Scientists have created an AI-powered tool named MELD Graph that detects 64% of brain abnormalities associated with epilepsy, a condition that human radiologists frequently overlook. According to researchers, this tool has the potential to revolutionize care for approximately 30,000 patients in the UK and 4 million individuals worldwide who suffer from one specific cause of epilepsy.
A study published today in JAMA Neurology by a team at King’s College London and University College London (UCL) highlights how MELD Graph significantly enhances the detection of focal cortical dysplasia (FCDs), which is a leading cause of epilepsy. FCDs are common structural causes of epilepsy, and in people with this type, seizures often cannot be controlled by medications alone. Surgery to remove these lesions can effectively stop the seizures.
However, detecting FCDs presents challenges as they can be subtle and difficult for radiologists to identify with the naked eye. Up to half of these abnormalities may go undetected by human reviewers. Delays in diagnosis and surgery lead to more frequent seizures, increased visits to accident and emergency departments (A&E), and greater disruption in patients’ daily lives.
The study involved pooling MRI data from 1185 participants—703 with FCDs and 482 controls—from 23 epilepsy centers around the world. Half of the dataset consisted of children’s scans. The researchers then trained MELD Graph on these images to detect subtle brain abnormalities that might otherwise remain unnoticed.
Radiologists are often overwhelmed by a large volume of imaging data they must review, making their work both time-consuming and challenging. Utilizing an AI tool like MELD Graph can support radiologists in their decision-making processes, thereby enhancing efficiency within the NHS, speeding up treatment times for patients, and reducing unnecessary tests and procedures.
Dr. Konrad Wagstyl from King’s College London highlighted a specific case where MELD Graph identified a subtle lesion missed by many radiologists in a 12-year-old boy who experienced daily seizures despite trying nine different anti-seizure medications with no success. This tool could help identify patients suitable for surgical intervention, facilitating more precise planning and improving outcomes.
Although the AI-tool is not yet clinically available, researchers have made it open-source software. They are organizing workshops to train clinicians and researchers worldwide on how to use MELD Graph effectively. These institutions include Great Ormond Street Hospital and the Cleveland Clinic.
The research team’s first author, Dr. Mathilde Ripart from UCL, emphasized hearing from doctors around the world who have already utilized their tools in helping their patients.
Professor Helen Cross, Prince of Wales’s Chair of Childhood Epilepsy and President of the International League Against Epilepsy, also expressed optimism about MELD’s potential. She noted that many children she encounters with epilepsy suffer years of seizures before a lesion is found. Initiatives like MELD aim to rapidly identify abnormalities for surgical removal and potentially cure epilepsy.
Co-lead Dr. Sophie Adler from UCL acknowledged the importance of international collaboration in this research, which involved 75 researchers and clinicians working towards “no missed epilepsy lesions worldwide.”
This innovative AI-powered tool demonstrates a significant leap forward in diagnosing FCDs and could greatly improve patient outcomes by reducing delays in diagnosis and treatment. As it moves closer to clinical application, MELD Graph promises to have substantial implications for managing epilepsy care globally.
