AI algorithm that detects brain defects could help cure hard-to-spot epilepsy, study suggests

Scientists have developed an AI that detect early signs of epilepsy that even experienced medics could miss.

It was able to detect six in 10 patients with a rare form of the condition, compared to an MRI which did not spot any.

Epilepsy, which affects one in 100 Britons and Americans, causes uncontrolled bursts of electrical activity in the brain that trigger seizures.

One cause of epilepsy is drug-resistant focal cortical dysplasia (FCD), a subtle abnormality in the brain that causes signals to misfire.

It can only be treated with surgery. But the changes it sets off in the brain are so subtle even experienced radiologists can miss them when examining MRI scans.

But a University College London team developed an algorithm that spotted 63 per cent of FCD cases that medics had previously been unable to find.

The researchers said their model could open up brain surgery to more people with epilepsy and offer them a chance of a cure.  

Around 600,000 people in the UK suffer epileptic seizures. But between 20 and 30 per cent don’t respond to drugs.

Brain cells, or ‘neurons’ normally form into organised layers of cells to form the brain ‘cortex’.

In FCD, these cells are disorgainsed, leading to much higher risk of seizures.

In children who have had surgery to control their epilepsy, FCD is the most common cause. In adults who require surgery, FCD is the third most common cause. 

However, medics struggle to identify FCD from MRI scans because they can appear normal. 

In the latest study, researchers gathered more than 1,000 MRI scans from 22 global epilepsy studies.

A team of expert radiologists labelled the scans as either healthy or having FCD. It then trained the algorithm to spot scans with abnormalities.

This involved inputting information on 300,000 locations in the brain. 

The findings, published in the scientific journal Brain, show that the algorithm spotted FCD in 538 of the scans.

This included 112 cases that radiologists had been unable to spot.

The researchers said this was ‘particularly important’ because those with FCD rely on it being detected, so they can undergo surgery to cure it.

Dr Konrad Wagstyl, a researcher at UCL’s Queen Square Institute of Neurology, said: ‘This algorithm could help to find more of these hidden lesions in children and adults with epilepsy, and enable more patients with epilepsy to be considered for brain surgery that could cure the epilepsy and improve their cognitive development. 

‘Roughly 440 children per year could benefit from epilepsy surgery in England.’ 

The algorithm can be run on any patient with a suspicion of having an FCD who is aged three and over and have had an MRI scan.

Co-first author, Dr Hannah Spitzer, a machine learning researcher at research institute Helmholtz Munich, said: ‘Our algorithm automatically learns to detect lesions from thousands of MRI scans of patients. 

‘It can reliably detect lesions of different types, shapes and sizes, and even many of those lesions that were previously missed by radiologists.’

Dr Sophie Adler, a researcher at UCL Great Ormond Street Institute of Child Health, added: ‘We hope that this technology will help to identify epilepsy-causing abnormalities that are currently being missed. 

‘Ultimately it could enable more people with epilepsy to have potentially curative brain surgery.’