AI predicts post-concussion headaches and fatigue 24 hours before symptoms appear

Foundation Models Detect Patterns in Post-Concussion Symptoms

Researchers at NTNU and the Norwegian Center for Headache Research (NorHead) have demonstrated that AI prediction can forecast post-concussion symptoms with moderate to good accuracy 24 hours before they occur1

1

. The study focused on individuals experiencing persistent headaches and fatigue long after their initial concussion, a condition affecting more than one in five people who sustain head injuries2

2

. Lead researcher Antonios Danelakis explains that symptoms don't occur randomly in people struggling with persistent post-concussion issues but follow specific patterns that machine learning models can detect through training on physiological data collected over time1

1

.

Physiological Signals Power Early Warning System

The proof-of-concept study at St. Olavs Hospital's Department of Rehabilitation involved 20 participants who logged their symptoms daily for 28 days while measuring heart rate, skin temperature, and tension in the trapezius muscle connecting the shoulder blade, back of head, and spine2

2

. Researchers combined health apps with foundation models—versatile, large-scale models trained on massive datasets similar to those powering ChatGPT and Claude1

1

. When tested against traditional machine learning models to predict moderate to severe symptoms over the next 24 hours, the foundation models yielded superior results for both headaches and fatigue prediction2

2

.

Smartwatch Alerts Could Transform Daily Planning

The ability to predict post-concussion headaches and fatigue through physiological signals opens possibilities for personalized healthcare for concussion recovery. Patients could receive notifications via smartwatch or digital app warning that symptoms are imminent, enabling them to prepare and plan their next day more effectively1

1

. This advance matters particularly given that more than 17,000 people in Norway visit emergency clinics or hospital emergency departments with minor head injuries annually, with the true figure likely twice as high when including general practitioner consultations and individuals who don't seek immediate medical attention2

2

. Healthcare professionals treating these patients would also gain valuable information for managing care.

Scaling Up Research for Clinical Applications

Danelakis acknowledges this remains an under-researched field requiring extensive next steps before solutions reach patients. The current study serves as a methodological proof-of-concept testing whether the idea is feasible1

1

. Critical requirements include more extensive testing, longitudinal patient monitoring studies, integration of alert-delivery technology into digital tools through user collaboration, and addressing legal regulations, privacy, and data security concerns2

2

. NorHead researchers have updated their app and plan to launch a new study in spring 2027, recruiting at least 200 patients with persisting post-concussion symptoms1

1

. The interdisciplinary collaboration required to bring this technology from research to clinical practice will determine whether AI can ultimately help patients regain control over their daily well-being, work participation, and social activities affected by long-term concussion complications.