Michelle ThompsonAn AI tool developed by UK researchers can identify two-thirds of epilepsy-related brain lesions that doctors often miss, potentially enabling more precise surgeries to prevent seizures.
Approximately 30,000 people in the UK—one in five epilepsy patients—experience uncontrolled seizures due to brain abnormalities too subtle to be detected in standard scans.
Experts in childhood epilepsy highlight the AI tool’s significant potential in expanding treatment options. However, further research is required to assess its long-term impact before clinical approval.
Focal cortical dysplasia, a frequent cause of epilepsy resistant to medication, can lead to symptoms such as jerking, stiffness, and loss of awareness, often necessitating emergency care. Removing a small brain region can be an effective solution, but undetected lesions may delay diagnosis, treatment, and surgery.
‘Really difficult’
In a study published in JAMA Neurology, researchers from King’s College London and University College London trained their AI tool, MELD Graph, using MRI scans from 1,185 patients across 23 hospitals worldwide, including 703 individuals with brain abnormalities.
MELD Graph analyzed the images faster and in greater detail than doctors, potentially enabling quicker treatment and reducing unnecessary tests, according to lead researcher Dr. Konrad Wagstyl.
Despite its capabilities, the AI still requires human oversight and does not detect all abnormalities. “It’s like searching for a single character within five pages of solid black text,” Dr. Wagstyl explained, noting that while AI catches two-thirds of missed cases, a third remain challenging to find.
At an Italian hospital, the tool identified a subtle lesion overlooked by radiologists in a 12-year-old boy who had daily seizures despite trying nine different medications.
Dr. Konrad Wagstyl and his research team are developing the AI algorithm, MELD Graph, to enhance the detection of subtle brain abnormalities linked to epilepsy.
Study co-author and childhood epilepsy specialist Prof. Helen Cross highlighted the AI tool’s potential to quickly detect abnormalities that could be removed to potentially cure epilepsy.
She described uncontrolled epilepsy as “incapacitating,” noting that many of her patients at Great Ormond Street Hospital undergo years of seizures and testing before a lesion is identified.
Epilepsy Action called the AI tool’s capabilities “exciting” for enabling faster diagnoses but emphasized that it does not address the shortage of specialist epilepsy nurses in England.
Ley Sander from the Epilepsy Society urged caution, stating that while it is still early, the tool could be “life-changing” by identifying more patients eligible for surgery.
The researchers are seeking official approval for MELD Graph as a diagnostic tool but must conduct further trials to assess its long-term impact. Meanwhile, they have released the tool as open-source software, allowing hospitals worldwide to use it for clinical research.
