Forecasting migraine with machine learning based on mobile phone diary and wearable data.

Area of Science:

• Neurology
• Data Science
• Digital Health

Background:

• Migraine attacks are preceded by triggers, premonitory symptoms, and physiological changes.
• Machine learning (ML) offers potential for predictive analytics in forecasting migraine attacks.

Purpose of the Study:

• To evaluate the effectiveness of ML in forecasting migraine attacks.
• To utilize preictal headache diary entries and physiological measurements for prediction.

Main Methods:

• A prospective study involved 18 migraine patients.
• Patients completed 388 headache diary entries and used app-based biofeedback sessions.
• Wireless sensors measured heart rate, skin temperature, and muscle tension; ML models were built to predict subsequent-day headaches.

Main Results:

• 295 days were used for predictive modeling.
• The best model, a random forest classifier, achieved an area under the receiver operating characteristics curve of 0.62.
• This indicates moderate predictive performance in forecasting migraine.

Conclusions:

• Mobile health apps and wearables combined with ML can forecast headaches.
• High-dimensional modeling is suggested to improve forecasting accuracy.
• Considerations for future ML-based migraine forecasting models using mobile health data are discussed.