Signal Processing - July 2016 - 95

slow variable remain unknown. But
fail when applied to different sets of
some have already been identified
patients [7]. One alternative solution is
during seizures, including extracelluto train the algorithm on a specific
lar K+ and O2 concentrations and
patient during a certain time and use
molecules linked to energy metabothis information to predict the incomlism, like adenosine triphosphate proing seizures in the same patient (perduction [4]. It would be particularly
sonalized medicine). This strategy was
interesting to monitor such molecular
used on a set of patients equipped with
activities in vivo before, during, and
electrocorticography grids for longafter seizures.
term recordings and a wireless system
Since seizure crossing can occur at
to transfer EEG signals to a data promultiple locations, the "force," and
cessing unit [8]. After training the syshence its biophysical
tem during several
mechanisms, may
weeks, the incoming
if the paths leading to
vary from one seiseizures could be
seizures are very diverse
zure to the other,
reasonably and reliin a given patient, their
evolve in time as a
ably detected profunction of environvided that the
fingerprints may also be
mental factors, or as
detection threshold
different.
the nature of the epiwas maintained at a
lepsy changes during
low value. But the
the patient's lifetime. These arguments
major caveat was a high rate of false
may also explain why seizure predicalarms [8]. Reliable seizure detection
tion based on EEG signals has mostly
would open the way to a closed-loop
failed, as its hidden assumption is that
system to stop seizures, e.g., with
preictal states follow rules that are unineurostimulation [9].
versal across patients and seizure types.
Perhaps the analysis of electroExperimental evidence demonstrates
physiological signals is not the best
the multiplicity of solutions, supporting
way to predict seizures and to underthe notion that the way networks
stand the mechanisms underlying their
approach seizure thresholds is not unigenesis. The fact that clinicians and
versal [4].
basic researchers focus on EEG signals has a historical origin, with the
discovery that brain activity could be
Conclusions
characterized by electrical signals. As
Epilepsy research has been conducted
mentioned previously, EEG signals
with the firm belief that magic bullets
reflect a highly integrated flux of
may be found to treat patients (with the
charged particles, mostly due to synultimate drug), predict seizures (the key
aptic activity. Since a seizure can be
EEG biomarker), and identify the epiobjectively characterized at the EEG
leptogenic zone (the key algorithm).
level, it was assumed that some
Perhaps, it is time to accept the comchanges in network activity would
plexity and multiplicity of solutions.
occur before the seizure. As menMarder's work is instrumental in that
tioned in this article, it is possible to
respect. Her laboratory rigorously dempostulate the existence of a "force"
onstrated that there exists a huge numdriving neuronal networks toward seiber of network configurations (or
zure threshold [4]. The main characdetailed molecular architectures) giving
teristic of this force is to evolve on a
rise to exactly the same type of network
very slow time scale (the fifth state
activity [3]. Seizures being an activity
variable in the epileptor), which natuendogenous to most neuronal networks,
rally points at slow molecular processthey are multiple ways to produce
es. The biophysical correlates of this
them. If the paths leading to seizures

IEEE Signal Processing Magazine

July 2016

are very diverse in a given patient, their
fingerprints may also be different.
Perhaps the solution lies in the use of
multimodal approaches, monitoring
different paths/mechanisms simultaneously. This would require the development of new technological tools [10],
[11] and conceptual approaches.

Author
Christophe Bernard (christophe.
bernard@univ-amu.fr) is the director of
research at Institut de Neuroscience des
Systèmes, Inserm UMR_S 1106, Aix
Marseille Université, France.

References

[1] E. M. Goldberg and D. A. Coulter, "Seizing the
opportunity: Stem cells take on epilepsy," Cell Stem
Cell, vol. 15, pp. 527-528, Nov. 2014.
[2] C. Bernard, S. Naze, T. Proix, and V. K. Jirsa,
"Modern concepts of seizure modeling," Int. Rev.
Neurobiol., vol. 114, pp. 121-153, Aug. 2014.
[3] A. A. Prinz, D. Bucher, and E. Marder, "Similar
network activity from disparate circuit parameters,"
Nat.Neurosci., vol. 7, no. 12, pp. 1345-1352, Dec.
2004.
[4] V. K. Jirsa, W. C. Stacey, P. P. Quilichini, A. I.
Ivanov, and C. Bernard, "On the nature of seizure
dynamics," Brain, vol. 137, pp. 2210-2230, Aug.
2014.
[5] R. G. Andrzejak, O. David, V. Gnatkovsky, F.
Wendling, F. Bartolomei, S. Francione, P. Kahane,
K. Schindler, and M. de Curtis, "Localization of epileptogenic zone on pre-surgical intracranial EEG
recordings: Toward a validation of quantitative signal
analysis approaches," Brain Topogr., vol. 28, no. 6,
pp. 832-837, Nov. 2015.
[6] H. E. Wang, C. G. Benar, P. P. Quilichini, K. J.
Friston, V. K. Jirsa, and C. Bernard, "A systematic
framework for functional connectivity measures,"
Front. Neurosci., vol. 8, pp. 405, Dec. 2014.
[7] W. Stacey, M. Le Van Quyen, F. Mormann, and
A. Schulze-Bonhage, "What is the present-day EEG
evidence for a preictal state?" Epilepsy Res., vol. 97,
no. 3, pp. 243-251, Dec. 2011.
[8] M. J. Cook, T. J. O'Brien, S. F. Berkovic, M.
Murphy, A. Morokoff, G. Fabinyi, et al. "Prediction
of seizure likelihood with a long-term, implanted seizure advisory system in patients with drug-resistant
epilepsy: A first-in-man study," Lancet Neurol., vol.
12, no. 6, pp. 563-571, June 2013.
[9] M. Morrell, "Brain stimulation for epilepsy: Can
scheduled or responsive neurostimulation stop seizures?," Curr. Opin. Neurol., vol. 19, no. 2, pp. 164-
168, Apr. 2006.
[10] A. Williamson, J. Rivnay, L. Kergoat, A. Jonsson,
S. Inal, I. Uguz, et al. "Controlling epileptiform activity with organic electronic ion pumps," Adv. Mater.,
vol. 27, no. 20, pp. 3138-3144, May 2015.
[11] D. Khodagholy, T. Doublet, P. Quilichini, M.
Gurfinkel, P. Leleux, A. Ghestem, et al. "In vivo
recordings of brain activity using organic transistors,"
Nat. Commun., vol. 4, 1575, 2013.

Table of Contents for the Digital Edition of Signal Processing - July 2016