Research and recent advances in epilepsy surgery

Can computer models help us improve the planning of epilepsy surgery?

The occurrence of seizures can pose a large burden on the quality of life of people suffering from epilepsy. When anti-epileptic medication is not enough to suppress the seizures, the clinical team often considers a resective surgery: the goal now is to remove the brain areas that are generating or generalizing the seizures. But the outcome is not as good as we would like: currently about 1 in every 3 patients who undergo epilepsy surgery continue to have seizures afterwards. In order to improve this outcome, so that patients have more changes of being seizure free after the resection, at Amsterdam UMC we are developing a computer model, personalized for each individual patient, that will help us identify the best resection for each case.

Want to know more? Check the video! You can also go through the information below to learn more about the model or our team. Do you have any inspiring ideas or questions? Check our forum, we will be happy to talk with you!

General Introduction

Basic concepts on the study of brain disorders

Computer Models

How do computer models of seizures work?

About us

Meet the researchers involved in this project and read about our current work

Model Demo

Visualize how seizures propagate in the model and test different resection strategies

Discussion Forum

In our discussion forum you can drop your questions or comments, and interact with the research team

Our pseudo-prospective study has now been published in the peer-reviewed journal Network Neuroscience:

In the special issue of December 2023 we give an overview of the project, describing the most important methodological improvements and results of the scientific papers mentioned below. For details, see: (in Dutch). An English translation can be found here.

In this preprint, we describe how the model can be applied in clinical practice. In the clinical setting, only information from various scans (EEG, MEG, MRI, PET) and clinical information is available before surgery. We used this information to estimate the probability that seizures originate in a particular brain region, and then the model accurately indicated how likely it is that a proposed resection would lead to seizure freedom. In case the model predicts a poor outcome, then this is an indication that more pre-surgical scans are needed, or that an alternative resection strategy is desired. Information from our model can be used to suggest such (optimal) alternative resection strategies.

In a new article in Network Neuroscience, we build upon the previously published model for which the parameters were optimized by comparing the modelled patterns seizure propagation to the patterns observed in invasive EEG recordings.

Here, we further improved the model by adding a recovery mechanism (Susceptible-Infected-Recovered; SIR), which allows us to more easily determine the effect of virtual resections. Additionally, the start of the seizure is now determined by the model, allowing us to generate alternative hypotheses about the area where seizures begin. We were still able to successfully test the effect of virtual resections. The model is now ready to be tested in a (pseudo-)prospective study.

Now that we are reaching the end of the first phase of research, we have made an effort to share our results with other scientists by participating in scientific conferences. In this manner we can gain feedback on our methods and projects, discuss new approaches, and share any interesting discoveries. It is a good opportunity to learn how other teams deal with similar scientific questions! During the first half of 2022 we have participated in the BIOMAT conference in Granada (Spain), the Dutch Neuroscience Meeting in Tiel (The Netherlands), the 31st Computational Neuroscience Meeting in Melbourne (Australia) and the BIOMAG conference in Birmingham (UK). You can find here the final poster we presented at BIOMAG with the details of our latest research.
Our second paper is live now in the journal Scientific Reports. Here we define individual seizure models using a simple spreading model: the Susceptible-Infected model. The modelled seizures spread throughout the brain network that was derived from a patient’s MEG, and we have shown that the model captures the fundamental aspects of clinically observed seizure propagation. We further demonstrate that our method can be used to test virtual resections in silico, in order to find optimal (smaller or alternative) resection strategies.
The Amsterdam Neuroscience Annual Meeting 2021 took place as a two-day hybrid conference (September 30 and October 1)  at the Johan Cruijff ArenA. This yearly event brings together neuroscientists and clinicians of Amsterdam UMC, VU Amsterdam and University of Amsterdam. We presented our computer model for epilepsy surgery and won one of the poster presentation prizes. Congratulations to the whole team!
In this new publication, we study how well individualized spreading models reproduce seizure propagation, and test a virtual resection model to find alternative, smaller resection strategies.
This is our first peer-reviewed publication defining and testing the spreading model for epilepsy surgery. Here we define the seizure propagation model, relate it to the network structure, and define a virtual resection optimization model to find alternative, smaller resections for each patient.