From computational models of epilepsy to clinical protocols

Epilepsy is a common neurological disease with a worldwide prevalence of
0.5-0.8%. About 25% of patients with seizures do not respond to drug treatment,
and the majority of these patients suffer from focal epilepsy. These patients
can successfully be treated with brain surgery if the region in the brain
responsible for the seizures can be identified. The epileptogenic zone is
delineated by using non-invasive measurements (MRI, EEG, MEG, PET). In 8-10% of
the patients, invasive measurements with seizure registration are required
after placement of electrodes directly on the cortical surface. The patient is
monitored for 1-2 weeks with video-EEG, waiting for spontaneous seizures to
occur and when the region of seizure onset is identified, this leads to a
strategy for surgical resection. At the same time, critical functional cortical
areas are localized by electrocortical stimulation mapping (ESM).

Intracranial EEG monitoring is stressful and prolonged, which poses risks of
infection and bleeding complications, and makes these recordings extremely
time-consuming, costly and scarcely available.
Procedures that shorten monitoring can be of great value, when these are
independed of the occurrence of seizures. Such methods, provided that they are
efficient and easy to use, could provide opportunities in avoiding chronic
registrations by intra-operative usage.

Single Pulse Electrocortical Stimulation (SPES) is such an alternative. Short
electrical stimuli are given at two neighbour electrodes on the cortex and the
EEG response in all the other electrodes is analyzed. These responses can be
classified by their timing and appearance as pathological (epileptiform) or
normal. It has been shown that pathological responses are reproducible and are
related to the seizure onset zone. Removal of the area of pathological
responses predicts favorable outcome of surgery. This implies that waiting for
spontaneous seizures might be unnecessary when replaced by systematically
screening the cortex for SPES pathological responses.
SPES and its analysis of large areas of cortex require many hours when all
electrode pairs are to be stimulated in a consecutive fashion. Therefore,
intra-operative implementation, which could even render prolonged monitoring
unnecessary, seems currently out of reach. Understanding of the effects of
cortical stimulation, necessary for optimization of the established clinical
protocol, is hampered by the limited possibilities to test all stimulus
parameters and response variables in patients. A translational model is needed
that allows us to investigate parameter settings, the best order of
stimulation, but also the effect of anti-epileptic or anesthetic drugs that
alter excitability. Traditional animal models for epilepsy are mostly based on
rodents and focus on hippocampal (archicortical) or generalized epilepsy, which
is hardly representative of human focal neocortical epilepsy. We feel that this
gap should be filled by computational models based on anatomical and
physiological knowledge.

Our goal is to build a faster SPES algorithm for execution and analysis which
produces the same information as the current, classic SPES protocol. This
faster protocol is obtained by predicting optimal stimulus parameters and
settings in a computer model, based on neural mass models that represent the
individual electrode space, and is self-adapting.

This is an observational study in 20 patients. Each patient will undergo the
two routine classic SPES sessions during chronic EEG monitoring. In the first
10-15 patients, the classic SPES will be used to build a computer algorithm. In
this algorithm, several settings for SPES will be tested. The settings with the
highest sensitivity, specificity and speed will be tested in these patients in
a third, extra SPES session for which we ask permission. The next 5-10 patients
are part of the validation set. In these patients, the computer will build a
SPES protocol itself with the algorithm, without using information from the
classic SPES recordings.

The extra SPES session will be performed while the patient has no obligations.
The patient does not notice anything during SPES and does not require any
instruction. This means that there is no burden and the risks are negligable.