Predicting surgical outcomes and seizure mechanisms in epilepsy using machine learning and computational modelling

Epilepsy is a serious neurological disorder in which patients experience repeated unprovoked seizures. It effects approximately 70 million people worldwide. Surgery is the last resort for many patients who do not respond to medical treatments. In a surgical procedure the brain tissues that are thought to be causing seizures are resected. Unfortunately, in around 40-60% cases surgery is not successful in rendering a patient seizure free.

In this talk, I will discuss how interdisciplinary computational techniques are providing new insights on the mechanisms underlying pathological brain function in epilepsy. I will describe mathematical models that simulate hallmark pathological activity featuring seizures. These models when informed with patient-specific data enables performing virtual resections, thereby, predicting surgical outcomes. I will illustrate some of our results upon retrospective application of these techniques on epileptic patients who underwent surgery.

I will conclude by demonstrating that computational techniques, such as modelling and development of associated mathematical theories offer a powerful approach to investigate brain function and dysfunction. These techniques are critical to investigate complex, multi-modal data we routinely acquire from the brain. Such a framework can potentially lead to the formulation and validation of new hypotheses, prediction of outcomes to a therapy, and novel patient-specific treatment strategies.