When seconds count: wearable data for seizure research
Reliably detecting epileptic seizures – and perhaps one day predicting them – is one of the most demanding tasks in clinical neurology. In the research project MEXT (Modular Extended Transcranial Magnetic Stimulation) at the Universität der Bundeswehr München, an interdisciplinary team is investigating how reliable patterns can be derived from vital data continuously recorded by wearables. Datico® contributes a central component: the MEXT app, which records high-resolution sensor data around the clock and makes it available for scientific analysis.
The Universität der Bundeswehr München at a glance
MEXT is an interdisciplinary research project at the Universität der Bundeswehr München, funded by the Bundeswehr’s Centre for Digitalisation and Technology Research (dtec.bw). It combines electrical engineering, physics, psychiatry, psychology and sports and materials science and, with around 50 participants at 13 sites, is a project of considerable scope. At its core, MEXT is developing a new generation of transcranial magnetic stimulation (TMS) – a non-invasive method for targeted stimulation of the brain – towards compact, mobile and data-driven systems. A central building block of this vision is data-based closed-loop approaches that continuously monitor vital parameters. Since TMS can in rare cases trigger an epileptic seizure as a side effect, this continuous monitoring of vital data is not only scientifically relevant but also safety-critical – and therefore the direct starting point for the wearable-based seizure detection project.
The challenge
Anyone who wants to detect seizures from physiological data needs two things: complete, high-resolution data capture over long periods and a reliable reference standard against which the wearable data can be validated. Conventional recording methods reach their limits here – media breaks between devices, limited sampling rates and the effort of securely combining data from different sources. For research in the MEXT project, this meant there was no way to record the variables of several wearables around the clock at maximum resolution, on the move and in a privacy-compliant way, and to pass them on to the analysing party.
The solution with Datico®
For this purpose, Datico® developed the MEXT app – based on Datico® Wearable Connect in its Bluetooth variant. The app records the variables of the wearables in use directly on a smartphone or mobile device – continuously around the clock and in high resolution.
The main devices used were the Polar H10 and the Polar 360 or Polar Loop band. The app records their variables live via Bluetooth – including the single-channel ECG of the Polar H10 at 130 Hz, acceleration data at up to 200 Hz and skin temperature and other parameters at correspondingly high sampling rates. The data is first stored locally on the mobile device and then transferred securely to the project partner TNG Technology Consulting, which analyses it using machine learning algorithms.
Connect, Understand and Act – on the foundation of “Secure” – in this specific project:
- Connect – several wearables (Polar H10, Polar 360 or Polar Loop) are connected via Bluetooth and recorded in parallel.
- Understand – high-resolution raw and vital data forms the basis from which patterns are derived using machine learning.
- Act – the recorded data is provided in a structured form for scientific analysis.
- Secure – transfer is protected and privacy-compliant – essential for sensitive patient data.
The approach is being validated in the neurology department of Ludwig-Maximilians-Universität München (LMU): in a specialised monitoring setting, patients are monitored around the clock by EEG. At the same time, the Polar wearables record the same period via the MEXT app. This produces a synchronised reference data set – the EEG as the established standard, the wearable data as the signal the models are to learn to interpret.
Results and benefits
The MEXT app has been in research use for several months. It reliably records the wearables’ variables at the intended high resolution, transfers them securely and provides them for analysis – so the technical foundation of the project is in productive use.
The scientific goal has two stages: first, the models are to learn to detect epileptic seizures from the wearable data at all; second, there is the prospect of predicting seizures early on the basis of the same data – that is, recognising them in advance. Datico®‘s decisive contribution lies in making the necessary data basis possible in the first place: complete, high-resolution, mobile and secure.
In the customer’s words
“With Datico®‘s MEXT app, we are recording our study participants’ vital data around the clock for the first time, at the resolution our analysis really needs – mobile, stable and privacy-compliant. That is the basis on which we can advance seizure detection scientifically.”
— Alexander Hainz, Project Lead Software Development, Universität der Bundeswehr München
Conclusion
The MEXT project shows how robust scientific findings can emerge from continuous wearable data. Datico® provides the data foundation for this – from connecting the sensors to securely handing the data over for analysis. In this way, high-resolution raw data becomes analysable patterns.
