Artificial Intelligence > Mayo Clinic Using AI to Pinpoint Migraine Treatments

Mayo Clinic Using AI to Pinpoint Migraine Treatments

By Rich Polikoff Photography by Paul Flessland Illustration by SERIFA

Treating migraine has long been an exercise in uncertainty.

The unpredictability of therapeutic responses frustrates both doctors and their patients suffering from migraine and other forms of recurring headache. While tension-type headaches occur more frequently in the population, migraine are the most common disabling form.

But Mayo Clinic scientists are using new technologies to demystify migraine treatment — advancing toward a future where personalized medicine delivers hope and healing to patients who have been disappointed in their search for relief.

“When treating migraine, about 30% to 50% of people who try a specific treatment will actually have a tangible benefit,” says headache specialist Todd Schwedt, M.D., who is chair of the Headache Division in the Department of Neurology at Mayo Clinic in Arizona. “We’ve been working for years on ways to identify which treatment is the best one for an individual patient, so that the best treatment is the first one prescribed.”

Chia-Chun Chiang, M.D., a Mayo Clinic headache specialist in Minnesota, centers her clinical and research focus on the application of artificial intelligence (AI) and machine learning to the study of migraine and vascular neurology.

Dr. Chiang, a Gerstner scholar, was the lead author of a 2024 study developing machine learning models to predict treatment responses to commonly used migraine preventive medications. For this work, she was the winner of the annual Harold G. Wolff Lecture Award from the American Headache Society, given for the best scientific paper on headache, head or face pain, or the nature of pain itself.

Related Content

Time Is Brain

Mayo Clinic experts are leveraging AI to transform stroke care and outcomes.

Neurologist Michael Cutrer, M.D., who was the longtime head of the Headache Division at Mayo Clinic in Minnesota, was the paper’s senior author. Dr. Cutrer has treated patients with complex headache syndromes for more than 25 years and is the creator of the Mayo Clinic Headache Subspecialty Clinic Database, a key foundation for Mayo Clinic’s research.

He says one of the biggest obstacles that has hampered the effective treatment of migraine has been a limited understanding of its underlying mechanisms.

"This challenge has resulted in a trial-and-error approach that can take months or years for each patient before finding a treatment that works for them,” Dr. Cutrer says.

Researchers demonstrated that migraine characteristics — such as headache intensity, location and frequency — are important in predicting treatment responses to individual classes of preventive medications. These results suggested that precision migraine treatment is feasible.

"We're developing self-learning algorithms that incorporate patients’ treatment outcomes,” Dr. Cutrer explains. "This continuous refinement will help us provide increasingly precise treatment recommendations."

The study led by Dr. Chiang analyzed data from the Mayo Clinic Headache Subspecialty Clinic Database prospectively collected from 2001 to 2023. The database includes detailed clinical characteristics and headache symptoms from about 17,000 patients who have completed in-depth questionnaires, which asked patients to describe their headaches, associated symptoms, family history of headache, migraine triggers and medications they’ve tried.

We have additional work to do, but we are close to having clinically useful predictive models that increase the chance that the first migraine medication a patient takes is the one that works.
— Todd Schwedt, M.D.

During follow-up visits, patients are asked about the change in their headache frequency. Their treatment responses provide valuable additional data and allow Mayo Clinic researchers to further use AI to personalize care as they develop algorithms that use natural language processing.

Headache researchers have historically needed to review patient records to obtain information from patients’ verbal responses to provider questions. This manual process significantly slows down the acquisition of potentially valuable data.

“We developed a model that can analyze language to extract headache frequency — a key parameter for measuring treatment response — directly from medical records," Dr. Chiang says.

In addition to the 17,000 patients in the Mayo Clinic Headache Subspecialty Database, Dr. Schwedt led the effort in constructing a Mayo Clinic Migraine Dataset that includes data from 200,000 patients with migraine who have been seen in Mayo Clinic Health System, which will further enrich the ongoing effort from the team using AI to tackle this unmet need.

The results of this study are beginning to take some of the guesswork out of therapeutic responses.

“Based on this work, we can identify some clinical features that predict treatment response,” Dr. Chiang says. “Headache-associated symptoms, things that trigger headaches, these are predictive of treatment response to many different migraine preventive medications.”

Researchers are also studying genetic biomarkers related to headache. Their analysis includes comprehensive biological data such as proteins within cells (proteomics) and small molecules involved in metabolism (metabolomics).

Mayo Clinic has one of the world’s leading research and clinical programs in headache medicine. Other Mayo Clinic contributors included scientists from the departments of Neurology, Cardiology and Radiology — spanning Mayo Clinic’s locations in Minnesota and Arizona.

“We’re leading the effort to use AI modeling to predict migraine treatment responses,” says Dr. Schwedt, who is the current president of the American Headache Society. “We have additional work to do, but we are close to having clinically useful predictive models that increase the chance that the first migraine medication a patient takes is the one that works.”