AI tool detects epilepsy-triggering brain lesions invisible to doctors—offering hope for surgery.

A groundbreaking artificial intelligence tool is revolutionising epilepsy diagnosis, detecting two-thirds of brain lesions that doctors frequently miss. Developed by researchers at King’s College London and University College London, this AI-driven breakthrough could lead to earlier, more targeted surgeries to stop uncontrollable seizures.

Currently, one in five epilepsy patients—around 30,000 in the UK—suffers from seizures caused by brain abnormalities too subtle for doctors to see on MRI scans. Known as focal cortical dysplasia, these abnormalities often leave patients struggling with uncontrolled seizures despite multiple medications. But now, MELD Graph, an AI-powered tool, is changing the game.

In a study published in JAMA Neurology, researchers fed MRI scans from 1,185 patients across 23 hospitals worldwide into MELD Graph. Among 703 patients with brain abnormalities, the AI identified subtle lesions faster and with greater accuracy than doctors—potentially paving the way for quicker treatment and fewer expensive tests.

Dr Konrad Wagstyl, the study’s lead researcher, described the challenge as trying to find a single character on five pages of solid black text. While MELD Graph successfully detected two-thirds of missed lesions, some still remained difficult to identify, highlighting the need for human oversight alongside AI.

At an Italian hospital, the AI uncovered a previously invisible lesion in a 12-year-old boy who had suffered daily seizures despite trying nine different medications. This breakthrough meant doctors could finally consider surgery as a potential cure.

A New Era for Epilepsy Treatment?

Professor Helen Cross, a consultant at Great Ormond Street Hospital and study co-author, called the AI’s potential “huge”, as it could quickly pinpoint abnormalities that can be removed to stop seizures. Many epilepsy patients endure years of suffering before a lesion is found, but this tool could significantly shorten that wait.

While the AI is not yet approved for clinical use, researchers have made it available as open-source software for hospitals to use in further studies.

However, epilepsy charities caution that while AI might speed up diagnoses, the NHS still faces a shortage of specialist epilepsy nurses—a crucial part of long-term patient care.

Despite this, experts believe MELD Graph could be life-changing for thousands, offering a clearer path to treatment and, for some, a potential cure.