AI algorithm that detects brain abnormalities could help cure epilepsy

An artificial intelligence (AI) algorithm capable of detecting subtle brain abnormalities that cause epileptic seizures has been developed by a team of international researchers led by the ‘UCL.

The Multicentric Epilepsy Lesion Detection (MELD) Project used more than 1 000 MRI scans of clients from 30 World Epilepsy Centers to develop the algorithm, which provides reports of the location of abnormalities in cases of drug-resistant focal cortical dysplasia (FCD) – a major cause of epilepsy.

FCD are areas of the brain that have developed abnormally and often cause drug-resistant epilepsy. It is usually treated with surgery, but identifying lesions from an MRI is a long-term challenge for clinicians, as MRIs in FCDs may appear normal.

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

The researchers then trained the algorithm on examples labeled by expert radiologists as either healthy brain or FCD, based on their patterns and characteristics.

The results, published in Brain, found that overall the algorithm was able to detect FCD in 67% of cases in the cohort ( 538 contributors).

Formerly, 178 of contributors were considered MRI negative, meaning that the radiologists had not been able to find the abnormality, but the MELD algorithm was able to identify the FCD in 63% of these cases.

This is particularly important because if doctors can find the abnormality in the brain scan, surgery to remove it may provide a cure.

Co-first author Mathilde Ripart (UCL Good Ormond Street Institute of Child Health and fitness) said, “We focused on creating an AI algorithm that was interpretable and could help doctors make decisions. Showing doctors how the MELD algorithm made its predictions was a critical part of this process.

Co-lead author Dr Konrad Wagstyl (UCL Queen Sq. Institute of Neurology) added: “This algorithm could help find more of these hidden lesions in children and adults with epilepsy, and enable more epileptic patients to be considered for brain surgery that could cure epilepsy and improve their cognitive development. Around 440 children a year could benefit from epilepsy surgery in England.

Around 1% of the world’s population suffers from epilepsy, a serious neurological condition characterized by frequent seizures.

In the United Kingdom, some 600 000 people are affected. Although drug treatments are available for the majority of people with epilepsy, 20 to 63 % not responding to medication.

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

Additionally, among clients with epilepsy who present with an abnormality in the brain that cannot be found on MRI scans, FCD is the most common finding.

Co-first author Dr Hannah Spitzer (Helmholtz Munich) said: “Our algorithm automatically learns to detect lesions from thousands of client MRI scans. It can reliably detect lesions of different kinds, shapes and sizes, and even many of these lesions. that radiologists previously lacked.”

Co-lead author Dr Sophie Adler (UCL Excellent Ormond Street Institute of Kid Health) added: “We hope this technology will help to identify abnormalities causing epilepsy that are currently overlooked. Ultimately, this could allow more people with epilepsy to have potentially healing brain surgery.

This FCD detection study uses the largest cohort MRI of FCD to date, which means it is able to detect all types of FCD.

The MELD FCD classification tool can be run on any affected person suspected of having FCD over 3 years old and having had an MRI.

The MELD project is supported by the Rosetrees Have faith in.

Limitations of the study

Different MRI scanners were used in the 22 hospitals involved in the study around the world, which allows the algorithm to be more robust but could also affect the sensitivity and specificity of the algorithm.

Related Articles

Back to top button