Epilepsy III

Part three of the blog post on epilepsy focuses on the control of seizures. When to stimulate? For how long to stimulate? Where to stimulate–is the epileptic focus known?

From predicting seizures to controlling them

The aberrant neural synchronization results in epileptic seizures. The previous blog discussed the necessary steps to achieve the “dynamical dream” of controlling these seizures: firstly, anticipating when the ictal event will occur, and secondly, sending a precisely timed, short electrical stimulation (Deep Brain Stimulation, DBS) to stop the abnormal synchrony pattern.

In this blog we will focus on the control of seizures. The attempt at stopping seizures is in reality an introduction, a precursor if you will, to the control of brain activity using electrical signals.

The relative success in Parkinson disease (will be a topic of a future blog) and epilepsy heralds a future – a future when brain control by electrical methods can be employed in a variety of scenarios.

Some of these scenarios still belong to science fiction, but fiction is becoming reality in a number of applications.

Thus, neurostimulation is used today not only in neurological pathologies but also to potentiate memory (Titiz, et al. 2017; Meisenhelter and Jobst, 2018).

In fact, there is an International Neuromodulation Society INS (www.neuromodulation.com), a non-profit group dedicated to the scientific development and awareness of neuromodulation. Only time will tell what may occur in the future. Now, let’s concentrate on the control of epileptic activity.

– D. Sornette, Critical Phenomena in Natural Sciences, Springer Verlag, Berlin, 2000.

When to stimulate?

The start of seizures represent dynamical bifurcations i.e. qualitative change in dynamics.
(A bifurcation occurs when a small change in the parameters of a system causes a sudden qualitative change in its behavior.)

Therefore, the start of seizures offers the best opportunity to use minimal stimulation currents to change the abnormal neural synchrony leading to the full ictus.

Stimuli (DBS, or any other that can be used for this purpose) are more efficient if given just at the critical point that occurs before the ictus. The reason for that is that systems are very sensitive to perturbations just at this precise moment.

Yes, these are somewhat abstract concepts(bifurcation, critical points…) but have been useful in some practical applications.

An additional advantage in this business of the control of brain activity is what Sornette’s sentence above says: we don’t need to care about the specific molecular/cellular mechanism that triggers the abnormal synchrony resulting in the seizure. This is because it all ends up in higher than normal neuronal synchronization that can be altered by delivering current pulses.

The activity of nervous systems has unstable/metastable states. The transient stabilization of these states manifests as synchronous oscillations in normal or abnormal brain rhythms. Therefore, it is a matter of favoring the normal, physiological brain rhythms.

In case of DBS, current stimulation is used, but medicines can be used too. However, chemical intervention is more difficult because of the reason aforementioned: the molecular events leading to high cellular synchronization may be very different from patient to patient, which has to be taken into account for molecular interventions.

But what kind of current pulses can be efficient at stopping an incoming seizure?

For how long do the DBS electrodes need to stimulate the brain areas where those electrodes are inserted?

To answer these questions needs research, basic research, so that a personalised DBS protocol can be created for each specific patient.

Many patients have very similar epileptic syndromes, and what is good for one will probably be good for the other. But an evaluation of the dynamics of epileptiform activity in each case is greatly advised before DBS is implemented.

Where to stimulate?

If the epileptic focus is known, then inserting the DBS electrodes there will be adequate. But even if the focus has not been found, there is hope.

A fundamental point has to be considered: seizures are a network phenomenon. One cell network over-stimulates another network, and that, in turn, stimulates others connected. In principle, the chain can be stopped at any stage.

In vitro and in vivo studies indicate that this is a possibility. Some cell networks (stages) will be better targets than others, but, again, the opportunity to stop the chain of abnormal high synchrony at any place remains. Hence, the issue of the focus may not be that crucial after all.

How to stimulate?

This depends on the dynamics leading to the epileptiform activity.

Our own 10-year research project ended up in an in vivo study. A closed-loop DBS was used to stop the generation of seizures in the rat models (Salam et al., 2015, 2016).

Specific signs of certain dynamical regimes were found leading to seizures. This led to the basic idea of stabilising the normal synchronization states using electrical stimuli (Deep Brain Stimulation / DBS). The whole procedure can be summarised like this: we recorded signals from the brain areas known as hippocampus. Our algorithm detected a change (decrease) in synchrony between two signals. That heralded a possible incoming seizure. At this moment a DBS of about 5–10 seconds was applied to those same brain areas at certain frequency (between 5 and 20 Hz) and low current amplitude (otherwise the tissue could be damaged).

Low frequencies of stimulation, 5 Hz specially, resulted in the rats becoming basically seizure-free. Higher frequencies (20 Hz) diminished seizures too but to a lesser degree than the low frequency. And too high frequencies (>50 Hz) could trigger seizures.

We chose low frequencies because both our own and other groups’ research showed that low frequency stimuli promoted interictal activity and avoided the route to the ictus. Interictal activity is what epileptologists call some brain activity that appears between seizures in the neurophysiological recordings.

As well, we tried an open-loop protocol. Stimuli that operate almost continuously, intermittently, are termed open-loop. An open-loop system is really a blind device. It does not have any “intelligent” mechanism built-in to monitor brain states and tune the stimulation schedule accordingly to improve seizure control.

In our work, the open loop paradigm that stimulated the brain at random, reduced seizures in the rats by 17%. In another research, the closed-loop (or on-demand, or feedback) paradigm that stimulated the brain when the change was detected, reduced seizures in the rats by ~90% (Salam et al., 2016).

The immense majority of clinical DBS protocols are of the open-loop variety. To my knowledge, there is only one closed-loop responsive DBS protocol approved for use in patients, the RNS system (Sun and Morrell, 2014).

Those interested in the full story can read chapters 2 and 3 of ‘The Brain-Behaviour Continuum’ [World Scientific] and some papers like Perez Velazquez et al., 2003, Khosravani et al., 2003.

A common approach to control pathophysiology, or how to circumvent molecular complexity

It is not without interest to note that the approach we used to study brain dynamics and stop seizures is very similar to the approach used to control cardiac arrhythmias (Christini et al., 2001).

I mentioned before the general applicability to many fields of the methods of study we conducted in epilepsy. I accentuated this fact using Sornette’s words: “The richness of out-of-equilibrium systems lies in the multiplicity of mechanisms generating similar behaviors”.

Let me emphasise it again: different causes, same results. This is very true in seizures. It is known that there are several molecular dysfunctions that will lead to the same result, higher than normal neuronal synchrony, and a subsequent ictus.

Some of these molecular events may sound even contradictory – to wit, promoting inhibition in some brain areas (thalamus mainly) favours seizures.

But how can this be, if seizures represent abnormal hyper-excitability of brain cells?

And, in addition, inhibitors of neuronal activity are normally used to stop seizures?

“Hyperexcitability normally leads to hypersynchrony that normally leads to neuropathology.” This admittedly very general scheme was proposed in my monograph The Brain-Behaviour Continuum.

Why sometimes inhibitory neurotransmission promotes ictal events? suffice to say, that sometimes the action of an inhibitory neurotransmitter leads to overexcitation of thalamic neurons. That will translate to hyperexcitability of cortical neurons, and the loop will be closed by the cortical cells exciting in turn those same thalamic cells already overexcited.

This is just an illustration of the complicated cellular/molecular machinery that, depending on the time and place, may go all wrong.

Nevertheless, the point I was trying to make is that a similar approach to finding dynamical regimes has been used in disparate systems: biochemical (Decroly and Goldbeter, 1987), physical (Bergé, et al., 1984), chemical (Roux, 1983), and cardiac (Christini and Glass, 2002; Christini et al., 2001) systems, plus our own study on epilepsy.

The methods are part of dynamical systems theory, popularly known as chaos theory, and they are quite technical. Chapter 2 of my aforementioned monograph contains a basic introduction to the theme.

The application of the knowledge gained by using these methods can be put to some very good use, like stopping Parkinsonian tremor, halting seizures or mending cardiac arrhythmias.

An extremely brief micro-survey of seizure control techniques

In addition to our own efforts, there have been many others that have used a wide variety of paradigms to stop the hypersynchrony in epilepsy.

The interest to control epileptiform activity has been among us for a long time. It was already in the 1960s that headsets with earphones that monitored EEG were used to trigger loud noises – in the hope that this could substantially perturb the brain activity before it goes into a seizure.

This may be seen as very non-specific, buta variety of non-specific perturbations seem to halt, transiently, paroxysmal activity. For instance, psychological interventions reduce seizure frequency in some patients (Schmid-Schönbein, 1998). The ketogenic diet (mainly used for children) help reduce seizure too, and so does deep brain stimulation (DBS) in the centromedian thalamic nucleus, plus many other brain areas (see the blog Epilepsy I).

Among the electrical protocols, we find proportional feedback, chaos control techniques, vagal nerve continuous stimulation, and many others. One review of the field is Lockman and Fisher 2009, but there is a very extensive literature on the topic that interested readers can find with any search.

I would like to point out that being non-specific may be good here. The reason is that we do not know what exact parameters are determining the state space where brain dynamics unfolds (the state space is used in dynamical systems theory to visualise and characterise the system’s dynamics).

Hence, to alter the “epileptic state space” and revert it to the “normal state space” with normal dynamics, one could change at the same time many parameters.

So, rather than being specific about one variable, just do something that will affect many parameters. Like vagal nerve stimulation that causes a global alteration of brain activity. Or the ketogenic diet aforementioned that results in many molecular events that reduce seizures in some patients.

And this non-specificity is, by the way, a feature of many medications. They promote inhibition, and thereby reduce propensity to ictal events by acting on several molecular mechanisms. Drugs are rarely very specific. It has been said that we know two things with certainty about a drug: the molecular mechanism by which it acts, and that there is another mechanism that it surely acts on, but remains unknown

On the other hand, what looks like very specific methods, are on the way. They are presaged by new technologies that are emerging and can be of great potential in this business. For example, optogenetics; some advances have been made recently in rodent models of epilepsy using this relatively novel technique, the method involving a closed-loop “optogenetic” stimulation. Details in Krook-Magnuson et al., 2013.

Technical addendum

Neurostimulation devices for treating epilepsy and other neurological disorders

In our epilepsy research, abovementioned chronically implanted electrodes were used to record and stimulate brain areas in rats. Novela electrodes, among others, were utilized. These have very good sensitivity and can capture individual neuronal spikes (action potentials).

In the end, local field potentials (LFPs) were used both for the analysis of the seizure precursor (commented in blog II) and to evaluate the results of the stimulation.

The Local Field Potential (LFP)

The reason that LFPs are so useful is that they represent the collective neuronal firing of many neurons.

One neuronal network activates another down the chain because many cells fire spikes in synchrony, but not all cells need to fire.

To understand the progression of excitability in brain tissue we would need to record from thousands of individual neurons, which is unfeasible with current methods. But the LFP offers us the solution: it already represents that collective firing. Hence, all these studies focus on LFPs.

The following figure from one of our papers (Salam et al., 2016), depicts an envisaged therapeutic neurostimulation system.

The neurostimulator could be a microchip with capabilities to analyze brain signals and to stimulate when needed. The device interfaces with the electrode arrays and wirelessly communicates with a computer.

Commercial neurostimulation systems for medically refractory epilepsy treatment are available from several companies, like Cyberonics, Medtronic or Neuropace.

Medtronic offered a programmable, open-loop DBS system, using about eight electrodes. The system delivers stimuli to the anterior nucleus of the thalamus at scheduled intervals.

As exemplified in the Figure 1, the main idea is to employ advanced electronic microchips. The devices need to perform the computations needed for early seizure detection, and to stimulate in a responsive, or closed-loop, fashion.

The advances in integrated circuit technologies enable miniaturization (less than 1cm3) to design single-chip implementations of neuroprostheses*

*neuroprosthesis, plural neuroprostheses = a replacement used to improve the function of an impaired nervous system.

Figure 2 shows one of our microchips we were developing while I worked in Toronto. It is compared in size with a Canadian coin. The small size would make it feasible to be implanted in a patient’s head (perhaps under the skin of the skull, as in Figure 1).

Regarding on-chip neural signal processing, ours is not the only chip that can perform online computations to detect seizures. Other integrated circuits that target seizure detection (but not early detection) have been reported (e.g. Verma, et. al., 2010).

Several neurostimulation systems with multiple recording and stimulation channels on the same chip have been reported in the very abundant engineering literature on this topic. For instance, a system with 128 biocompatible electrodes for recording and stimulation interface (Heer et al., 2006).

However, univariate algorithms* generally used in these devices/algorithms have been shown inferior for early detection of seizures. Detecting a seizure early significantly increases the chances of success in controlling the seizure by early electrical stimulation.

Bivariate algorithms** (such as neural synchrony computation on many channels) are effective but very expensive, and they cannot easily be implemented on a chip in real time.

A general survey on closed-loop neurostimulators where many technical details are expounded can be found in one of our recent papers (Kassiri et al., 2017).

*Univariate data consists of only one variable. The information deals with only one quantity that changes.

**Bivariate data involves two different variables. The analysis of this type of data deals with causes and relationships.

Univariate and bivariate data.

Author Jose Velazquez is a senior scientific advisory board member to Novela Neurotechnologies.

Leave a Reply

Your email address will not be published. Required fields are marked *