Abstracts

Rationale: Unrecognized nightly seizures pose a serious problem in monitoring treatment response of patients with epilepsy. Therefore, there is a need for reliable seizure detection devices. In recent years, there have been advances with different sensors and wearable devices and the sensitivity to detect seizures has increased considerably. Specificity, however, remains a challenge and the false alarm rates of current systems are high. Radar technology is excellent at detecting movements and is not blocked by materials such as blankets. Compared to video, it works well in darkness and also avoids privacy issues associated with video cameras. The objective of our study is to assess whether radar sensors could be used to detect nightly generalized tonic-clonic seizures (GTCS).

Methods: We installed three radar devices (Texas Instruments, IWR6843ISK) in an epilepsy monitoring unit. The devices were placed two meters above the foot end of the bed. The radar antennas emit electromagnetic pulses which reflect off the patient and the reflections are recorded by the devices’ receiver channels. Stationary objects like the bed produce a static signal that can be subtracted, so that only movements are detected. Different properties of the reflected signal can be measured: angle of incidence, movement extent and movement velocity. The seizures were classified by an experienced epileptologist based on the video, recorded as part of the clinical routine.

Results: We recorded 83 seizures in 25 patients, of those 19 were GTCS. In all recorded seizures with a clear clonic phase, the radar signatures of the movements’ velocity and extent clearly show the rhythmic jerks of decreasing frequency typical for the clonic phase of GTCS. The frequency matches the one observed in the video and by electromyogram-artifacts, that can be seen on scalp electroencephalography. Conversely, during the tonic phase, there is very little signal. Compared with physiological movements occurring during sleep, GTCS could be detected with a sensitivity of 93% and a specificity of 84%.

Conclusions: This is a proof-of-concept study, which shows that radar devices are capable of detecting the clonic phase of GTCS with a high sensitivity and specificity and may be a viable tool in detecting previously unrecognized nightly seizures. This could improve therapeutic monitoring and also contribute to SUDEP prevention. The study is part of an ongoing project aimed at combining the radar data with wearable devices for a multimodal seizure detection system and training an artificial intelligence to detect and classify seizures based on these multimodal signals.

Funding: Research grant from the German Federal Ministry of Education and Research (Bundesministerium für Bildung und Forschung - BMBF)