AI algorithm that detects brain abnormalities may help treat epilepsy

An international team of researchers led by University College London (ucl) has developed artificial intelligence (AI) algorithms that can detect subtle brain abnormalities that cause seizures

The Multicenter Epilepsy Lesion Detection Project (MELD) used more than 1,000 patient MRI scans from 22 epilepsy centers around the world to develop an algorithm that provides reporting of abnormalities in cases of drug-resistant focal cortical dysplasia (FCD), FCD is the main cause of epilepsy

FCDs are abnormally developed regions of the brain that often lead to drug-resistant epilepsy

To develop the algorithm, the research team quantified cortical characteristics of MRI scans, such as the thickness or folding of the cerebral cortex/brain surface, and used approximately 300,000 locations in the brain

The researchers then trained the algorithm on samples labeled by radiologists, which were either healthy brains or brains with FCD—depending on their patterns and characteristics

The study, published in the journal Brain, found that overall, the algorithm was able to detect FCD in 67 percent of the cases in the cohort (538 participants)

Prior to this, 178 participants were considered MRI-negative, meaning the radiologist couldn't find an abnormality -- yet the MELD algorithm was able to identify FCD in 63 percent of these cases

This is especially important, as if a doctor could find an abnormality on a brain scan and remove it surgically to cure it

Co-first author Mathilde Ripat (UCL Great Ormond Street Institute for Child Health) said: "Our focus is to create an explainable AI algorithm that can help doctors make decisions

Co-senior author Dr Konrad Wagstyl (UCL Queen Square Institute of Neurology) added: "The algorithm could help uncover these hidden lesions in more children and adults with epilepsy and make more epilepsy patients consider brain surgery , cure epilepsy and improve their cognitive development

About 1% of the world's population suffers from epilepsy, a severe neurological disorder characterized by frequent seizures

In the UK, around 600,000 people are affected

In children with surgically controlled epilepsy, FCD is the most common cause, and in adults it is the third most common cause

In addition, FCD was the most common cause of epilepsy in patients with epilepsy whose brain abnormalities were not detected on MRI scans

Co-first author Dr Hannah Spitzer (Helmholtz Munich) said: "Our algorithm automatically learns to detect lesions from thousands of patient MRI scans

Co-author Dr Sophie Adler (UCL Great Ormond Street Institute for Child Health) added: "We hope this technology will help to identify abnormalities that are currently overlooked as causing epilepsy

This FCD detection study used the largest FCD MRI cohort to date, meaning it was able to detect all types of FCD
.

The MELD FCD classifier can be used in any patient over 3 years of age with suspected FCD who has an MRI scan
.

The MELD project is supported by the Rose Tree Trust
.