AI-Assisted EEG Interpretation Is Here to Stay
When asked about the long-term future of AI-assisted EEG interpretation, Dr. Friedman predicted that it will be “critical” for helping improve the accuracy of epilepsy diagnoses, particularly because most EEGs worldwide are interpreted by non-experts, leading to the known issue with epilepsy misdiagnosis.
“However,” he added, “it is important to note that epilepsy is a clinical diagnosis ... [EEG] is only one piece of evidence in neurologic decision making. History and accurate eyewitness description of the events of concern are extremely critical to the diagnosis and cannot be replaced by AI yet.”
Dr. Kerr offered a similar view, highlighting the potential for SCORE-AI to raise the game of non-epileptologists.
“My anticipation is that neurologists who don’t use SCORE-AI will be replaced by neurologists who use SCORE-AI well,” he said. “Neurologists who use it well will be able to read more EEGs in less time without sacrificing quality. This will allow the neurologist to spend more time talking with the patient about the interpretation of the tests and how that impacts clinical care.”
Then again, that time spent talking with the patient may also one day be delegated to a machine.
“It is certainly imaginable that AI chatbots using large language models to interact with patients and family could be developed to extract consistent epilepsy histories for diagnostic support,” Dr. Wesley said.
This work was supported by a project grant from the Canadian Institutes of Health Research and Duke Neurology start-up funding. The investigators and interviewees reported no relevant conflicts of interest.