AI Detects Tiny Brain Lesions to Cure Pediatric Epilepsy
- 1 oct
- 2 Min. de lectura
A recent study has revealed that an advanced Artificial Intelligence (AI) tool, dubbed the "AI epilepsy detective," can detect minuscule brain lesions that cause severe epilepsy in children, promising faster diagnosis, more precise treatment, and a potential cure.
The development, led by a team at the Murdoch Children's Research Institute (MCRI) and The Royal Children's Hospital (RCH) in Australia, focuses on the detection of focal cortical dysplasias (FCDs). FCDs are malformations that develop while the baby is in the womb and are a common cause of drug-resistant seizures.
Epilepsy affects approximately one in 200 children, and FCDs are among the causes of seizures that typically start during preschool or early school years. However, these lesions are hard to spot on routine Magnetic Resonance Imaging (MRI) scans, often being impossible to identify using traditional techniques. Before the AI detector was used, the study found that 80 percent of patients had their diagnosis missed during the human examination of their MRI results.
Dr. Emma Macdonald-Laurs, a RCH pediatric neurologist who led the team, explained that failure to locate the abnormal tissue slows down the pathway to a definitive diagnosis and can stop a child from being referred for potentially curative epilepsy surgery. Prolonged, uncontrolled seizures increase the likelihood of the child developing learning difficulties, including intellectual disability.
The new AI tool was trained using information gathered from both MRI and FDG-positron emission tomography (FDG-PET scans). The detector proved capable of finding lesions the size of a blueberry or smaller.
In the test cohort, the medical imaging-assisted tool recorded its best result with a success rate of 94 percent in lesion detection. Dr. Macdonald-Laurs noted that while the tool acts as a "detective" that helps put the puzzle pieces together, it does not replace radiologists or epilepsy doctors.
The results have been impressive. Of 17 children in the test group, 12 underwent surgery to remove their lesions, and 11 of them are now seizure-free. This more accurate diagnosis leads to faster referrals for surgery. It allows neurosurgeons to develop a safer surgical roadmap, avoiding important blood vessels and brain regions that control speech, thought, and movement, thereby removing less healthy brain tissue. The family of Royal, a 5-year-old child who previously suffered multiple seizures daily, testified that, following AI-assisted surgery, he is now seizure-free.
Although the use of PET scans involves higher costs, is not as widely available as MRI, and is associated with a dose of radiation, the researchers are excited about the results. They plan to test the detector in additional pediatric hospitals across Australia with further funding.
🔖 Sources
