Australian researchers announced on Tuesday that an artificial intelligence tool capable of detecting subtle brain abnormalities that are difficult to detect in children with epilepsy may help patients undergo potentially life-changing surgery more quickly.
According to experts, epilepsy has multiple causes, and about three out of every ten cases are attributed to structural abnormalities in the brain. But MRI scans often miss these abnormalities, especially the smallest lesions that may be hidden deep in the crease of the brain.
A team led by pediatric neurologist at the Royal Melbourne Children’s Hospital Emma Macdonald-Lowers trained an artificial intelligence tool on brain images of children to detect lesions the size of blueberries, or even smaller.
“These lesions often go undetected, and many children are not considered to need surgery,” MacDonald-Lors noted during a press conference before the study was published in the journal “Epilipia.”
She explained, “This tool does not replace the need for radiologists or those who specialize in epilepsy. Rather, it is more like an investigator that helps put together the parts of the disassembled picture more quickly, which allows suggesting surgery that may change the patient’s life.”
Of the patients in the study with cortical dysplasia and focal epilepsy, 80 percent had previously undergone an MRI scan with normal results.
When the researchers used the AI tool to analyze both MRI scans and another type of medical test, positron emission tomography, its success rate was 94 percent for one test group and 91 percent for the other.
Of the 17 children in the first group, 12 underwent surgery to remove brain lesions, and 11 were relieved of their seizures, according to MacDonald-Lurz’s team at the Murdoch Children’s Research Institute.
She noted, “The next step is to test this detection tool in a more realistic hospital environment on new patients who have not previously been diagnosed.”
Epilepsy, which causes recurrent seizures, affects about one in 200 children, and medications are ineffective in about a third of patients.
