Kujawski et al. (2025)
- Authors: Sławomir Kujawski, Hanna Tabisz, Karl J. Morten, Aleksandra Modlińska, Joanna Słomko, Paweł Zalewski
- Institutes: Department of Hygiene, Epidemiology, Ergonomics and Postgraduate Education, Nicolaus Copernicus University, Toruń, Poland, Nuffield Department of Women’s and Reproductive Health, University of Oxford, Oxford, UK
- Publisher: Journal of Translational Medicine
- Link: DOI
Summary
This research represents a significant step toward an objective diagnostic test for ME/CFS by using artificial intelligence to identify specific patterns in heart and nervous system function. By finding clear physiological differences in blood pressure and heart rate variability, the study provides scientific evidence that validates patient experiences and offers a blueprint for more accurate clinical diagnosis. The high accuracy of the AI model suggests that autonomic data could eventually replace current exclusionary diagnostic methods.
What was researched?
The study investigated the use of an artificial intelligence pipeline to differentiate ME/CFS patients from healthy controls using high-frequency, beat-to-beat measurements of the autonomic nervous system and cardiovascular function.
Why was it researched?
ME/CFS lacks universally accepted objective diagnostic markers, meaning patients often face long delays while doctors rule out other conditions. Researchers aimed to find reliable physiological indicators that could provide an objective basis for diagnosis.
How was it researched?
This prospective case-control study involved 112 patients and 61 healthy controls. Researchers used a Task Force Monitor to record beat-to-beat heart rate, blood pressure, and stroke volume, then analyzed this data using a sequential machine learning approach combining a Transformer model and an XGBoost classifier.
What has been found?
The AI classifier achieved a high subject-level accuracy of 89%. Patients with ME/CFS were characterized by reduced cardiac vagal tone, higher sympathetic vascular tone, and lower stroke volume compared to healthy individuals.
Discussion
The findings suggest that autonomic dysfunction is a core measurable component of ME/CFS. While the AI model performed exceptionally well, further validation in larger, more diverse cohorts is necessary to ensure the results are applicable across the entire patient population.
Conclusion & Future Work
Beat-to-beat autonomic measurements significantly enhance the objective diagnosis of ME/CFS when paired with advanced AI analysis. This methodology offers a promising path toward establishing a standardized, evidence-based diagnostic tool for clinical use.