Using multiple wearable devices, the My Seizure Gauge project has made strides in gathering data for the development and validation of a noninvasive, wearable biosensor system for forecasting seizure probability. This according to research presented at the 2019 American Epilepsy Society Annual Meeting, held December 6 to 10, in Baltimore, Maryland.
Despite therapeutic advancements, individuals with epilepsy continue experience seizure associated morbidity. While a reliable warning system for seizures has the potential to help patients manage breakthrough seizures, at present seizure forecasting has only been successfully demonstrated via intracranially implanted electroencephalograph (EEG) devices. As such devices may not be a viable option for all candidates, researchers sought to use EEG-based seizure records collected from the My Seizure Gauge project to develop and validate a noninvasive wearable biosensor to forecast the probability of seizures.
This 3-year project began year 1 with the evaluation of commercially available, noninvasive, wearable sensors to assess patient acceptability, signal quality, and ability to predict and detect seizures in patients undergoing various forms of EEG. In year 2, participants trialing sub-scalp EEG devices and/or ambulatory intracranial EEG devices wore sensors to correlate biosignal records with EEG seizure annotations over multiple months. In year 3, a machine-learning competition was conducted to develop seizure forecasting algorithms on biosignals. Biosignals were obtained from commercially available wearable sensors, including electrodermal activity, accelerometry, photoplethysmography, electromyography, heart rate, scalp EEG, and temperature.
Investigators enrolled 59 patients with epilepsy, 53% of whom are women (median age 31), and 14% of whom are in the pediatric age group. Current data includes ≥305 days of monitoring, during which 153 seizures were recorded. Additionally, 32% (n=19) of patients had stereotactic EEG recordings, 49% (n=29) had ambulatory scalp EEG recordings, 15% (n=9) had in-hospital scalp EEG recordings, 2% (n=1) had subdural invasive EEG recordings, and 2% (n=1) had an implanted device.
Seizure localization was follows: 15% (n=8) generalized or nonlocalized; 4% (n=2) right occipital, 2% (n=1) right frontal, 5% (n=3) right temporal, and 7% (n=4) left temporal.
In total, 96 seizures were recorded via wrist electrodermal activity, 76 via chest accelerometry, 122 via wrist accelerometry, 37 via chest photoplethysmography, 96 via wrist photoplethysmography, 9 via wireless scalp EEG, and 32 via electromyography.
Investigators concluded, “The My Seizure Gauge project has made significant progress gathering seizure data using multiple different wearable devices.”
Brinkmann BH, Nurse E, Nasseri M, et al. My Seizure Gauge: Seizure detection and prediction with noninvasive wearable devices. Presented at the American Epilepsy Society Annual Meeting; December 6-10, 2019; Baltimore, MD. Abstract 1.092.