Real-time segmentation and tracking of brain metabolic state in ICU EEG recordings of burst suppression

We provide a method for estimating brain metabolic state based on a reduced-order model of EEG burst suppression. The model, derived from previously suggested biophysical mechanisms of burst suppression, describes important electrophysiological features and provides a direct link to cerebral metabolic rate. We design and fit the estimation method from EEG recordings of burst suppression from a neurological intensive care unit and test it on real and synthetic data.
Real-Time Segmentation and Tracking of Brain Metabolic State in ICU
EEG Recordings of Burst Suppression
M. Brandon Westover
, ShiNung Ching
, Mouhsin M. Shafi
, Sydney S. Cash
and Emery N. Brown
Abstract We provide a method for estimating brain
metabolic state based on a reduced-order model of EEG burst
suppression. The model, derived from previously suggested
biophysical mechanisms of burst suppression, describes impor-
tant electrophysiological features and provides a direct link
to cerebral metabolic rate. We design and fit the estimation
method f rom EEG recordings of burst suppression from a
neurological intensive care unit and test it on real and synthetic
Burst suppression is an electroencephalographic (EEG)
pattern in which periods of high voltage activity (bursts)
alternate with periods of isoelectric quiescence (suppression)
(see Figure 1). It is characteristic of a profoundly inacti-
vated brain and occurs in conditions such as deep general
anesthesia [1], hypothermia [2] and coma [3]. That these
different conditions lead to seemingly similar brain activity
suggests that burst suppression is the result of a f undamental,
low-order process that is prominent when higher-level brain
activity is depressed.
The main features of burst suppression have been well de-
scribed [4], [5], [6]. Classically, burst suppression is thought
to be a global state where bursts begin and end nearly
simultaneously across the entire scalp. It is different from
typical faster EEG oscillatory patterns, in that suppression
epochs can be very irregular and may last several seconds.
Importantly, burst suppression is not a homogeneous state
but, instead, varies continuously as a function of brain inacti-
vation. As the brain becomes progressively more inactivated,
the amount of suppression, relative to the amount of burst,
increases. This variation has been traditionally quantified
with the burst suppression ratio [6], which measures the
amount of suppression in a sliding window of EEG data.
Recent research on the burst suppression probability [7]
*This work has been supported by NIH DP1-OD003646 (to ENB).
SC holds a Career Award at the Scientific Interface from the Burroughs-
Wellcome Fund
Department of Anesthesia, Critical Care and Pain Medicine, Mas-
sachusetts General Hospital & Harvard Medical School, Boston, MA
Department of Brain and Cognitive Science, Massachusetts Institute of
Technology, Cambridge, MA
Department of Electrical and Systems Engineering, Washington Univer-
sity in St. Louis, St. Louis, MO, USA
Harvard-Massachusetts Institute of Technology Division of Health Sci-
ences and Technology, Massachusetts Institute of Technology, Cambridge,
Institute for Medical Engineering and Sciences, Massachusetts Institute
of Technology, Cambridge, MA
Department of Neurology, Massachusetts General Hospital & Harvard
Medical School, Boston, MA
Fig. 1. Example of burst suppression. (A) Continuous EEG activity, (B)
Burst suppression
(BSP) has provided a statistically rigorous, and window-free,
approach to estimating the burst suppression state.
Here, we introduce a method for estimating not simply
burst suppression, but the underlying brain metabolic state.
Our method is based on a recent nonlinear, biophysical
model [5], which attributes the parametric increase in sup-
pression duration with brain inactivation to decreases in brain
We begin by characterizing the relationship between brain
metabolic state and observable EEG features, namely the
lengths and variability of bursts and suppressions. We then
introduce and fit a reduced state-space model of burst
suppression to recordings from neurological intensive care
unit (ICU) patients. From this model, we demonstrate the
inference of the underlying metabolic state.
The remainder of this paper is organized as follows.
Section II provides a brief background on the biophysical
mechanisms of burst suppression and the resulting models.
Section III introduces the reduced state-space model and
methods for metabolic state inference. Brief conclusions are
formulated in Section IV.
A. Neurophysiology of Burst Suppression
Although many features of burst suppression have been
described, the neurophysiological mechanisms that are re-
sponsible for creating it are less well understood. In the
context of general anesthesia, the early work by Steriade [8]
helped establish certain neural correlates of burst suppres-
sion, describing the participation of different cell types in
bursts and suppressions, though an underlying mechanism
was not suggested. Other studies [9] have suggested that
burst suppression involves enhanced excitability in cortical
networks, and have implicated fluctuations in calcium as
related to the alternations between bursts and suppressions.
B. Existing Models of Burst Suppression
A unifying biophysical model for burst suppression one
that accounts for its characteristics, and also its range of
etiologies was recently proposed [5]. The key insight of
35th Annual International Conference of the IEEE EMBS
Osaka, Japan, 3 - 7 July, 2013
978-1-4577-0216-7/13/$26.00 ©2013 IEEE 7108
Fig. 2. ATP-based mechanism for burst suppression. ATP is depleted
through the course of each burst, leading to suppression. During suppression
ATP gradually recovers until, eventually, activity begins again
the model is that each of the conditions associated with burst
suppression (general anesthesia, hypoxic/ischemic coma, hy-
pothermia) is associated with decreased cerebral metabolic
rate of oxygen (CMRO). The model links this decrease in
CMRO to deficiencies in ATP (adenosine triphosphate, the
energetic substrate for neuronal activity) production in corti-
cal networks (see Figure 2). The termination of each burst is
a reflection of ATP consumption due to the neuronal activity
underlying fast EEG oscillations, whereas suppressions are
governed by the slow dynamics of ATP regeneration.
This model provided an explanation for why three cardinal
features of burst suppression the spatial synchrony of bursts
onsets across the scalp, the increase in suppression durations
with increasing brain inactivation, and the long timescales
of suppressions can arise across its disparate etiologies.
The present paper is intended to provide a simplification
of the model in [5], and simultaneously, to describe a
fourth cardinal feature that was not previously explored,
namely the variability of burst lengths at different burst
suppression levels. This, in turn, enables the estimation of
brain metabolism (CMRO) from EEG recordings.
A. Simplified Burst Suppression Model
Based on [5], we present a reduced order state-space
model for burst suppression governed by the following:
˙a = k
(1 a) k
u (a) . (1)
Here, a(t) is the concentration of local ATP in a cortical
region, k
is the rate of ATP consumption during each burst,
is the rate of ATP regeneration during each suppression,
and u (a ) indicates whether burst activity can or cannot be
sustained. We select
u (a) =
1 ˙a > 0 and 0 a < α
0 otherwise,
meaning that burst activity can only be initiated when ATP
levels increase beyond the threshold α.
By fixing the parameter k
= 1, (1) can be rewritten as
˙a = x (1 a) u (a) , (3)
where x, a value from 0 to 1, is the brain metabolic state.
A value of x = 0 corresponds to f ull CMRO (when ATP
regeneration equals consumption), while x = 1 is complete
metabolic depression.
Fig. 3. Example of model output for different values of metabolic state.
(A) x = 0.8, (B) x = 0.1. Simulated EEG signal shown for schematic
purposes only.
Figure 3 illustrates the output of the model for two dif-
ferent values of x. When x is moderate, the model produces
epochs of burst and suppression that are commensurate in
length. When x is reduced to a low value, the bursts are much
shorter (due to more rapid consumption) and the suppressions
are longer (due to slower regeneration).
The model (1) offers increased analytical tractability as
compared to the full nonlinear model in [5]. In particular,
we can derive explicit expressions for burst and suppression
lengths (L
and L
) at different metabolic state levels as:
(x) = log ¯α/x
(x) = log
1 ¯αx
¯α = 1 α. (5)
The burst suppression state itself can then be quantified in
terms of the suppression lengths, relative to the total length
of a burst-suppression cycle, specifically:
(x) =
(x) + L
log ¯α
log ¯α + log
1 ¯αx
Note that in practice, (6) can be estimated using the burst
suppression probability (BSP) [7] algorithm. Through (4)
and ( 6), we can estimate x based on measurement of burst
suppression and calculation of burst and suppression lengths
from the EEG.
B. Automatic EEG Segmentation
In order to infer the metabolic state in our model, we
must first establish a method to segment EEG recordings into
Fig. 4. Examples of ICU burst suppression with automatic segmentation.
Segmented bursts (i.e., n
= 0) are shown in red, while suppressions
(i.e., n
= 1) are blue. (A,B) Patterns containing epileptiform spikes,
(C,D) Patterns with distinct bursts and suppressions, (E,F) Patterns with
less distinct bursts
bursts and suppressions. That is, if x
, t = 0, 1, 2, ... is the
sampled EEG signal, then we must obtain a corresponding
binary series n
where n
= 1 if x
is in a suppression and
0 if it is in a burst.
While several algorithms have been developed for this
purpose [10], [11], we choose to use adaptive variance
thresholding as follows:
= γx
+ (1 γ)¯y
= γ(x
+ (1 γ)s
1 s
< v
0 otherwise,
where γ is a tunable filter parameter and v
is an
amplitude threshold. We have applied this method to a variety
of EEG recordings of burst suppression from the neurological
ICU [12] and, as illustrated in Figure 4, it can reliably
segment the EEG into bursts and suppressions.
From the binary signal n
it is straightforward to obtain
empirical lengths of bursts and s uppression (simply, the
lengths of consecutive 0s or 1s), facilitating estimation of
metabolic state.
C. Inference of Metabolic State
In order to estimate the metabolic state x as a function
of time, and to account for anticipated stochastic effects in
burst and suppression lengths, we introduce a probabilistic
model as follows:
= min (max (x
+ v
, 0) , 1) , v
N (0, σ) (10)
This model is a rectified Gaussian random walk and, if σ
is suitably small, implies that x does not exhibit large and
sudden temporal changes.
We will, furthermore, make a Markovian assumption that
p (x
, x
, ..., x
) = p (x
) (11)
and, in particular, that
p (n
|H (n, L, x)) = p (n
, L
, x
) , (12)
where L
denotes the length of the i
event (either a burst
or suppression) and H(·) denotes the entire history.
What remains is to define the probabilities for continua-
p (n
= 1|n
= 1, L
, x
p (n
= 0|n
= 0, L
, x
and switching:
p (n
= 1|n
= 0, L
, x
p (n
= 0|n
= 1, L
, x
Based on the characterization from (4) and (6), we choose to
model these probabilities using the Weibull hazard function
h (t; λ, θ) =
, (15)
and its cumulative distribution function (CDF)
F (t; λ, θ) = 1 exp
. (16)
Note that (15) and (16) are common in medical survival
analysis and reliability engineering.
We proceed to fit (16) to the burst suppression level,
which can be well-estimated from the segmented EEG using
the burst suppression probability (BSP) algorithm [7]. In
particular, we compute an empirical CDF for (13) and ( 14)
by finding, for each suppression and burst, the correspond-
ing BSP level. We then fit (16) to these CDFs using the
λ (BSP ) = a
exp (BSP × b
) , θ = c
for bursts and
λ (BSP ) = a
exp ((1 BSP ) × b
) , θ = c
for suppressions. For this, we use a nonlinear least squares
numerical method over the free parameters a
, b
, c
. Figure
5 illustrates the empirical CDF for switching f rom the EEGs
of 20 ICU patients
and the resulting fit for two BSP levels.
In both cases, the functions (17)-(18), together with (16),
are able to closely match the empirical CDFs. Figure 6
illustrates these fits, as compared to the empirical CDFs for
switching, across the entire range of BSP values. As shown,
the resulting model characterization is close to what we find
from our data.
The one-to-one relationship (6) relates our continuation
and switching functions (f or BSP) directly to metabolic state.
We can thus proceed t o perform inference of the metabolic
state through a direct application of Bayes formula to (12).
We illustrate the estimation using synthetic data generated
from the model (1). Figure 7A illustrates the burst and
suppression output (n
) from the model when x(t) is a
realization of the random walk (10). Through (12)-(16), and
These data were collected at the Massachusetts General Hospital as part
of routine clinical monitoring and with institutional review board approval.
Fig. 5. Example of CDF for switching and resulting fits for two BSP
levels. (A) BSP of 0.2, (B) BSP of 0.7
Fig. 6. Empirical
and fit switching probability functions vs. BSP for
suppressions (A,B) and bursts (C,D). The fitted functions (B,D) closely
match the empirical CDFs (A,C). White indicates values close to 1 (high
probability of switching) whereas black indicates values close to 0 (low
probability of switching).
the fits of (17)-(18) obtained empirically from our ICU data
(i.e., Figure 6), we obtain the posterior probability density
function of metabolic state x at each point in time. The
mean of each distribution is the metabolic state estimate,
which is plotted in Figure 7C and compared with the true
value. Clearly, the estimate closely tracks the true value. One
feature of note is that the estimate does not immediately
change at each switch from burst to suppression. Instead, and
consistent with our model, it remains stable during each burst
and suppression until such time as its length is improbable
given the current BSP estimate.
We have provided a reduced-order model for burst sup-
pression that links the EEG directly to reductions in cerebral
metabolic rate. From this model, we developed a probabilis-
tic inference s cheme to estimate brain metabolic state from
measured EEG activity. The resulting method was fit and
tested on EEG data gathered from patients in the neurological
ICU. We then tested the method on synthetic burst suppres-
sion data, showing correct inference of metabolic state.
Further testing is, of course, necessary to validate the
use of this method in the clinical setting. Nevertheless, the
model provides justification for t he practice of pharmaco-
logically inducing burst suppression as a therapeutic target
for brain protection in neurological intensive care settings
such as unrelenting seizures (refractory status epilepticus),
severe traumatic brain injury, and in cardiac surgery during
Fig. 7. Example of inference of metabolic state from simulated burst
suppression. (A) Simulated bursts and suppressions from (1), (B) Proba-
bility d ensity function of metabolic state x estimated from (12)-(18) (and
corresponding fits). (C) Inferred x (red trace) as compared to the true value
used to generate (A) (blue trace).
circulatory arrest [13]. The model and estimation scheme
may also help inform strategies for optimizing burst sup-
pression when using anesthetic drugs. An eventual goal is to
provide a neurophysiologically-principled basis for inferring
and tracking brain metabolism in the ICU or surgical settings.
[1] D. L. Clark and B. S. Rosner, “Neurophysiologic effects of general
anesthetics. i. the electroencephalogram and sensory evoked responses
in man.” Anesthesiology, vol. 38, no. 6, pp. 564–582, Jun 1973.
[2] M. M. Stecker, A. T. Cheung, A. Pochettino, G. P. Kent, T. Patterson,
S. J. Weiss, and J. E. Bavaria, “Deep hypothermic circulatory arrest:
Ii. changes in electroencephalogram and evoked potentials during
rewarming.” Ann Thorac Surg, vol. 71, no. 1, pp. 22–28, Jan 2001.
[3] G. B. Young, “The eeg in coma.” J Clin Neurophysiol, vol. 17, no. 5,
pp. 473–485, Sep 2000.
[4] F. Amzica, “Basic physiology of burst-suppression.” Epilepsia, vol.
50 Suppl 12, pp. 38–39, Dec 2009.
[5] S. Ching, P. L. Purdon, S. Vijayan, N. J. Kopell, and E. N. Brown,
“A neurophysiological-metabolic model for burst suppression.” Proc
Natl Acad Sci U S A, vol. 109, no. 8, pp. 3095–3100, Feb 2012.
[6] W. P. Akrawi, J. C. Drummond, C. J. Kalkman, and P. M. Patel,
“A comparison of the electrophysiologic characteristics of eeg burst-
suppression as produced by isoflurane, thiopental, etomidate, and
propofol.” J Neurosurg Anesthesiol, vol. 8, no. 1, pp. 40–46, Jan 1996.
[7] J. J. Chemali, K. F. K. Wong, K. Solt, and E. N. Brown, A state-space
model of the burst suppression ratio, IEEE EMBC 2011, Boston, MA.
[8] M. Steriade, F. Amzica, and D. Contreras, “Cortical and thala-
mic cellular correlates of electroencephalographic burst-suppression.”
Electroencephalogr Clin Neurophysiol, vol. 90, no. 1, pp. 1–16, Jan
[9] D. Kroeger and F. Amzica, “Hypersensitivity of the anesthesia-induced
comatose brain.” J Neurosci, vol. 27, no. 39, pp. 10 597–10 607, Sep
[10] J. Lofhede, N. Lfgren, M. Thordstein, A. Flisberg, I. Kjellmer, and
K. Lindecrantz, “Classification of burst and suppression in the neonatal
electroencephalogram.” J Neural Eng, vol. 5, no. 4, pp. 402–410, Dec
[11] M. Sarkela, S. Mustola, T. Seppnen, M. Koskinen, P. Lepola,
K. Suominen, T. Juvonen, H. Tolvanen-Laakso, and V. Jntti, Auto-
matic analysis and monitoring of burst suppression in anesthesia.” J
Clin Monit Comput, vol. 17, no. 2, pp. 125–134, Feb 2002.
[12] M. B. Westover, S. Ching, M. Shafi, S. Cash, and E. N. Brown, “Real-
time segmentation of burst suppression patterns in critical care eeg
monitoring, submitted, 2012.
[13] R. Hall and J. Murdoch, “Brain protection: physiological and pharma-
cological considerations. part ii: The pharmacology of brain protec-
tion.” Can J Anaesth, vol. 37, no. 7, pp. 762–777, Oct 1990.
    • "All EEG data were segmented into a binary sequence of burst and suppression epochs using a previously validated automated suppression detection algorithm [11]. The resulting binary signal was then filtered to produce a continuous measure of burst suppression depth, the BSP ('burst suppression probability'), which quantifies the instantaneous the probability of being in the suppressed state [12]. "
    [Show abstract] [Hide abstract] ABSTRACT: Millions of patients are admitted each year to intensive care units (ICUs) in the United States. A significant fraction of ICU survivors develop lifelong cognitive impairment, incurring tremendous financial and societal costs. Delirium, a state of impaired awareness, attention and cognition that frequently develops during ICU care, is a major risk factor for post-ICU cognitive impairment. Recent studies suggest that patients experiencing electroencephalogram (EEG) burst suppression have higher rates of mortality and are more likely to develop delirium than patients who do not experience burst suppression. Burst suppression is typically associated with coma and deep levels of anesthesia or hypothermia, and is defined clinically as an alternating pattern of high-amplitude " burst " periods interrupted by sustained low-amplitude " suppression " periods. Here we describe a clustering method to analyze EEG spectra during burst and suppression periods. We used this method to identify a set of distinct spectral patterns in the EEG during burst and suppression periods in critically ill patients. These patterns correlate with level of patient sedation, quantified in terms of sedative infusion rates and clinical sedation scores. This analysis suggests that EEG burst suppression in critically ill patients may not be a single state, but instead may reflect a plurality of states whose specific dynamics relate to a patient's underlying brain function.
    Full-text · Conference Paper · Aug 2015 · Clinical Neurophysiology
    • "8. Seizure detection algorithms and automated background assessment: Most automated seizure detection algorithms were developed for ictal patterns seen in patients with established epilepsy and have not been validated in ICU populations with acute symptomatic seizures (Sackellares et al., 2011). Automated analysis of background patterns (e.g., burst suppression periodic patterns) is an active research area but is not in routine clinical use (Cloostermans et al., 2011; Shibasaki et al., 2014; Westover et al., 2013). "
    [Show abstract] [Hide abstract] ABSTRACT: Introduction: Critical Care Continuous EEG (CCEEG) is a common procedure to monitor brain function in patients with altered mental status in intensive care units. There is significant variability in patient populations undergoing CCEEG and in technical specifications for CCEEG performance. Methods: The Critical Care Continuous EEG Task Force of the American Clinical Neurophysiology Society developed expert consensus recommendations on the use of CCEEG in critically ill adults and children. Recommendations: The consensus panel describes the qualifications and responsibilities of CCEEG personnel including neurodiagnostic technologists and interpreting physicians. The panel outlines required equipment for CCEEG, including electrodes, EEG machine and amplifier specifications, equipment for polygraphic data acquisition, EEG and video review machines, central monitoring equipment, and network, remote access, and data storage equipment. The consensus panel also describes how CCEEG should be acquired, reviewed and interpreted. The panel suggests methods for patient selection and triage; initiation of CCEEG; daily maintenance of CCEEG; electrode removal and infection control; quantitative EEG techniques; EEG and behavioral monitoring by non-physician personnel; review, interpretation, and reports; and data storage protocols. Conclusion: Recommended qualifications for CCEEG personnel and CCEEG technical specifications will facilitate standardization of this emerging technology.
    Article · Apr 2015
  • [Show abstract] [Hide abstract] ABSTRACT: Objective: Deep hypothermia induces 'burst suppression' (BS), an electroencephalogram pattern with low-voltage 'suppressions' alternating with high-voltage 'bursts'. Current understanding of BS comes mainly from anesthesia studies, while hypothermia-induced BS has received little study. We set out to investigate the electroencephalogram changes induced by cooling the human brain through increasing depths of BS through isoelectricity. Methods: We recorded scalp electroencephalograms from eleven patients undergoing deep hypothermia during cardiac surgery with complete circulatory arrest, and analyzed these using methods of spectral analysis. Results: Within patients, the depth of BS systematically depends on the depth of hypothermia, though responses vary between patients except at temperature extremes. With decreasing temperature, burst lengths increase, and burst amplitudes and lengths decrease, while the spectral content of bursts remains constant. Conclusions: These findings support an existing theoretical model in which the common mechanism of burst suppression across diverse etiologies is the cyclical diffuse depletion of metabolic resources, and suggest the new hypothesis of local micro-network dropout to explain decreasing burst amplitudes at lower temperatures. Significance: These results pave the way for accurate noninvasive tracking of brain metabolic state during surgical procedures under deep hypothermia, and suggest new testable predictions about the network mechanisms underlying burst suppression.
    Full-text · Article · Jan 2015
Show more