Multivariate epileptogenic abnormalities in the human brain : towards classification, prediction, and mechanistic understanding

Abstract

Epilepsy is a heterogeneous disorder characterised by unprovoked and recurrent seizures. Despite a wealth of knowledge and advancements in treatment such as anti-seizure medications and surgical intervention, there are still individuals whose seizures remain refractory. Given the impact that epilepsy can have on an individual’s life it is clear that work remains to personalise patient treatment plans in order to try and achieve seizure freedom. Developing quantitative markers of abnormality relative to healthy populations using neuroimaging and neurophysiological data has proven to be a valuable framework to complement current qualitative clinical practices. In this thesis we extend the quantitative study of abnormalities from a typical univariate framework, to a multivariate framework which can incorporate and account for the physiological and pathological heterogeneity within the brain. We developed two quantitative measures of multivariate abnormality using healthy cohorts as a baseline. Deriving patient specific measures of abnormality, we investigated whether or not our multivariate abnormalities could be leveraged in a clinical setting as (1) markers of post-operative surgical success, (2) to help guide invasive intracranial electrode placement, and (3) to develop a better understanding of epilepsy progression within each individual. Implantation of intracranial electrodes, and then the subsequent resection of the strongest multivariate functional abnormalities was associated with post-operative surgical success for individuals living with drug refractory neocortical epilepsy. Moreover, novel clustering of functional abnormalities highlighted areas of tissue that overlapped strongly with the resection in seizure free individuals only. Multivariate structural abnormalities were associated with duration of epilepsy, with the findings more robust to outliers than equivalent univariate measures of abnormality. In this thesis we highlight that quantitative multivariate abnormalities which account for the heterogeneity within the brain and individuals can complement current clinical practice and aid in the treatment of individuals living with drug refractory epilepsy. The data-driven measures proposed could be used in conjunction with others to provide personalised treatment options that are tailored to each individual’s neuroimaging and neurophysiological data.