A migraine is a moderate or severe headache felt as a throbbing pain on one side of the head. It is a chronic, incapacitating neurovascular disorder. In the United States, approximately 1 in 7 adults and 1 in 5 of those in their peak employment years (aged 18 to 54 years) reported severe headaches or migraines in 3 months, according to a 2018 study published in the American Journal of Managed Care. In 2016, healthcare costs and decreased productivity amounted to an estimated total annual cost of $36 billion in the country.
Researchers from the Biomimetics and Intelligent Systems Group at the University of Oulu in Finland conducted a study to see if wearable sensors can detect migraine attacks beforehand. The researchers in the team were: Pekka Siirtola, Heli Koskimäki, Henna Mönttinen, and Juha Röning.
The aim of their study was to build reliable recognition models for the early detection of migraine attacks based on sleep time data collected using wearable sensors. The study is based on two hypotheses: (1) wearable sensor data can be used to detect an attack of illness beforehand, and (2) sleep time data contains information about the forthcoming migraine attack. Related work supports both of these hypotheses.
The researchers recruited 7 participants. Among the participants, 5 were female and 2 were male. They were aged between 30 to 60 years. Each of them was given an Empatica E4 wristband. Only sleep time data were used in this study. Data was collected for 27 days.
Results showed that the wearable device could not accurately detect a migraine attack beforehand.
“According to our preliminary results, the migraine attacks cannot be detected reliably beforehand using user-independent models,” the researcher wrote. “However, the used data set was most likely not comprehensive enough to build reliable user-independent models due to a limited number of study subjects. On the other hand, the small number of study subjects is not an issue when recognition is based on personal models instead of user-independent models. In fact, the results achieved using personal models indicate that early detection of migraine attacks is possible. When a personal early detection model based on QDA classifiers was used, the average balanced accuracy was over 84%. However, there is a great deal of variation between study subjects. In fact, when results are studied study-subject-wise, it can be seen that balance accuracy varies between 60.4% and 95.2%.”
They said that the small number of participants may be the reason the study didn’t produce accurate results.
“With a larger data set, it would be possible to experiment with the accuracy of personal models with all migraine types. If the findings of this article can be confirmed in a larger population, it may potentially contribute to early diagnoses of migraine attacks,” the researchers concluded.