Everyone says that Artificial Intelligence, or AI is here to stay. The accepted credence is that AI won’t replace humans but people who know how to effectively use it will replace those who don’t. Perhaps because I’m a middle-aged parent to two teenagers and a professional writer, AI inspires at least as much anxiety as excitement in me.
Of course, AI is impacting the medical world, as well, including epilepsy treatment. Based on what I’ve read, this is a sphere where AI developments have potential to be more inspirational than nerve wracking. For example, scientists in Scotland are developing a headset that uses AI to detect oncoming seizures by analyzing brainwaves and heart rhythm. This innovative device could warn people when a seizure is coming, providing them with the opportunity to take steps to stop it (e.g. a rescue medication), or at least get to a safe place. Prior to successfully treating my drug-resistant seizures with brain surgeries, I remember dozens of situations where such a warning would have been helpful, like the day I burst out moaning mid-seizure during a staff meeting, the room full of my colleagues. Had I known the seizure was coming, I could have discreetly excused myself first. (That would have made this scene from my book, Seizing Control less stirring.) Having managed drug-resistant epilepsy, or DRE for years, I know this headset is a truly exciting example of how AI can improve life-quality and health outcomes for those of us who have it. Alas, the headset research is in its infancy phase in Europe, so it likely won’t be available in America for many years.
Nevertheless, AI use by medical practitioners is on a steady upswing. A 2025 American Medical Association survey found that 66% of doctors are using AI for a variety of tasks including drafting medical charts or visit notes, creating discharge instructions, care plans or progress notes, translation services, and assistive diagnosis. The 66% rate marks a 78% increase from the previous year, but this substantial leap is not being matched by appropriate training for practitioners. Medical experts fear that the lack of guidance is mitigating AI’s positive impacts on healthcare.
According to Dr. Kimberly Lomis, vice president of the American Medical Association’s, or AMA’s ChangeMedEd Initiative, residents and fellows “know how to use specific [AI]applications, but they haven’t had the fundamentals, in many cases,” comparing it to someone driving around a parking lot for the first time. “You haven’t fully learned how to drive. You’ve learned how to operate that machine, but you still need the rules of the road and the risks, such as the dangers of texting and driving, or drinking and driving.” A key takeaway: medical educators need to provide basic guidance on AI-best practices to their students. Without this training, medical students can’t figuratively learn how to drive. A quick online search found some movement on this front: Harvard Medical School is offering a three-day workshop this June on AI in Clinical Medicine.
While that AI-powered headset is probably a distant pipe dream for now, the Autonomous Registry and Analytics, or AURA platform is currently being piloted at West Virginia University’s Rockefeller Neuroscience Institute. The AURA leverages generative AI agents to analyze epilepsy patient records, including doctors’ clinical notes, imaging test results, and assessments. The AURA also tracks incomplete or missing records, for example whether a patient needs an additional imaging test to determine if they’re a good brain surgery candidate. AURA can summarize its findings prior to scheduled neurology appointments, providing practitioners with useful details beforehand, and assessing whether the patient has DRE. The Rockefeller Institute’s study found that generative AI–powered registries like AURA may improve epilepsy care by gathering actionable insights, reducing clinical burden, and enhancing pre-visit readiness.
As a DRE patient, I have experienced how the AURA could potentially make a neurology appointment more beneficial. When I had my first brain surgery in 2014, my neurosurgeon scheduled a last-minute MRI a few days before because she wanted a clearer image. I was annoyed about having to arrange yet another pre-operative test, but given the circumstances, I quickly complied. I wanted the neurosurgeon to have the clearer image she needed—she was cutting into my brain, after all. If the AURA had been around back then, maybe the MRI wouldn’t have been an eleventh-hour request.
But given the unease about the medical professionals’ training scarcity, I’m concerned that using the AURA (or other AI-generative tools) could lead to the equivalent of what Dr. Lomis might describe as a medical practitioners’ car collision. Without proper practitioner training, the platform might generate misinformation about a patient’s test results or provide inaccurate diagnosis assessments. The AURA platform was designed by Dr. D’Haese who has a Ph.D. in Electrical Engineering and Computer Science, but no medical background. Based on practitioners’ own feedback on AI use, they can’t fully benefit from something like the AURA without clear guidance and training on how to correctly use it.
So where does that leave patients with DRE (or any chronic illness)? Some are filled with anticipatory hope and optimism when we hear about the latest innovations, like the above headset, or the AI-powered Meld Graph, which I blogged about last year that can analyze MRI images. Many DRE patients who don’t have easy access to a neurologist who specializes in epilepsy, or an epileptologist are likely thrilled to use Dr. ChatGPT as a resource when they experience an unusual medication side effect or seizure symptom. While AI-generated medical advice could seem like a dubious source, if I lived in rural America where the nearest epileptologist was a four-hour drive, I’d likely rely on it.
Patients have been using AI for informal medical advice since ChatGPT’s 2022 release. A 2023 survey found that 21% of patients used ChatGPT as a source of advanced online health information. Users tended to be younger, in worse health, or have less access to traditional healthcare. While increased access to health information is a net positive, that benefit comes with complications. Patients may be making major medical decisions based on advice from generative AI-systems, rather than credentialed clinicians. The dynamic of automated tools influencing patients’ trust and choice raises multiple concerns including the tools’ embedded biases to mirror users’ expectations, and equity based on ease of access to them. Navigating this is tricky for both patients and providers but given that AI is here to stay, Dr. Lomis’s call to educate medical students and medical practitioners is key to successful steering. As Dr. ChatGPT looms larger in patients’ lives, healthcare providers must be prepared to proactively counsel their patients about how to optimize its use. This guidance is key to ensuring solid patient-doctor relationships and safe healthcare in our increasingly AI-driven world. Hopefully that Harvard Medical School AI workshop is well attended in June.
Leave a Reply