AI Tool Detects Tiny Brain Lesions in Children with Epilepsy, Improving Diagnosis and Treatment Outcomes
An advanced AI tool developed by researchers at the Murdoch Children's Research Institute (MCRI) and The Royal Children's Hospital (RCH) can detect tiny brain lesions linked to severe epilepsy in children with remarkable accuracy. The AI system, dubbed the "AI epilepsy detective," identifies focal cortical dysplasias—abnormalities the size of a blueberry—in up to 94% of cases when combined with medical imaging data. Led by Dr. Emma Macdonald-Laurs, a neurologist at RCH and MCRI, the team created the tool to address a major challenge in pediatric epilepsy: the difficulty of spotting these small, often hidden lesions on standard MRI scans. Before the AI’s intervention, human experts missed the diagnosis in 80% of cases studied. The AI was trained using MRI and FDG-PET scan data from 71 children at RCH and 23 adults at the Austin Hospital. The model’s performance was tested on a separate group of children, achieving a 94% detection rate. Of the 17 children in the test group, 12 underwent epilepsy surgery, and 11 are now seizure-free. Cortical dysplasias, which develop during fetal development, are a leading cause of drug-resistant epilepsy in children, affecting about one in 200. Seizures typically begin in early childhood and can escalate to multiple episodes daily, often requiring emergency care or hospitalization. Without timely treatment, ongoing seizures can lead to learning difficulties, behavioral issues, and even intellectual disability. Dr. Macdonald-Laurs emphasized that early detection allows for faster referrals to neurosurgery, more precise surgical planning, and better outcomes. With the AI’s help, surgeons can avoid critical brain areas responsible for speech, movement, and cognition, reducing the risk of complications and preserving healthy tissue. She noted that traditional MRI scans often fail to reveal these lesions, delaying diagnosis and blocking access to potentially curative surgery. The AI tool offers a faster, more reliable path to identifying the source of seizures. With further funding, the research team plans to expand testing across pediatric hospitals in Australia. The study, published in the journal Epilepsia, also involved collaborators from the University of Melbourne, the Florey Institute of Neuroscience and Mental Health, Harvard Medical School, and the Austin Hospital.
