AI uses body data to predict post-concussion headaches, fatigue 24 hours ahead

Researchers at NTNU’s Norwegian Centre for Headache Research have shown that artificial intelligence may be able to warn people recovering from concussion up to 24 hours before they develop headaches and fatigue. The study, published in the Journal of Head Trauma Rehabilitation, suggests that symptoms experienced the following day by patients with persistent postconcussion problems do not occur randomly, but follow patterns that can be identified in continuous physiological data. The work was led in part by Antonios Danelakis, a researcher at NTNU’s Department of Computer Science and the Norwegian Centre for Headache Research. The study used machine learning models to analyze symptoms and body data recorded by 20 people with ongoing issues after concussion. Participants took part in the research at the Rehabilitation Department of St. Olavs Hospital and logged their symptoms daily for 28 days. They also recorded physiological measurements including heart rate, skin temperature and tension in the trapezius muscle, a region involving the shoulders, back of the head and spine. The team tested both a traditional machine learning model and a foundation model, a type of large-scale AI system trained on broad data and later adapted for specific tasks. The foundation model performed best in predicting moderate to severe headaches up to 24 hours in advance. Researchers then used the same approach to assess whether fatigue could also be predicted, with early indications that symptom patterns may be detectable before they fully develop. The findings point toward future digital tools that could alert patients through a smartwatch, phone or health app that a headache or period of fatigue may be approaching. Such warnings could allow people to plan activities, rest or take preventive steps in advance. The tools could also support clinicians by providing objective, continuous data on how symptoms develop over time. The study also highlighted the scale of the problem. Researchers noted that more than 17,000 people in Norway visit emergency clinics or hospital emergency departments each year for minor head injuries, although the true number is likely much higher when general practitioner consultations and unreported cases are included. More than one in five people who sustain a concussion may still experience symptoms more than three months later, affecting daily life, work, school and social participation. The researchers emphasized that the work is an early methodological proof of concept. More extensive clinical testing is still needed before AI-based prediction tools can be used in routine care. Next steps include longer-term monitoring of patients, improved digital alert systems, user-centered design, and compliance with legal regulations, privacy protections and data security requirements. The NorHead team has now updated the study app and plans to launch a larger trial in spring 2027. Recruitment is expected to begin soon, with the goal of collecting data from at least 200 patients who experience persistent symptoms after concussion.