UK researchers develop AI that detects brain lesions missed by doctors

The development of artificial intelligence tools for medical diagnosis has reached a significant milestone in epilepsy treatment, with UK researchers creating an AI system that can detect previously invisible brain lesions. This breakthrough could particularly impact the 30,000 UK patients with uncontrolled epilepsy caused by subtle brain abnormalities that traditional scanning methods cannot identify.

The breakthrough explained: MELD Graph, developed by researchers at King’s College London and University College London, can detect two-thirds of epilepsy-causing brain lesions that doctors typically miss on traditional scans.

• The AI tool analyzes magnetic resonance imaging (MRI) scans to identify focal cortical dysplasia, a common cause of medication-resistant epilepsy
• Researchers tested the system using 1,185 MRI scans from 23 hospitals worldwide, including 703 cases with confirmed brain abnormalities
• The tool processes images more quickly and in greater detail than human radiologists, potentially accelerating diagnosis and treatment

Clinical impact and limitations: While MELD Graph shows promising results in identifying subtle brain abnormalities, the technology still requires human oversight and has some limitations in detection capability.

• The tool successfully identified a previously undetected lesion in a 12-year-old patient who had tried nine different medications without success
• Despite its advanced capabilities, the AI still misses approximately one-third of brain lesions
• The system could reduce the need for costly tests and procedures while enabling more targeted surgical interventions

Expert perspectives: Medical professionals and advocacy organizations have expressed optimism about the tool’s potential while emphasizing the need for further research.

• Professor Helen Cross, a childhood epilepsy consultant, highlights the tool’s potential to identify removable abnormalities that could cure epilepsy
• Epilepsy Action praised the technology’s potential for faster diagnosis while noting it doesn’t address the shortage of specialist epilepsy nurses in England
• The Epilepsy Society’s Ley Sander emphasized the need for cautious implementation while acknowledging the life-changing potential for surgical candidates

Implementation roadmap: The path to clinical deployment of MELD Graph involves several key steps and considerations.

• The research team has made the tool available as open-source software for worldwide clinical research
• Additional trials are required to evaluate long-term patient benefits before official approval for diagnostic use
• The technology must demonstrate consistent reliability across different healthcare settings and patient populations

Future implications: The development of MELD Graph represents a significant step forward in combining AI with medical expertise, though questions remain about its full potential and limitations in clinical practice. Success in this application could pave the way for similar AI tools in other areas of neurological diagnosis, while highlighting the importance of maintaining human oversight in medical decision-making.