growth
matrix is 128 ms. This corresponds to having a single data point for each
or decay along a corresponding eigenvector) is given by the absolute value of
coefficient. Thus, 500 ms is a reasonable compromise between the two
the eigenvalue . This can be understood if we consider the formal solution
constraints. Note that, in our previous work with human ECoG, the data
to Equation 1, x(n ) Anx(0), where is the time step (1 ms in our case).
were sampled at 10 kHz using a 64-channel ECoG array. This allowed us
Writing the eigenvalue decomposition A UU*, where U is orthonormal,
to use 200 ms windows to estimate the models (Alonso et al., 2014).
is the diagonal eigenvalue matrix, and * represents the transpose conjugate,
The choice of the window size limits the bandwidth of frequencies that
we obtain x(n ) UnU*x(0). The norm of the solution can be ex-
can be fit by the model. Furthermore, filtering is a nontrivial operation
pressed as xn k1D
ckn x0 , where D 128 and
that can, in principle, affect the results. Thus, we varied the size of the
ck k, (2) window (350 750 ms) and studied the effect of the high-pass filter on the
stability analysis using the following procedure.
which we will call the criticality index. If ck 1, the mode is stable: a small For each choice of window size and filter, we estimated autoregressive
perturbation along the eigenvector will decay and return the system to its models and computed the distribution of the criticality indices. To study
original trajectory. Conversely, if ck 1, the mode is unstable: any per- how this distribution evolves in time as the monkey is induced into
turbation along the eigenvector will grow exponentially. ck 1 corre- unconsciousness and subsequently recovers, we computed the z statistic
sponds to a critical value at the transition between the stable and unstable using the nonparametric MannWhitney U test between the distribution
modes. In other words, rather than reconstructing system dynamics, in in the initial window and each subsequent window. Thus, every experi-
this work, autoregressive models are used to perform linear stability ment is represented by a sequence of z statistics. To quantify the degree of
analysis of the ECoG potentials. consistency between different windows and filter conditions, we com-
Methodology controls: order of the autoregressive models. Assuming that puted the correlation between the sequences of z statistic. Table 2 shows
Equation 1 is a good local approximation, the evolution matrix A(T ) is the correlation coefficients averaged across all 16 experiments. Thus,
representative of ECoG dynamics in the corresponding short window although we used high-pass-filtered data and window size of 500 ms in
around T. Thus, complexity emerges from the time evolution of the this work, within limits, the choice of the window length or high-pass
model parameters, which allows us to follow abrupt changes in neuronal filtering has only minimal effect on the stability analysis.
dynamics that accompany induction of anesthesia. Spectral analysis. Power spectra were estimated using Thomsons mul-
Therefore, it is critical that the model actually captures the dynamics of titaper method (Thomson, 1982) implemented in Chronux (http://
the ECoG signals within the short windows. Because autoregressive mod- chronux.org/; Mitra and Bokil, 2008) with time bandwidth product of 5
els are stochastic, the goodness of fit can be estimated by the fraction of (nine tapers). Spectra for each channel were estimated in sliding windows
the total covariance of the dataset that is captured by the model. For 128 (window duration, 3 s; window step, 1 s). The spectrum in each window
channels sampled at 1 kHz, the highest-order model that can be fit to a was normalized by total power such that all power estimates are ex-
window of 500 ms is p 3. We evaluated the goodness of fit for orders 1, pressed as fraction of total power in the corresponding window. Fre-
2, and 3 by computing the ratio C/Cx, where C and Cx are the quency bands were defined as follows: , 0 4 Hz; , 4 7 Hz; , 714 Hz;
covariance matrices for the error (Eq. 1) and the ECoG time series, where , 15- 30 Hz; , 30 100 Hz; high , 100 250 Hz. Fraction of power
. . . stands for the matrix norm defined as the largest singular value of contained in each frequency band was computed by summing the power
the matrix. Using data from monkey M1, we obtained mean ratio values across all frequency elements in each band.
of 0.011, 0.0088, and 0.0011 for orders 1, 2, and 3, respectively. Using data For the purposes of the mean spectrum (see Fig. 6 A, B), the spec-
from monkey M2, the mean over the rest and anesthesia conditions were trum for each channel was estimated using 10 s windows using a window
0.0082, 0.0044, and 0.00066; however, during the recovery condition, the step of 1 s. Then spectra from different channels within each experiment
model performed considerably worse, with means of 0.069, 0.063, and were averaged. Spectra were then expressed as deviation from the mean
0.004. As expected, although order 3 yields a better fit (at the expense of awake spectrum computed by averaging spectral windows from 4 min
three times the number of regressors), order 1 models capture 99% of culminating in drug injection.
the covariance of the ECoG time series and are therefore a good approx- Connectivity. We did not require the autoregressive matrices to be
imation of the ECoG dynamics. Thus, we use order 1 models for the rest sparse, and, thus, each electrode is connected to all others. For the pur-
of the analysis of both real and surrogate datasets (see below). Alonso et pose of simplifying the content of the matrices, we binarized each one
al. (2014) has shown that the choice of the order of the model does not according to the following procedure: we estimated the mean and SD of
Solovey, Alonso et al. Stabilization of Cortical Dynamics in Unconsciousness J. Neurosci., July 29, 2015 35(30):10866 10877 10869

Figure 2. Stabilization of cortical activity after injection of an anesthetic drug. AC, Ketaminemedetomidine; DF, propofol. Evolution of the distribution of the criticality indices as a function
of time since anesthetic injection for monkey M2 under ketaminemedetomidine (A, top) and under propofol (D, top). For each time window, the probability distribution of the criticality indices (see
Materials and Methods) is shown by color (red shows low probability, yellow shows high probability). Note that, in the awake state, criticality indices crowd the critical line (1) between stable
(1) and unstable (1) regions. With drug injection, the fraction of criticality indices above 0.98 drops abruptly (A, D, bottom). Results obtained for monkey M2 are representative of recordings
in all monkeys (see Figs. 3, 4). C and F show the average drop in the number of the most critical modes 4 min after ketaminemedetomidine and propofol, respectively. The differences in the
distribution of criticality indices after drug injection are statistically significant in all monkeys (B, E): p value of a KolmogorovSmirnov test comparing criticality indices at t 0 and at subsequent
times. **p 0.01.

the distribution of values, A Aij, A Aij2 A 2 (where A is the vicinity of the critical value (criticality index 1). During the
autoregressive matrix in Eq. 1), respectively, so that only elements 2 SDs induction of the anesthetic state, the number of critical modes
above the mean were considered. Formally, the summary matrix B is decreases: the number of modes with an associated criticality
defined as: index larger than 0.98 decreased from 17.5 to 13.4% and

B ij 1 if Aij A 2a
0 otherwise
from 15 to 4.5% within 4 min after the injection of ketamine
medetomidine and propofol, respectively (Fig. 2C,F ). Although
0.98 is a somewhat arbitrary threshold, as can be seen in Figure 2,
For each experiment, we took 200 s of data and binarized each autore- A and B (as well as in Figs. 3, 4) induction of the anesthetized state
gressive matrix using the above procedure. is accompanied by the decrease in the density of the criticality
Then we constructed a summary matrix S out of these binarized ma-
indices near 1, and, thus, the observed stabilization does not de-
trices as follows: Sij(i j) 1 if Bij 1 for any of the matrices within the
200 s time window. This was performed for control conditions before pend strongly on the threshold. Thus, consistent with our hy-
induction of anesthesia over a 200 s window. Because the kinetics of pothesis, dynamical criticality normally present in the awake
ketaminemedetomidine and propofol effects were different (see Figs. brain (Solovey et al., 2012) is abolished after the administration
2 4), for the ketaminemedetomidine experiments, we started the 200 s of anesthetics.
anesthesia period at 500 s after drug injection, and for the propofol Although both anesthetic regimens result in statistically sig-
experiment, the anesthesia period started at 150 s. nificant stabilization of cortical activity, the onset of effect of
The diagonal and off-diagonal elements of the matrix capture two propofol is faster than that for the ketaminemedetomidine
different aspects of neuronal dynamics. Diagonal elements reflect the mixture and is less persistent. Note that ketamine and medetom-
dependence of activity in each electrode on its own past history, the idine were administered intramuscularly, while propofol was
off-diagonal elements represent the effect of signal at one electrode on
given intravenously (Yanagawa et al., 2013). Pharmacokinetic
future signal in a different electrode. Off-diagonal elements i,j are shown
as arrows originating at j and pointing toward i. The arrow color changes models (Schnider et al., 1999) suggest that the peak in the effect
from blue to red (origin target). site concentration occurs at 1.7 min after intravenous bolus of
Note that S reflects only the off-diagonal elements. For diagonal ele- propofol. Consistent with this, we observed robust stabilization
ments, we created a summary vector Di t1 T
Bii . This was estimated for of cortical activity at 2 min. The pharmacokinetics of ketamine
the window before, Dib, and after, Dia, anesthesia. We show Di Dib Dia and medetomidine are more complex, but the peak in the plasma
normalized for the purposes of visualization to lie between 0 (red) and 1 ketamine concentration after intramuscular injection in children
(blue). That is, if the diagonal element was higher while awake, it appears occurs at 15 min (Grant et al., 1983). Consistent with this, we
as blue; otherwise, it appears as red. observed sustained decrease in the number of critical modes 15
min after administration of ketamine and medetomidine.
Results
Induction of anesthesia results in stabilization of
cortical dynamics Stability is an emergent phenomenon dependent on
Our main finding is that cortical dynamics become stabilized correlations in cortical activity
during induction of anesthesia, with both ketaminemedetomi- Induction of anesthesia is associated with complex changes in the
dine and propofol (Figs. 2 A, D for representative examples, power spectrum (John et al., 2001), phase relationships among
B, C, E, F for group data; 3, 4). In the awake state before drug different oscillations (Mukamel et al., 2011), and changes in co-
injection (t 0; Figs. 2 and timing of drug administration is herence (Cimenser et al., 2011) and correlations (Ku et al., 2011).
indicated by vertical lines in 3, 4), the criticality indices given by To address specifically the role of correlations among cortical
Equation 2 (see Materials and Methods) crowd in the immediate sites to the global stability, we constructed time-shuffled surro-
10870 J. Neurosci., July 29, 2015 35(30):10866 10877 Solovey, Alonso et al. Stabilization of Cortical Dynamics in Unconsciousness

Figure 3. Stabilization of cortical activity after injection of ketaminemedetomidine for all monkeys in different sessions. Two-dimensional histograms of the estimated criticality indices and the
fraction of modes with stability indices above 0.98 (bottom plots). Drug injection occurred at time indicated by vertical lines. Soon after drug injection, the fraction of the most unstable modes drops
abruptly. Different histograms correspond to different experiments separated by at least 1 d.

Figure 4. Stabilization of cortical activity after injection of propofol for monkeys M1 and M2 in different sessions. Two-dimensional histograms of the estimated criticality indices and the fraction
of modes with stability indices above 0.98 (bottom plots). Drug injection indicated by vertical lines. Soon after drug injection, the fraction of the most unstable modes drops abruptly.
Solovey, Alonso et al. Stabilization of Cortical Dynamics in Unconsciousness J. Neurosci., July 29, 2015 35(30):10866 10877 10871

Figure 5. Distributions of the criticality index in time-shuffled surrogates are internally inconsistent. A, To construct time-shifted surrogates for each channel in the 128-electrode grid, we chose
a time shift from a Gaussian distribution ( 0, 50 ms). A single shift was chosen for each channel and applied to the entire recording spanning the awake state, induction of the anesthetized
state, and recovery. Example of a single channel of data (black) and the time-shifted surrogate (red) are shown. Note that the time shift preserves all features of the signal, including frequency and
phase information. B, Time shifting the signals in the surrogate (red) produces a very mild perturbation of the real data (black). Sixteen of 128 channels are shown, yet in every experiment on every
monkey, we find time windows that have significant differences in the distribution of the criticality indices ( p 0.05, MannWhitney U test). C, Procedure used to compare the internal consistency
of the data and the surrogate during the awake state. First, reference distribution of criticality indices was constructed from the first 5 min of recording (600 500 ms windows). Then p value was
computed using MannWhitney U test between the reference distribution and each of the subsequent windows during the awake state (before drug injection). This procedure was applied to the
real and time-shifted surrogate datasets (each was compared to its own reference distribution). If, while the monkey is awake, the distribution of criticality indices (Figure legend continues.)
10872 J. Neurosci., July 29, 2015 35(30):10866 10877 Solovey, Alonso et al. Stabilization of Cortical Dynamics in Unconsciousness

gates produced by independent random constant shift of the time reference distribution of the criticality indices over the first 5 min
stamps for each electrode, with values drawn from a Gaussian for real and time-shifted surrogate datasets (using 500 ms win-
distribution with 0 and 50 ms (Fig. 5A). Note that this is dows to estimate each autoregressive matrix independently) and
a very subtle perturbation to the signal (Fig. 5B); it selectively compared this distribution with those obtained subsequently
disrupts higher-order correlation but preserves all features of during wakefulness using MannWhitney U test. This procedure
each individual channel taken in isolation, and all of the pairwise is illustrated in Figure 5C. Figure 5D shows cumulative distribu-
correlations between signals are maintained but arbitrarily time tions of p values for real (blue) and time-shifted (red) datasets.
shifted by a small amount. However, in every experiment on The abscissa shows the p value for rejecting the null hypothesis
every monkey, we observed statistically significant differences in that the monkey is awake, and the ordinate shows the probability
the distributions of the eigenvalues obtained in corresponding of rejecting the null hypothesis at a given level of statistical signif-
time windows for the real and surrogate datasets ( p 0.05, icance. The results indicate that the eigenvalues for the real data
MannWhitney U test). appear to be drawn from one single (consistent) distribution
Dynamical criticality is a highly unlikely state of the system (probability of falsely rejecting the null hypothesis is low). In
and is not generically expected. Thus, the presence of dynamical contrast, the eigenvalues for the surrogate data fluctuate in dis-
criticality implies some self-tuning. It has been proposed (for tribution; in other words, the surrogate data appear to have been
review, see Chialvo, 2010) that many aspects of neuronal activity drawn from a distribution that fluctuates haphazardly in time.
are found at a critical point of a second-order phase transition. Thus, given the surrogate dataset, the null hypothesis monkey is
These critical points are characterized by power law distributions awake can be readily falsely rejected with high statistical confi-
of events such as sizes of avalanches in models of sand piles (Bak dence. This result is consistent with the notion that the brain is
et al., 1987) and activity avalanches in cultured neurons (Beggs self-tuned to exhibit dynamical criticality and that this self-
and Plenz, 2003), as well as other biological systems (for review, tuning involves global correlations. That is, dynamical criticality
see Mora and Bialek, 2011). One appealing aspect of this theory is is an emergent phenomenon.
that it has been shown that some systems can spontaneously ex- Note that we chose 50 ms as the width of the Gaussian distri-
hibit such critical behavior without the need for fine tuning of bution from which the time shifts were drawn. This preferentially
parameters (Bak et al., 1987). disrupts eigenmodes with higher frequencies. Our objective here
However, the analysis of statistical properties of neuronal ac- was to produce the most subtle perturbation to the data that can
tivity, such as avalanche size distribution, glosses over the dynam- nonetheless disrupt the results of the stability analysis. Broaden-
ical properties, such as stability. It is possible that the same kind of ing the distribution of time shifts is expected to further alter the
tuning process that produces power law distributions of states data and exacerbate the differences between real and time-shifted
also yields marginally stable dynamics (Magnasco et al., 2009), datasets.
but currently the relationship between dynamical and statistical
aspects of criticality is not well understood (Mora and Bialek, Power spectrum is uninformative of the action of anesthesia
2011). It is well known that induction of anesthesia is associated with
Although statistically significant differences between the sta- changes in the spectral signatures of cortical signals (Brown et al.,
bility of real and time-shuffled surrogates essentially rule out the 2011). Note, however, that in contrast to the uniform increase in
possibility that stability depends on any set of features of each stability observed with both ketaminemedetomidine and with
signal in isolation, this analysis does not directly address self- propofol, changes in spectral content induced by these drugs are
tuning, which we address by comparing the internal consistency distinct (Fig. 6A). Changes observed with propofol are, on aver-
of the real and time-shuffled surrogates. age, much larger than with ketaminemedetomidine. In addition
It is not easily knowable when, exactly, the monkey loses con- to the increase in the slow oscillations 5 Hz, complex cha-
sciousness after the administration of the anesthetic agent, but nges in other frequency ranges variably accompany induction of
the monkey is known to be awake before drug administration. anesthesia.
If the awake brain was self-tuned to exhibit dynamical criticality, Although it is true that, on average, spectral signatures of cor-
the distribution of the criticality indices obtained for different tical signals change with induction of anesthesia, these average
time windows during the awake state ought to be stationary, and quantities conceal both the temporal and spatial heterogeneity of
hence the distribution of eigenvalues for any two distinct win- changes in the power spectrum (Fig. 6CF ). For instance, al-
dows should remain consistent. In other words, the eigenvalues though, on average, we observe the expected increase in the low
obtained in any given epoch while the monkey is awake should be frequency () power with both ketaminemedetomidine and
drawn from a single probability distribution that remains stable propofol (Fig. 6C,D, thick lines), each individual channel can
in time, and therefore we should not be able to reject the null deviate quite significantly from the mean behavior (thin lines).
hypothesis the monkey is awake based on the distribution of Furthermore, the spatial distribution of changes in power is dif-
the criticality indices. To test this prediction, we computed a ferent between the two anesthetic regimens (Fig. 6 D, E). In many
frequency bands, profound fluctuations in the spectral content
4 are observed such that the changes in the power spectrum depend
strongly on time, the location of the electrode, and the anesthetic
(Figure legend continued.) remains unchanged (internally consistent), one expects to have regimen. The complexity of changes in the spectral content ob-
very few small p values. D, Cumulative distribution of p values obtained using the procedure served here is consistent with that observed by Breshears et al.
outlined in C for real (blue) and surrogate (red) datasets. The abscissa of this plot corresponds to
(2010) in human ECoG recordings obtained during induction
the desired statistical significance for rejection of the null hypothesis that the subject is awake.
The ordinate shows the probability of rejecting this null hypothesis at the desired significance and recovery from propofol anesthesia.
threshold. Data from multiple experiments on the same monkey are combined into a single It is not entirely surprising that the distribution of the stability
distribution. Note that the probability of rejecting the null hypothesis is much higher for the parameter behaves differently from the power spectrum. In a
surrogate than for the real data. Thus, disruption of global correlations induced by the time shift steady-state autoregressive process, there is a complex relation-
disrupts the internal consistency of the distribution of criticality indices. ship between the power spectrum of the process (the absolute
Solovey, Alonso et al. Stabilization of Cortical Dynamics in Unconsciousness J. Neurosci., July 29, 2015 35(30):10866 10877 10873

Figure 6. Changes in the power spectrum are not consistent among anesthetic regimens and cortical areas. A and B show time-resolved spectrograms averaged across all monkeys for each of the
anesthetics. First, spectra from all channels were computed for each monkey and averaged across channels. Spectra were then normalized such that power at each frequency is expressed as a fraction
of total power. To emphasize the effect of anesthetics, spectra are plotted as differences from the mean awake spectrum (computed over 4 min culminating in drug injection for each monkey). Note
that the spectral signatures of the two anesthetic regimens are clearly distinct. Furthermore, note that, in addition to the increase in the power of low (5 Hz) frequencies, power in many frequencies
changes in a nontrivial manner. These average quantities gloss over both spatial and temporal variability of the spectral changes that accompany induction of anesthesia (CF). All data in CF come
from the same monkey that was subjected on different occasions to either propofol or ketaminemedetomidine. C and D show changes in power for each of the different frequency bands for
ketaminemedetomidine and propofol, respectively. Thin lines show spectra from each individual channel. Thick lines show average across channels. Pronounced differences from the mean
behavior are observed for many frequency bands. Changes in spectral power are projected onto the electrode grid (C, ketaminemedetomidine; D, propofol). Power in each frequency band was
averaged across the window shown by the rectangle in C and D. All channels were ordered from highest power to lowest power and colored from red to blue. This shows that, although there are some
similarities between the anesthetics in terms of spatial distribution of spectral power, there are also pronounced differences.
10874 J. Neurosci., July 29, 2015 35(30):10866 10877 Solovey, Alonso et al. Stabilization of Cortical Dynamics in Unconsciousness

value squared of its Fourier transform)

and the distribution of the eigenvalues in
the complex plane. Summarily, each
eigenvalue contributes to the power spec-
trum a Lorentzian distribution, whose
center is the frequency given by the imag-
inary part of the eigenvalue and whose
width is given by the (negative) real part of
the eigenvalue. The height is given by the
projection of the noise term onto the cor-
responding eigenvector, meaning that
each channel can have in principle a
sharply different spectrum. Therefore,
very similar power spectra can be achieved
by very different eigenvalue distributions,
and, conversely, the distribution of eigen-
value frequencies by itself does not deter-
mine the spectrum. The objects in which
we are most interested in this study, the Figure 7. Induction of anesthesia is associated with changes in functional connectivity. All data are from monkey M2. Strong,
criticality indices, do not determine the consistent connections were extracted from the autoregressive matrices and plotted for the awake and anesthetized conditions. A,
spectrum at all. For instance, it is easy to Off-diagonal elements; B, diagonal elements. Note that the effects of different anesthetics on the connectivity patterns differ
conceive of a system in which a high- between anesthetics. Although propofol leads to a dramatic decrease in the number of connections, ketaminemedetomidine
frequency mode decays either faster or leads to the reversal of flow of connections in parallel with the weakening of the autoconnections (diagonal elements) in the
slower than a low-frequency mode and primary sensory cortices.
thus can be either more or less stable.
Non-steady-state autoregressive processes, such as in this Frequency dependence of stabilization
work, add additional complexity to the relationship between sta- The focus of our analysis thus far has been the criticality index
bility and spectral characteristics because the dependence on the given by the , but generally, the eigenvalues ( values) of the
initial conditions cannot be brushed aside, as the system never autoregressive matrices are complex: j jei j. Thus, the fre-
j
decays into its asymptotic state before the matrix is changed. The quency of the jth eigenmode is given by fj
potential presence of unstable modes generate growing oscilla- 2 t, and the
tions that can only truly be taken into account by remembering log pj
damping rate is given by j , where t is the sampling
that the window is finite and that, at later times, the process will t
move that eigenvalue onto the stable regime. rate (1 ms). Note that 0 corresponds to critical dynamics,
0 corresponds to damped modes, and 0 corresponds to
unstable modes. Thus, a more complete analysis of changes in
Global reorganization of functional connectivity stability examines the changes in the distribution of the eigenval-
Although the eigenvalues of the autoregressive matrix (see Materials ues induced by anesthetics in the plane spanned by the frequency
and Methods) define the stability of the neuronal dynamics, the and damping rate (Fig. 8). Although propofol anesthesia resulted
autoregressive coefficients forming these matrices define a network in the increased damping of eigenmodes in the broad range of
of interactions between individual cortical areas. Thus, characteriza- frequencies from 5 to 100 Hz, stabilization induced by keta-
tion of the structure of the autoregressive matrices could be used to minemedetomidine was limited to the high-frequency range
understand changes in directed functional connectivity during greater than 50 Hz.
transition from consciousness into the anesthetized state (Fig. 7).
Although the fitting procedure was blind to the anatomy, note Discussion
the consistency in the overall flow of connections in the awa- Our main finding is that reversible loss of consciousness is accom-
ke state from the primary sensory (occipital) cortices toward the panied by reversible stabilization of brain dynamics such that critical
frontal association and inferotemporal cortical areas. Thus, at oscillations in the awake brain become damped during uncon-
least to some degree, the autoregressive models fitted to the ECoG sciousness. The consistency of the stabilization is remarkable given
capture the patterns of flow of information in the primate cortex. the clear inconsistencies in the spectral characteristics (Fig. 6), di-
Loss of consciousness induced with both propofol and the keta- rected functional connectivity (Fig. 7), and frequency dependence of
minemedetomidine mixture results in the disruption of the global damping (Fig. 8) observed with different anesthetics.
connectivity pattern. Propofol anesthesia results primarily in the dis- In dynamical systems theory, stability analysis refers to an
appearance of connections. In contrast, anesthesia induced with ket- evaluation of whether the trajectory of the system would be sim-
aminemedetomidine results in the reversal of direction such that ilar or vastly different after a perturbation. It is performed by
the inferotemporal cortex became the origin of the bulk of the con- evaluating a linear approximation to the full nonlinear dynamics
nections. This is consistent with the decrease in the strength of self- centered at the observed trajectory. This is what our analysis at-
connections observed in the occipital cortex (Fig. 7B). These tempts to do: to capture coarse-grained linear components of the
results were consistent across experiments and monkeys. dynamics valid for a short time window. As the system moves in
Thus, although loss of consciousness results in the disruption its phase space, the linear approximation changes, which is ob-
of functional connectivity patterns, the specific nature of the served as changes to our matrix. An established result of dynam-
disruption depends quite strongly on the identity of the anes- ics systems theory is that the dynamics is strongly dominated by
thetic agent. the linear components, except when the linear components are
Solovey, Alonso et al. Stabilization of Cortical Dynamics in Unconsciousness J. Neurosci., July 29, 2015 35(30):10866 10877 10875

Figure 8. Frequency dependence of stabilization. All data are from monkey M2. Histogram of complex eigenvalues is plotted in the plane spanned by the damping timescale and frequency.
Histograms were normalized such that occupancy spans the range from 0 to 1. The histograms are encoded in the greenred blue color scheme (green for the awake state before the induction of
anesthesia; red for the anesthetized state that ensues after drug administration; blue for the awake state observed after the effect of the anesthetic subsides). The rightmost panels correspond to
the overlay of the three plots. This overlay is constructed such that pure red, green, or blue indicate regions in the plane that are occupied preferentially during each one of the three conditions.
Locations with similar occupancy in the awake and recovery states appear cyan (mixture of green and blue), whereas regions that are exclusive to the anesthetized state appear red. The absence of
pure green or blue in the overlay plot shows that the distribution of eigenmodes is essentially restored after the effect of the anesthetic subsides. The leftward shift from cyan to red in the overlay
panels demonstrates stabilization observed with both anesthetic regimens. Although in the case of propofol stabilization is observed across the entire frequency range, with the ketamine
medetomidine regimen, stabilization is most prominent at higher frequencies.

critical, in which case the higher-order nonlinear terms decide to a stable regime (Yan and Magnasco, 2012). Imas et al. used volatile
the ultimate fate of the system. Thus, within our analysis to find anesthetics and Ferrarelli et al. used midazolam, neither one of which
that a large number of eigenvalues are critical means that the shares significant similarities in terms of molecular mechanisms or
system is not long-term predictable by this very analysis. What changes in the cortical power spectrum with the ketaminemedeto-
our analysis is capable of demonstrating is that the system hovers midine mixture. Thus, loss of dynamical criticality maybe a universal
over a regime with many critical eigenvalues, the central thesis of feature of loss of consciousness induced by anesthetics.
this study. Anesthetics may at least in part impinge on the same neuronal
What are the functional consequences of this? Critical dynam- mechanisms as sleep (Brown et al., 2010). There is indeed evidence
ics render the system exquisitely sensitive to perturbations (e.g., that sleep and anesthesia share essential features in terms of the qual-
sensory stimuli), not in an exponentially unstable butterfly ef- itative dynamics. For example, the effect of TMS is almost identical
fect manner but in a more subtle propagate the perturbation in the naturally sleeping (Massimini et al., 2005) and anesthetized
across the system manner. Consistent with this intuition, our (Ferrarelli et al., 2010) subjects. Furthermore, the effect of TMS-
results suggest that the decrease in the responsiveness of an anes- induced perturbation during disorders of consciousness is similar to
thetized subject is heralded by a shift of the dynamics away from sleep and anesthesia (Casali et al., 2013). This leads us to make a bold
the critical regime. and tantalizing suggestion that dynamical criticality is an essential
Several seemingly disparate results can be explained by the stabi- requirement for wakefulness in the brain.
lization hypothesis and integrated into a deeper conceptual frame-
work. Imas et al. (2005) demonstrate that the long-latency Relationship to functional connectivity
component of an evoked potential is suppressed under anesthesia Many methods have been applied previously to show changes in the
and that the evoked potentials exhibit damping. Perturbation need functional connectivity in anesthesia correlation between fMRI
not be sensory. The effect of transcranial magnetic stimulation voxels (Peltier et al., 2005) and EEG channels (Lee et al., 2011),
(TMS) in the awake subject persists longer and is more complex than band coherence (Cimenser et al., 2011), phase lags between EEG
in the anesthetized subject (Ferrarelli et al., 2010). This is exactly recordings (Ku et al., 2011), nonlinear generative models of EEG
what one expects when the system moves from a dynamically critical (Boly et al., 2012), and frequency-specific modes of large-scale
10876 J. Neurosci., July 29, 2015 35(30):10866 10877 Solovey, Alonso et al. Stabilization of Cortical Dynamics in Unconsciousness

communications (Yanagawa et al., 2013) revealed using spectral For systemic drug administration, our stabilization hypothe-
Granger causalityall show changes with anesthetic-induced sis predicts that the combined system including both hemi-
unconsciousness. spheres should exhibit stabilization during loss of consciousness.
Here, we use autoregressive matrices to define directed func- Furthermore, we would expect that spatial projections of some
tional connectivity. Although our results with propofol are eigenmodes would span both hemispheres. These predictions can
consistent with the notions of disconnected brain, under keta- be tested experimentally in the future. What is less clear is what
minemedetomidine anesthesia, we observe reversal of the direc- should happen with direct administration of the anesthetic
tion of many connections rather than disappearance of agents to a single hemisphere. In light of the observations with the
connections. One manifestation of this reversal is that, in the Wada test and hemispherectomy, we would hypothesize that
brain anesthetized with ketaminemedetomidine, signals from modes localized to the ipsilateral hemisphere and those that span
the visual cortex become less self-driven and more influenced both hemispheres may exhibit stabilization, whereas the modes
by the activity of the inferotemporal cortex. Ketamine is known localized only to the contralateral hemisphere may be spared.
to elicit visual hallucinations and has been used as a pharmaco-
logical model of positive symptoms of schizophrenia (Krystal et Outlook
al., 2003). Visual hallucinations are associated with increased ac- One common interpretation of dynamical criticality in the brain
tivity in the occipital cortex (Ffytche et al., 1998). Our results is that a system can only achieve optimal information processing
suggest that, during visual hallucination, this activity in the oc- in the vicinity of a phase transition (Langton, 1990), yet most
cipital cortex is driven by other brain areas rather than sensory physical systems enjoy physical stability: they preserve their
experience to generate the illusion or hallucination. qualitative behavior under perturbations, i.e., the defining pa-
For propofol, stabilization can be related directly to the de- rameters of most systems are found away from a bifurcation. In
crease in functional connectivity. Elimination of the off-diagonal stark contrast to this, our results and those of others (Solovey et
elements in the autoregressive matrices has the effect of decreas- al., 2012; Alonso et al., 2014) suggest that a significant fraction of
ing the number of feedback loops that can destabilize the system. cortical dynamics is found close to a bifurcation. This conclusion
The precise nature of the relationship between functional con- is not unique to cortical dynamics. The most compelling case for
nectivity and stability is less clear for the brain anesthetized with dynamical criticality has been made for the auditory periphery in
ketaminemedetomidine. which Hopf bifurcation is thought to give rise to some of the
However, regardless of the specifics, our methodology offers fundamental aspects of hearing, such as compressive gain, fre-
significant advantages over other methods. While autoregressive quency tuning, and spontaneous otoacoustic emissions (Hud-
modeling is conceptually closer to the notion of causation pro- speth et al., 2010).
posed by Granger than correlation-based connectivity (Garg et The very fact that a system is found close to a bifurcation
al., 2011), the main advantage in the present context is that it implies self-tuning because bifurcations occupy a small region in
allows for direct interrogation of the global brain dynamics that the parameter space. Although our analysis is phenomenological
arise out of this correlation structure through eigenmode decom- and thus does not allow for the direct interrogation of underlying
position. By fitting autoregressive models to short temporal mechanisms, in theoretical models (Magnasco et al., 2009), sim-
windows, our method allows for globally nonstationary and non- ple and biologically plausible anti-Hebbian synaptic plasticity
linear dynamics and yet takes advantage of the well tractable lin- leads robustly to a dynamically critical regime. It is possible that
ear stability analysis. the slow increase in stability seen with induction of anesthesia is a
manifestation of the changes in the synaptic plasticity.
Neurophysiological measures of the depth of anesthesia
In clinical and most research settings, depth of anesthesia (DOA) Notes
is estimated through changes in the various statistical measures of Supplemental material for this article is available at http://sur.rockefeller.
EEG channels taken in isolation, such as the bispectral index and edu/Plone/lab-members/leandro-alonso/. Animation showing evolu-
spectral entropy (Palanca et al., 2009). Despite years of research tion of stability analysis during reversible loss of consciousness. This
and use in clinical practice, these DOA monitors fail to prevent material has not been peer reviewed.
intra-operative awareness and postoperative recall (Avidan et al.,
2011). One reason for this failure is that, in contrast to the slow References
Alonso LM, Proekt A, Schwartz TH, Pryor KO, Cecchi GA, Magnasco MO
progression measured by the criticality index, there is no uniform
(2014) Dynamical criticality during induction of anesthesia in human
signature of DOA in the statistical structure of individual record- ECoG recordings. Front Neural Circuits 8:20. CrossRef Medline
ings. This is consistent with the notion that consciousness is likely Austin GM, Grant FC (1955) Physiologic observations following total
an emergent phenomenon that involves correlated activity dis- hemispherectomy in man. Surgery 38:239 258. Medline
tributed among multiple brain regions. Here, we suggest that Avidan MS, Jacobsohn E, Glick D, Burnside BA, Zhang L, Villafranca A, Karl
stability is one such emergent property exhibiting robust changes L, Kamal S, Torres B, OConnor M, Evers AS, Gradwohl S, Lin N, Palanca
BJ, Mashour GA (2011) Prevention of intraoperative awareness in a
with onset of anesthesia and is independent of individual differ-
high-risk surgical population. N Engl J Med 365:591 600. CrossRef
ences in the spectral properties of individual recordings and of Medline
the specific identity of anesthetic agents. Thus, stability analysis Bak P, Tang C, Wiesenfeld K (1987) Self-organized criticality: an explana-
may prove useful in characterizing DOA in clinical settings. tion of the 1/f noise. Phys Rev Lett 59:381384. CrossRef Medline
Our recordings were confined to a single hemisphere. Al- Beggs JM, Plenz D (2003) Neuronal avalanches in neocortical circuits.
though administration of anesthetic to a single hemisphere J Neurosci 23:1116711177. Medline
(Wada test) can trigger confusion, disorientation, and amnesia, it Boly M, Moran R, Murphy M, Boveroux P, Bruno MA, Noirhomme Q,
Ledoux D, Bonhomme V, Brichant JF, Tononi G, Laureys S, Friston K
does not render one unconscious (Posner et al., 2007). Further- (2012) Connectivity changes underlying spectral EEG changes during
more, en bloc resection of an entire cerebral hemisphere does not propofol-induced loss of consciousness. J Neurosci 32:70827090.
routinely produce a comatose or a stuporous state (Austin and CrossRef Medline
Grant, 1955). Breshears JD, Roland JL, Sharma M, Gaona CM, Freudenburg ZV, Tempel-
Solovey, Alonso et al. Stabilization of Cortical Dynamics in Unconsciousness J. Neurosci., July 29, 2015 35(30):10866 10877 10877

hoff R, Avidan MS, Leuthardt EC (2010) Stable and dynamic cortical and emergent computation. Phys D Nonlinear Phenom 42:1237.
electrophysiology of induction and emergence with propofol anesthesia. CrossRef
Proc Natl Acad Sci U S A 107:21170 21175. CrossRef Medline Lee U, Muller M, Noh GJ, Choi B, Mashour GA (2011) Dissociable network
Brown EN, Lydic R, Schiff ND (2010) General anesthesia, sleep, and coma. properties of anesthetic state transitions. Anesthesiology 114:872 881.
N Engl J Med 363:2638 2650. CrossRef Medline CrossRef Medline
Brown EN, Purdon PL, Van Dort CJ (2011) General anesthesia and altered Leuthardt EC, Schalk G, Wolpaw JR, Ojemann JG, Moran DW (2004) A
states of arousal: a systems neuroscience analysis. Annu Rev Neurosci brain-computer interface using electrocorticographic signals in humans.
34:601 628. CrossRef Medline J Neural Eng 1:6371. CrossRef Medline
Casali AG, Gosseries O, Rosanova M, Boly M, Sarasso S, Casali KR, Casarotto Magnasco MO, Piro O, Cecchi GA (2009) Self-tuned critical anti-Hebbian
S, Bruno MA, Laureys S, Tononi G, Massimini M (2013) A theoretically networks. Phys Rev Lett 102:258102258104. CrossRef Medline
based index of consciousness independent of sensory processing and be- Maksimow A, Sarkela M, Lngsjo JW, Salmi E, Kaisti KK, Yli-Hankala A,
havior. Sci Transl Med 5:198ra105. CrossRef Medline Hinkka-Yli-Salomaki S, Scheinin H, Jaaskelainen SK (2006) Increase in
Chen G, Gao W, Reinert KC, Popa LS, Hendrix CM, Ross ME, Ebner TJ high frequency EEG activity explains the poor performance of EEG spec-
(2005) Involvement of kv1 potassium channels in spreading acidification tral entropy monitor during S-ketamine anesthesia. Clin Neurophysiol
and depression in the cerebellar cortex. J Neurophysiol 94:12871298. 117:1660 1668. CrossRef Medline
CrossRef Medline Massimini M, Ferrarelli F, Huber R, Esser SK, Singh H, Tononi G (2005)
Chialvo DR (2010) Emergent complex neural dynamics. Nat Phys 6:744 Breakdown of cortical effective connectivity during sleep. Science 309:
750. CrossRef 2228 2232. CrossRef Medline
Cimenser A, Purdon PL, Pierce ET, Walsh JL, Salazar-Gomez AF, Harrell PG, Miller KJ, Sorensen LB, Ojemann JG, den Nijs M (2009) Power-law scaling
Tavares-Stoeckel C, Habeeb K, Brown EN (2011) Tracking brain states in the brain surface electric potential. PLoS Comput Biol 5:e1000609.
under general anesthesia by using global coherence analysis. Proc Natl CrossRef Medline
Acad Sci U S A 108:8832 8837. CrossRef Medline Mitra P, Bokil H (2008) Observed brain dynamics. New York: Oxford UP.
Concas A, Santoro G, Mascia MP, Serra M, Sanna E, Biggio G (1990) The Mora T, Bialek W (2011) Are biological systems poised at criticality? J Stat
general anesthetic propofol enhances the function of gamma- Phys 144:268 302. CrossRef
aminobutyric acid-coupled chloride channel in the rat cerebral cortex. Mukamel EA, Wong KF, Prerau MJ, Brown EN, Purdon PL (2011) Phase-
J Neurochem 55:21352138. CrossRef Medline based measures of cross-frequency coupling in brain electrical dynamics
Devor M, Zalkind V (2001) Reversible analgesia, atonia, and loss of con- under general anesthesia. Conf Proc IEEE Eng Med Biol Soc 2011:1981
sciousness on bilateral intracerebral microinjection of pentobarbital. Pain 1984. CrossRef Medline
94:101112. CrossRef Medline Nagasaka Y, Shimoda K, Fujii N (2011) Multidimensional recording
Ferrarelli F, Massimini M, Sarasso S, Casali A, Riedner BA, Angelini G, (MDR) and data sharing: an ecological open research and educational
Tononi G, Pearce RA (2010) Breakdown in cortical effective connectiv- platform for neuroscience. PLoS One 6:e22561. CrossRef Medline
ity during midazolam-induced loss of consciousness. Proc Natl Acad Sci Neumaier A, Schneider T (2001) Estimation of parameters and eigenmodes
U S A 107:26812686. CrossRef Medline of multivariate autoregressive models. ACM Trans Math Softw 27:2757.
Ffytche DH, Howard RJ, Brammer MJ, David A, Woodruff P, Williams S CrossRef
(1998) The anatomy of conscious vision: an fMRI study of visual hallu- OConnor T, Wong HY (2012) Emergent properties. In: The Stanford en-
cinations. Nat Neurosci 1:738 742. CrossRef Medline cyclopedia of philosophy (Zalta EN, ed). http://plato.stanford.edu/
Franks NP (2008) General anaesthesia: from molecular targets to neuronal Palanca BJA, Mashour GA, Avidan MS (2009) Processed electroencephalo-
pathways of sleep and arousal. Nat Rev Neurosci 9:370 386. CrossRef gram in depth of anesthesia monitoring. Curr Opin Anaesthesiol 22:553
Medline 559. CrossRef Medline
Garg R, Cecchi GA, Rao AR (2011) Full-brain auto-regressive modeling Peltier SJ, Kerssens C, Hamann SB, Sebel PS, Byas-Smith M, Hu X (2005)
(FARM) using fMRI. Neuroimage 58:416 441. CrossRef Medline Functional connectivity changes with concentration of sevoflurane anes-
Grant IS, Nimmo WS, McNicol LR, Clements JA (1983) Ketamine disposi- thesia. Neuroreport 16:285288. CrossRef Medline
tion in children and adults. Br J Anaesth 55:11071111. CrossRef Medline Posner JB, Saper CB, Schiff N, Plum F (2007) Plum and Posners diagnosis
Hudspeth AJ, Julicher F, Martin P (2010) A critique of the critical cochlea: of stupor and coma (Contemporary neurology series, Book 71), Ed 4.
Hopfa bifurcationis better than none. J Neurophysiol 104:1219 New York: Oxford UP.
1229. CrossRef Medline Ritaccio A, Brunner P, Cervenka MC, Crone N, Guger C, Leuthardt E, Oost-
Imas OA, Ropella KM, Ward BD, Wood JD, Hudetz AG (2005) Volatile enveld R, Stacey W, Schalk G (2010) Proceedings of the first interna-
anesthetics enhance flash-induced gamma oscillations in rat visual cortex. tional workshop on advances in electrocorticography. Epilepsy Behav
Anesthesiology 102:937947. CrossRef Medline 19:204 215. CrossRef Medline
Jia F, Pignataro L, Schofield CM, Yue M, Harrison NL, Goldstein PA (2005) Schnider TW, Minto CF, Shafer SL, Gambus PL, Andresen C, Goodale DB,
An extrasynaptic GABAA receptor mediates tonic inhibition in thalamic Youngs EJ (1999) The influence of age on propofol pharmacodynamics.
VB neurons. J Neurophysiol 94:4491 4501. CrossRef Medline Anesthesiology 90:15021516. CrossRef Medline
John ER, Prichep LS, Kox W, Valdes-Sosa P, Bosch-Bayard J, Aubert E, Tom Solovey G, Miller KJ, Ojemann JG, Magnasco MO, Cecchi GA (2012) Self-
M, di Michele F, Gugino LD, diMichele F (2001) Invariant reversible regulated dynamical criticality in human ECoG. Front Integr Neurosci
QEEG effects of anesthetics. Conscious Cogn 10:165183. CrossRef 6:44. CrossRef Medline
Medline Thompson E, Varela FJ (2001) Radical embodiment: Neural dynamics and
Krystal JH, DSouza DC, Mathalon D, Perry E, Belger A, Hoffman R (2003) consciousness. Trends Cogn Sci 5:418 425. CrossRef Medline
NMDA receptor antagonist effects, cortical glutamatergic function, and Thomson DJ (1982) Spectrum estimation and harmonic analysis. Proc
schizophrenia: toward a paradigm shift in medication development. Psy- IEEE 70:10551096. CrossRef
chopharmacology (Berl) 169:215233. CrossRef Medline Yanagawa T, Chao ZC, Hasegawa N, Fujii N (2013) Large-scale information
Ku SW, Lee U, Noh GJ, Jun IG, Mashour GA (2011) Preferential inhibition flow in conscious and unconscious states: an ECoG study in monkeys.
of frontal-to-parietal feedback connectivity is a neurophysiologic corre- Valdes-Sosa PA, ed. PLoS One 8:e80845. CrossRef Medline
late of general anesthesia in surgical patients. PLoS One 6:e25155. Yan XH, Magnasco MO (2012) Input-dependent wave attenuation in a
CrossRef Medline critically-balanced model of cortex. PLoS One 7:e41419. CrossRef
Langton CG (1990) Computation at the edge of chaos: Phase transitions Medline