Epilepsy surgery is an invasive procedure that removes the part of the brain thought to cause seizures. This procedure fails to completely stop seizures in many patients, in part due to the difficulty of accurately localising the problematic neural tissue. In our pilot study and review in 2014, we suggested approaches for improving localisation and predicting surgical outcome using new data and analysis methods. Since then, we have shown that interictal functional networks combined with computational models can predict surgical outcome and suggest alternative surgery locations. Another study used diffusion imaging data with a machine learning model to find white matter connections that best predict surgical outcome. 

It is well-documented that epileptic seizures show variability in their symptoms, duration, and severity in most patients with focal epilepsy. We have recently reported that brain activity patterns also change from seizure to seizure. Intriguingly, this variability is not random, but can be explained by modulations on circadian and other timescales. We have since also confirmed these findings using chronic recordings over months and years.

The implication is that the abnormal seizure brain activity and its ensuing symptoms and severity is modulated by (slow) time-varying processes. Our future research will elucidate these modulators in more detail and explore their therapeutic utility.

One of the most striking visual features of the human brain is that its surface is folded. This folding appears different in certain diseases, and there are also stark changes during healthy ageing.

Our goal is to understand and characterise cortical folding in a comprehensive manner in humans. Together with our collaborator Prof. Bruno Mota, we have shown that the folding pattern follows a strict scaling law in humans. Most recently, we further demonstrated that this scaling law is even obeyed by different parts of the same cortex and can be generalised to any point on the cortex. The compliance across individuals and regions indicates a universal mechanism underlying human cortical folding.

Cognitive dysfunction is common across neurological, psychiatric, and neurodegenerative disorders and decreases quality of life for patients. There is currently a limited understanding of cognitive disease profiles and their neurobiological underpinnings, hindering development of cognitive interventions. We are tackling challenges in this field by developing innovative computational approaches to better understand the neuropsychology of disease. This includes: 

• reliably modelling cognitive processes using nuanced statistical methods

• investigating disease-specific profiles of cognitive impairment

• accurately estimating structural brain abnormalities and predicting disease trajectories by leveraging normative modelling and independent components of cortical morphology

• applying multivariate approaches to assess reliable brain-cognition associations in disease

• developing translatable tools for the community to enhance research and inform clinical practice

This work bridges the gap between computational and neuropsychological science and has transdiagnostic applications. For more information, please see the published work of Dr Beth Little. 

Individuals with epilepsy often have cognitive and behavioural comorbidities attributed to co-occurring neurodevelopmental conditions. The additional burdens of co-occurring medical conditions are sometimes overlooked or misdiagnosed resulting in suboptimal patient care and poorer outcomes later in life. There remains a knowledge gap in the underlying neurobiology of co-occurrence. We are developing methodologies and approaches to investigate co-occurring conditions with epilepsy. We aim to further our understanding of co-occurring conditions and build translatable tools to benefit and improve patient care. This work includes:

• Identifying imaging markers for phenotypic assessment of co-occurring traits in autism and epilepsy

• Assessing paediatric health and disease by exploring functional mechanisms of neurodevelopment in early life (see published work of Dr Sonja Fenske)

Brain stimulation is a treatment strategy that is currently being developed for epilepsy patients. To aid in this development, our work has included using computational models to predict the response to single pulse stimulation and more generally finding optimal stimulation protocols for aborting seizures. We also summarised the field of computational modelling in brain stimulation in a review in 2015.

Spike and slow wave oscillations are often observed on EEG recordings from patients with generalised seizures.  By combining structural connectivity derived from diffusion weighted MRI with computational models, we have suggested potential epileptogenesis mechanisms and accounted for patient-specific spatiotemporal heterogeneity.

We have developed custom data processing pipelines to generate structural brain networks at a much higher resolution than those traditionally investigated. By applying these techniques to data from healthy controls, we have discovered structural modules exist within brain regions, which may subserve function. We have also used these techniques in the context of epilepsy surgery.

Lithium is used as a treatment in bipolar disorder, but we do not know where it acts on the brain to achieve an effect, nor do we know if lithium actually reaches all parts of the brain.

Through a collaboration with the local Lithium group, who have developed methods to image lithium in the brain in vivo, we performed some initial analysis on the lithium concentration profile in the brain and related this to disease-associated changes in brain structure.

We have developed novel mathematical models using differential equations to comprehensively describe brain electrical activity in epilepsy. We have further extended these models to incorporate spatial aspects of pathological dynamics.  Using bifurcation theory, we have also investigated transient dynamics following perturbation. 

We further applied the principles from dynamical systems to epileptic seizures, which are pathological brain dynamics that evolve in space and time. To understand these events, we considered ways in which seizures can arise and categorised them according to their dynamic mechanisms. We have also hypothesised how the EEG waveform at seizure onset relates to these mechanisms.

In addition to applying existing data analysis and computational modelling techniques to new research questions, we also actively participate in developing new methods, advancing existing algorithms, and improving interpretability of analysis techniques.

One example of our recent work is developing multifractal measures in EEG with collaborators from University College London.

Traumatic brain injury is heterogeneous - affecting different patients in different ways. We have applied high dimensional statistical approaches to identify how different a patient is from controls in high dimensional 'connectivity space'. We have also used machine learning techniques to identify features that explain cognitive recovery after brain injury.