Revealing the dynamic mechanisms of seizures: An integrated mathematical and clinical approach

The complexity of the brain is reflected in both its repertoire of 'normal' behaviour, as well as in the pathologies underlying neurological disorders. Finding appropriate treatment for disorders such as Parkinson's disease, Alzheimer's disease and epilepsy is an on-going, challenging task for scientific, clinical and medical communities. To address the level of complexity in these diseases, scientists collaborate by integrating computational, experimental and clinical neuroscience. Due to these collaborations, it is becoming clearer that several neurological disorders are characterised by rhythmic oscillations. A good example of this can be found in the field of epilepsy, where very slow as well as very fast oscillations are associated with the generation of seizures, and daily hormonal rhythms have been shown to increase the number of seizures' incidents.

Clinical diagnosis of epilepsy relies on a detailed case history of the seizures a subject encountered, the use of electroencephalogram (EEG) monitoring (typically 90 minutes or more) and the observation of seizures in the clinic. However, conventional EEG has low sensitivity and specificity, and people may have non-epileptic seizures or mannerisms similar to the symptoms typically observed during seizures, but that are not caused by the synchronous neuronal activity underlying epileptic seizures. Epilepsy is challenging to diagnose and misdiagnosis of epilepsy is a serious issue, associated with several negative personal and socio-economic consequences for a subject, as well as significant costs at the national level. Furthermore, of all cases with epilepsy, on average 70% will eventually be able to successfully control their seizures by using appropriate anti-epileptic drugs (AEDs). In the other 30% of the cases ('refractory epilepsy') there is no combination of AEDs that is effective.

Mathematical models provide a powerful tool with which to identify and understand biological mechanisms that can lead to the generation, propagation, and termination of seizures. Mathematical models integrating experimental and clinical detail at diverse scales of activity have revealed the importance of many microscopic and macroscopic mechanisms in the generation of seizure-like activity, ranging from genetic and molecular mechanisms to changes in the excitability of neural populations leading to the generation of pathological oscillations. Despite this, no recent treatment has made a significant impact on the success rate of AEDs in treating epilepsy.

We have recently demonstrated through analysis of routine clinical recordings provided by our collaborators at King's College London that there are differences between brain networks of people with idiopathic generalised epilepsies (IGE), their unaffected first-degree relatives, and healthy controls. Identifying features characteristic of people with epilepsy is particularly exciting because it will start building towards a deeper understanding of how pathological rhythms originate in brain networks of people with epilepsy, but not their relatives, controls, or people with non-epileptic seizures. This is particularly exciting because identification of biomarkers - features characteristic of people with epilepsy or high seizure susceptibility - is potentially beneficial for the diagnosis of epilepsy, as well as with respect to predicting whether people with epilepsy benefit from specific AEDs.

Epileptic seizures are characterised by periods of abnormal synchronous activity distributed over large areas of the brain. This has lead to the current understanding of epilepsy as a dynamic disorder of brain networks, where patients have an enduring susceptibility to seizures, which occur unpredictably from otherwise healthy brain function. By understanding how seizures emerge from healthy background activity from clinical recordings, we can start to understand how networks of people with epilepsy differ from the networks of people with non-epileptic seizures (e.g. syncope), and how AEDs might alter the network structures to decrease the seizure-likelihood. This will put us in a better position to support clinicians in making decisions with respect to diagnosis and prognosis of people with epilepsy.

I have used a combination of mathematical modelling and data-analysis to show that connectivity structures within resting-state functional brain networks of people with idiopathic generalised epilepsy (IGE) and their first-degree relative have altered properties to those of healthy controls. This implies that the underlying mechanisms responsible for seizures are reflected in resting state EEG, but that additional mechanisms play a role as well because the abnormal features are characteristic of both people with IGE as well as their unaffected relatives.

In order to identify a sufficient set of dynamic biomarkers characteristic of people with epilepsy, I will develop a model framework that combines connectivity structures inferred from resting state EEG with mathematical descriptions of the dynamics of individual brain regions (e.g. neural mass). By using a novel parameter estimation technique, I will then find optimal model frameworks that allows me to infer a set of dynamic biomarkers generating the best discrimination between people with epilepsy and people without epilepsy, as well as seizure-free people and people with ongoing seizures using follow-up data.

This research will provide benefits to various groups of people, potentially leading to significant societal impact. In particular people with epilepsy and their relatives would benefit from a change in the way we diagnose epilepsy: moving from clinic to GP surgery or even their home. Further, more accurate diagnosis will benefit people who are currently diagnosed with epilepsy, but do not in fact have the condition. Neurologists, clinical neurophysiologists and epileptologists will benefit from our research by receiving evidence-based support to the process of diagnosis and the identification of appropriate clinical interventions.

By understanding the underlying mechanisms specific to the individual case of epilepsy and seizures, the methods described in this proposal will optimise treatments (e.g. clinicians prescribing ineffective AEDs considering surgical resection if our model-driven framework shows these interventions would have limited effect). By increasing the accuracy of these interventions, there may be a decrease in negative side effects. These effects have the potential to generate significant health benefits to people with epilepsy, as the process of diagnosis could be significantly sped up, as well as preventing unnecessary prescription of AEDs. This will lead to significant cost savings to private and public healthcare providers. At present, the costs of managing and treating epilepsy in Europe are believed to exceed 15.5B euros per year, of which well over £30M per annum is due to misdiagnosis in the UK alone.

Moreover, the advances of this research could be used as an exemplar of how the general approach of using state-of the art computational research within the framework of clinical and biomedical research can support the improvement of diagnosis and treatment of neurological disorders. Further, demonstrating how our model-driven framework enhanced the process of diagnosing and treating people with epilepsy could potentially raise the profile of this collaborative approach in the larger discipline of systems and translational medicine, thereby promoting collaborations between computational modellers and clinicians rooted in concrete biomedical and clinical issues.