For patients with focal epilepsy that can't be controlled with medication, surgery can provide a potential cure. To decide whether surgery is feasible, it is vital for neurosurgeons to firstly localise language function in the brain. Functional magnetic resonance imaging (fMRI) is a non-invasive method of mapping language representation. This innovative project led by Dr Reutens aimed to develop a method for language mapping that uses a passive naturalistic approach in patients with this condition.

Refractory focal epilepsy means that the seizure activity starts in a smaller part of the brain and may spread to a larger area. People affected by it do not respond to anti-epilepsy medication. Doctors can then investigate whether a potentially curative surgery is feasible, and if it is, plan the extension of resection. To do so, patients need to go through a fMRI. Current methods require the patient to perform various language-related tasks actively during the fMRI scan and rely heavily on patient performance. The task can be even more challenging for children, people with disabilities and the elderly.

Dr Reutens’ team trained a machine learning (ML) algorithm to learn the activation pattern in language areas using a naturalistic approach. That means the stimuli mimic real-world experiences that involve cognitive and emotional experiences common in daily life. By using movie watching as the naturalistic stimuli, the team has trained the algorithm to recognise the differences in the delivery of blood to the brain in people with refractory focal epilepsy. Analysis and continuous development of the algorithm are ongoing. The final outcomes will ultimately benefit the state-wide Comprehensive Epilepsy Surgery Program at RBWH and facilitate decisions about the feasibility of surgery for the patient.