Researchers in Exeter are using mathematics to identify the areas of the brain that contribute most to epileptic seizures. Their approach could significantly improve the success rate of epilepsy surgery.

A bit of background

People who are about to have epilepsy surgery often require an invasive procedure called an intracranial EEG (iEEG), which tries to identify where in the brain seizures are coming from. During the iEEG, electrodes are placed onto and sometimes into the brain and the electrical activity is recorded. By studying this electrical activity, clinicians aim to identify seizure-generating tissue, and if possible remove it, in the hope that seizures will stop.The success rate of epilepsy surgery is currently about 50-60%, which is not nearly as high as it could be. The group in Exeter is trying to address this.

Computer modelling of the brain

In earlier work, the team used existing data from ‘real’ iEEG recordings and imaging scans to develop a complex computer model of a functioning brain. The model contains ‘nodes’, which correspond to the regions of the brain underneath the iEEG recording electrodes. Connections between these nodes (which correspond to neural networks in the brain) were mapped by studying how the recorded electrical activity from each electrode related to that of the other electrodes. This resulted in a brain network model. In the model, individual nodes have different numbers of connections to other nodes, and they don’t all connect to each other.

What the researchers did and what they found

For this study, the researchers used a more simplified system and managed to simulate the development of focal seizures in different types of brain network model. They then used mathematical methods to identify which nodes would need to be removed for the seizures to stop in their simulations. The team found that in networks containing nodes with lots of connections, and which connected preferentially to each other (rich-club networks), there were specific nodes that were critical for generating the seizures. These needed to be removed for the simulated seizures to stop. This was the same for another type of network with different numbers of connections and connection patterns (scale-free networks). In a third type of network (small-world network), the scientists didn’t identify any ‘crucial’ nodes for generating seizures. If this mathematical approach can be applied real iEEG data (collected from people before epilepsy surgery), it might help surgeons plan their surgical strategy. For people whose seizures develop in rich-club- or scale-free-type networks, the approach could pinpoint brain tissue that needed to be removed for seizures to stop. For people whose seizures begin in small-world-type networks, simply removing sufficient brain tissue might be effective.

Testing the method on real data

The researchers started to test these theories using real iEEG recordings from 16 people who had already had epilepsy surgery. When the mathematical approach was applied to the data, rich-club-type networks were revealed. As the team predicted, subjects had better long-term seizure outcomes if a greater proportion of the ‘rich club’ (tissue with especially high numbers of neural connections) had been removed.

What happens now?

The researchers aim to validate their method further using data from many more surgical patients. They will also explore whether the approach can be improved by integrating information from additional brain imaging techniques. If successful, this could significantly improve the success rate of epilepsy surgery and reduce surgery-associated side effects. Senior Author, Professor John Terry, is currently being funded by ERUK to explore hidden information within EEG recordings to help improve epilepsy diagnosis and treatment.

Click here to read more about this project.

This work was published in the journal PLOS Computational Biology.