AI and brain mapping combination advances personalised epilepsy surgery

For people with drug-resistant epilepsy, surgery can offer a life-changing opportunity to control seizures. But it also raises questions about how a procedure could affect language, memory, and communication.
A new $2.32 million NIH-funded project led by the University of Florida aims to give patients, families, and clinicians more personalised information to help answer those questions before surgery.
Led by Abbas Babajani-Feremi, associate professor in the UF Department of Neurology and director of the Magnetoencephalography (MEG) Laboratory at the Norman Fixel Institute for Neurological Diseases at UF Health, the project will combine advanced brain mapping, intracranial recordings and artificial intelligence to predict how different surgical approaches could affect a patient’s language function. UF is collaborating with Cincinnati Children’s Hospital Medical Centre on this project.
Babajani-Feremi said: ‘For a patient and family, one of the most difficult questions is not only whether surgery will control the seizures, but also what they might lose because of the surgery.’
To create a more individualised picture of those networks, researchers will combine MEG, functional MRI, diffusion MRI and intracranial recordings. Each technology provides complementary information on where language processing occurs and how those regions are connected.
Machine-learning models will then integrate the brain-mapping data with patient-specific clinical information and characteristics of a proposed surgical resection. The goal is to estimate an individual patient’s risk of language changes rather than relying on population averages or a single brain-mapping technique.
A major goal of the project is to develop a virtual resection tool that could allow clinicians to compare potential surgical approaches on a patient’s own brain before an operation.
Clinicians could identify an area of brain tissue to remove or treat and overlay it onto the patient’s individualised language maps. An AI model would then estimate how the proposed approach could affect different aspects of language. Clinicians could adjust the proposed surgical area and compare different scenarios.
The researchers hope the approach could ultimately help clinical teams balance seizure control with the preservation of language and quality of life, while providing patients and families with more individualised information when making treatment decisions.
The project brings together expertise in epilepsy, neurosurgery, neuroimaging, neuropsychology and artificial intelligence. At the Fixel Institute, researchers have access to a 306-channel whole-head MEG system and an advanced 3T MRI, enabling the team to integrate multiple forms of brain imaging with intracranial recordings. UF’s AI and high-performance computing infrastructure will support developing machine-learning models and a virtual resection platform.
Babajani-Feremi said: ‘Today, much of presurgical mapping asks, “Where is language located?” I hope we can move towards the much more clinically meaningful question, “For this individual patient, what is likely to happen to language if we perform this particular surgery?” What excites me most is the possibility of changing the question we ask before epilepsy surgery.’