Continuous and automated data collection in migraine research - Extending the data collection capabilities of the Empatica E4

Abstract

Migraine is a recurrent headache disorder that afflicts significant portions of the global population. There is no current cure and migraines are mainly managed through symptomatic medical treatments and manual biofeedback routines. Automated data collection and prediction of migraine attacks through machine learning could be viable approaches for helping migraineurs and for reducing the impact of migraines, both on a societal and an individual level. However, machine learning approaches require access to large amounts of high-quality real-time data for facilitating prompt and reliable prediction under everyday conditions and within useful timeframes. The Empatica E4 is an unobtrusive wearable sensor device that can satisfy these data collection needs, although not without flaws and shortcomings. Several studies have reported issues with E4 data collection, most regarding participant involvement and the logistical aspects of the collection process. On top of this, the native systems provided by Empatica for storing, retrieving, and utilizing collected data do not properly facilitate real-time data analysis or machine learning approaches. This project creates a flexible data collection solution based on the E4 for facilitating real-time prediction of migraine attacks. It incorporates features and elements for increasing user involvement and for maximizing the data collection potential of the E4. Additionally, the solution is integrated with the mSpider data storage platform, facilitating reliable and flexible data storage and retrieval options. The prototype system was tested on three potential end-users under everyday conditions over the course of 20 days. After the data collection period, each user attended a semi-structured interview. Testing and interview results show that the data collection capabilities of the prototype system are on-par with other similar systems, it offers stable data collection under everyday conditions, and it can store data in the mSpider system. However, the added features for increasing participant involvement had little discernible effect on the data collection process or the amount of collected data. This was probably caused by the low intensity of the added features or the short duration of the testing period. Additionally, the testing process found that the high technical proficiency requirements and the necessary daily maintenance of the E4 makes it unsuited for continuous migraine treatment purposes, although it is a good tool for migraine research. Future prototype iterations should increase the intensity of the participant involvement features and greatly increase the length of testing periods.