Using AI to Assist in Identifying Cardiovascular Risk in Patients With Migraine

Chia-Chun Chiang, MD

(Credit: LinkedIn)

Migraine remains a complex neurological disorder without clear diagnostic biomarkers or standardized treatment pathways, making the identification of response predictors increasingly important for clinicians.¹ As artificial intelligence (AI) continues to expand in neurology, its potential to improve care for patients with migraine is gaining attention in the headache field. From possibly improving diagnostic accuracy and patient classification to analyzing brain imaging and predicting treatment response, AI-based tools may offer promising support for both clinicians and their patients navigating migraine management.

At the 2025 American Headache Society (AHS) Annual Meeting, held June 19-22, in Minneapolis, Minnesota, neurologist and headache specialist Chia-Chun Chiang, MD, presented findings from a retrospective Mayo Clinic study that evaluated how AI-ECG analysis, echocardiographic parameters, and migraine-specific clinical characteristics could help predict cardiovascular risk in patients with migraine. The study also explored differences in vascular risk between patients with migraine with aura versus without aura. In the session, she highlighted that certain features may be associated with increased risk of stroke and adverse cardiovascular outcomes.²

Chiang, associate professor of neurology at Mayo Clinic Rochester, sat down with NeurologyLive during the meeting to discuss the findings in detail. During the interview, she spoke about the use of AI-predicted atrial fibrillation probability and delta ECG age as potential predictors of adverse vascular outcomes in a large patient cohort. Additionally, she talked about how migraine-specific and cardiac imaging characteristics were also shown to correlate with stroke and cardiovascular disease risk, pointing to the potential for multimodal AI-assisted screening in clinical headache management.

NeurologyLive: Discuss your presentation in greater detail–what were the most notable takeaways?

Chia-Chun Chiang, MD: I did a presentation at one of the plenary sessions, the stroke and migraine session, on our study using different AI tools to study the risk of stroke and cardiovascular outcomes in patients with migraine. The reason we wanted to do this study is that, as a practicing headache and stroke specialist, I think about this question very frequently. We see patients with migraine very often in the clinic, literally every day, but some of these patients end up in the hospital with a stroke or debilitating cardiovascular disease. Many of these patients are young and don’t have vascular risk factors. So, how do we identify these patients early to avoid developing these adverse vascular outcomes? That was mainly the motivation for this study.

In the study, we integrated several AI tools—one of the main ones being an AI ECG–based analysis. In addition, we also incorporated other factors, like echocardiography (heart ultrasound results) and detailed headache characteristics, to see whether these factors, individually or together, can help us identify patients with migraine who are at risk for vascular outcomes.

What may possibly be the clinical implications of integrating AI-based ECG prediction tools into routine migraine management?

We investigated AI ECG, which is a very cool tool available at Mayo Clinic. For every ECG done at Mayo Clinic, we can view the results from different AI algorithms. For example, we have one AI algorithm that calculates the probability that someone may have undiagnosed paroxysmal atrial fibrillation. Another algorithm estimates the ECG-based age. We used those 2 algorithms, primarily the one calculating atrial fibrillation probability and the ECG-estimated age, to calculate what we call delta age. That’s the AI-estimated age minus the patient’s actual age, which we refer to as "ECG age." A higher delta age has been shown to be a marker of endothelial dysfunction.

So, we used those 2 AI ECG algorithms to predict cardiovascular outcomes in a cohort that included almost 30,000 patients with migraine. We found that higher outputs from the atrial fibrillation prediction model and larger delta age values were predictive of adverse cardiovascular outcomes in patients with migraine. That suggests AI ECG can be a very quick and effective tool. It’s just a 12-lead ECG that can be done quickly in the clinic, and it could help stratify patients into high- and low-risk groups.

I think that’s one of the most important results and clinical implications. In the future, in addition to what we already do in clinical practice, assessing vascular comorbidities and other health conditions, we could incorporate tools like a 12-lead ECG into the evaluation to assess whether a person is at higher risk of developing cardiovascular outcomes.

Separately, we also analyzed echocardiogram features, meaning heart ultrasound, and identified several abnormalities, such as those in the tricuspid valve, left and right ventricles, and regional wall motion. These abnormalities were predictive of adverse vascular outcomes. Similarly, in terms of migraine characteristics, patients with more frequent or more severe migraine were also found to be at increased cardiovascular risk.

These are all very important research findings. Going forward, we are working on building a multimodality prediction model that incorporates all these results. Hopefully, in the future, when patients come into our Headache Clinic, we can use this model to say, “Based on these features, and incorporating AI tools, this patient is at higher risk of stroke or other vascular outcomes.” That would allow us to monitor their vascular risk more closely and implement prevention strategies earlier to reduce that risk.

In your perspective, what challenges or limitations could exist when applying AI ECG screening specifically for stroke and CVD risk in migraine populations?

These research results are certainly exciting, and we’re still working on validating them in a separate cohort. As for challenges, I do think this could be a very useful clinical tool. Using an AI-based screening approach to identify high-risk patients is promising.

However, access is always a potential limitation. Not all institutions or healthcare systems have access to ECG, echocardiography, or AI-based tools. So, access will be a big issue. Another challenge is the understanding, by both clinicians and patients, of how to interpret these results. That’s something we’ll need to work on. Perhaps by providing more education, clinicians can better understand what these tools mean and how to apply them in clinical practice. I think those are the main limitations and challenges.

Do you have a closing remark based on your presentation given at AHS 2025?

I’d just like to add that, based on our study, we identified an association that there’s an increased risk of stroke and cardiovascular disease in patients with migraine, especially those with migraine with aura. We used different modalities like ECG, AI-based tools, heart ultrasound results, and detailed migraine characteristics for risk stratification. We found that several of these features can predict stroke and cardiovascular outcomes. I think, in the future, these tools can be implemented into clinical practice. It’s also very important to have discussions with our migraine patients, especially emphasizing the potential vascular risks and how to best prevent them.

Transcript edited for clarity. Click here for more coverage of AHS 2025.

REFERENCES
1. Torrente A, Maccora S, Prinzi F, et al. The Clinical Relevance of Artificial Intelligence in Migraine. Brain Sci. 2024;14(1):85. Published 2024 Jan 16. doi:10.3390/brainsci14010085
2. Chiang C. Artificial Intelligence ECG-Based Screening and Prediction Tool to Identify Migraine Patients at Risk for Stroke and CVD (CL). Presented at: 2025 AHS Annual Meeting; June 19-22; Minneapolis, MN. Plenary 2 Session: STROKE.