AI Helps Detect Epilepsy Before Doctors

Accordingly, Meld Graph software, developed by scientists at King's College London and University College London (UCL) in the UK, is an advanced diagnostic support tool to detect brain abnormalities that radiologists may miss.

The technology is designed to identify focal cortical dysplasia (FCD), a malformation that occurs when nerve cells fail to develop normally in certain areas of the brain.

This is one of the common causes of drug-resistant epilepsy, that is, cases that do not respond to treatment with conventional antiepileptic drugs.

Statistics show that there are currently around 630,000 people with epilepsy in the UK, of which one in five have epilepsy caused by brain damage.

Remarkably, many of these cases can be effectively treated by surgical removal of the affected brain area. However, detecting FCD is not simple, because the lesions are often very subtle and easily overlooked in conventional imaging tests.

"Radiologists have to process large amounts of medical data, including very small and difficult-to-detect lesions. Using an AI-powered tool like Meld Graph can help them make faster and more accurate decisions," said Dr Konrad Wagstyl from King's College London.

“This not only improves the efficiency of the healthcare system, but also shortens treatment times, helping patients avoid unnecessary and costly tests or procedures,” said Dr. Konrad Wagstyl.

How Meld Graph Works

In the study published in the prestigious American medical journal JAMA Neurology, the team used magnetic resonance imaging (MRI) data from 1,185 patients, including 703 patients with FCD, collected from 23 epilepsy treatment centers worldwide. Half of the data came from children, a group of subjects particularly vulnerable to epilepsy.

Meld Graph was trained by analyzing brain images, focusing on detecting small abnormalities that might not be picked up by the naked eye or traditional methods.

Results showed that this AI tool can detect 64% of brain malformations, significantly improving the early diagnosis rate for epilepsy patients.

The research team believes that, if applied in clinical practice, Meld Graph could revolutionize the diagnosis and treatment of partial epilepsy, helping more patients have the opportunity for more timely and effective treatment.

Dr. Luca Palma from Bambino Gesu Children's Hospital (Italy), co-author of the study, shared a specific case: "Meld Graph discovered an extremely subtle lesion that many radiologists had previously missed in a 12-year-old boy. This boy had daily seizures and had tried up to 9 different anti-epileptic drugs without improvement."

"Thanks to this tool, we can precisely locate the lesion, help make surgical decisions, minimize risks, save costs and optimize treatment outcomes," added Dr. Luca Palma.

Next steps

Although Meld Graph is not yet widely used in clinical practice, the team has released the software as open source and organized training workshops to help clinicians access and apply the technology in practice.

“One of the most memorable moments for me was hearing from doctors from all over the world, from the UK, Chile, India to France, how they had used our tool to help their patients,” said Dr Mathilde Ripart from UCL.

In the UK, Great Ormond Street Children's Hospital is one of the pioneering facilities participating in the Meld Graph trial.

Professor Helen Cross, consultant epilepsy physician at the hospital and director of the UCL Great Ormond Street Institute of Child Health, stressed: "Many of the children I treat have had years of seizures and countless tests before we find damage to their brains."

Professor Helen Cross added: "The epilepsy medical community is working hard to speed up diagnosis and treatment. Innovative technologies like Meld Graph can help quickly identify surgically treatable abnormalities, even giving patients the chance of a complete cure."

With its enormous potential, Meld Graph could become a groundbreaking tool in the diagnosis and treatment of epilepsy, opening up new hope for millions of patients worldwide.