Better treatment of epilepsy using computer models

| Hans Wolkers

Scientists from the University of Twente and UMC Utrecht are developing a mathematical model that can simulate the activity in different parts of the brain of epileptic patients. This helps to identify which brain parts play a role in epileptic attacks and may improve the surgical treatment. The project received a ZonMw Pearl, an honor only awarded to the most outstanding projects.

Epilepsy is a chronic brain condition, where patients suffer periodically from a massive, uncontrolled and highly synchronized electrical activity of the brain cortex, often resulting in a seizure. In some forms of epilepsy, such an attack starts in a specific brain region, from where it spreads across the brain. The patient may pass out and completely lose control, while unaware of what’s going on. When this happens at the wrong place and time, it can obviously be very dangerous.

Since the attacks are highly unpredictable, epilepsy puts an enormous psychological pressure on the patient and impacts their quality of life. ‘About two-thirds of all 200.000 epileptic patients in The Netherlands can successfully be treated using medication,’ says mathematician Hil Meijer, scientist at the faculty of EEMCS, University of Twente. ‘For some patients, surgery may be an option, where the brain parts, responsible for the attacks, are removed. However, it is tricky to exactly determine which brain parts are responsible for the attacks. This is where our model comes in.’ 

Complex cascade

According to UMC Utrecht neurologist Dr. Leijten, partner in the project, roughly half of the surgeries to treat epilepsy are successful. But in roughly fifty percent of the operated patients the attacks return after some time, or don’t disappear at all. Meijer: ‘Specific brain parts often are responsible for epileptic attacks, but also the nerves connecting different brain areas may play a role. The seizures are thus a complex cascade of events, where complete brain networks are involved.’

To better understand in which parts of the brain the attacks start, and how it spreads through the brain by connecting nerves, the scientists are developing a mathematical model that can simulate the different electrical signals in the complex brain network. ‘Mathematics is very suitable to study these dynamic networks and model the signals,’ Meijer explains. ‘Knowing where the attack starts and how the neural connections and other brain regions are involved in its propagation, may increase our understanding of how and where attacks occur and help us find more effective treatments.’ 

Identify connections

To build a reliable model that is capable of simulating not only electrical signals, but also an epileptic attack, requires a lot of patient data. The scientists collect these from epilepsy patients whose brain activity is monitored in the hospital. To measure the multitude of electrical signals in the brain and identify the different brain connections, doctors place a small silicon mat containing several electrodes on the brain cortex. During this procedure, they selectively stimulate certain areas of the brain and check for a response in other brain parts.

‘We try to map the neural pathways by which the electrical signals travel’

‘Clearly, the golden standard is to measure a spontaneous seizure and to see what’s going on in the patient’s brain. Where does the attack start and which neural connections play a role in its spreading across the brain?’, Meijer says. ‘In other words, we try to map the neural pathways by which the electrical signals travel, while spreading the attack’. These unique, patient-specific response data are subsequently put into the model, eventually resulting in a ‘digitalized patient’ in the computer. Meijer: ‘The ultimate goal is to model the entire brain of a patient, but currently, we can only model small brain parts. But that already has led to very promising results.’

Provoke an attack

Scientists are using the current model to simulate the effect of a possible brain operation to cure patients. First, they provoke an attack in the computer model. During this simulation, they can analyze and evaluate the problem areas in the brain as well as the nerves involved in the spreading. In a next step, they determine the impact of surgery by mimicking the removal of these brain parts, whereafter they try to arouse another digital seizure. If they fail to do so, this is good news and proof that the ‘simulated surgery’ was effective in stopping the attacks. If not, they do more research: will the outcome be better if other connections between brain parts are removed in the model?

‘Our model helps physicians to make smarter and more effective clinical decisions’

By modelling the different possible scenarios of the operation and the consequent impact on the seizures, the surgeon can select the best procedure before the actual surgery starts. ‘The models make it clear that the whole brain network is important in causing epileptic seizures, not just small brain parts,’ Meijer explains. ‘Our model helps physicians to make smarter and more effective clinical decisions. But the doctor remains fully in charge.’


Although an important step has been taken with building the first-generation models, and results are already applied in the clinical practice, improvements are needed. ‘Although we can digitally evoke an attack and simulate surgery to some degree, the translation to an individual patient remains somewhat tricky,’ Meijer says. ‘I would like to measure the activity of the whole brain in individual patients. Such data would provide more extensive input for the current model and make it more patient-specific, so that it can be used more effectively in a clinical setting.’

Stay tuned

Sign up for our weekly newsletter.