To design a wearable, wireless system for detecting convulsive seizures in the community.

Epilepsy affects 1% of the population and is associated with increased morbidity and mortality. Accurate identification of seizures often relies on eyewitness accounts. Patients with convulsive seizures are at higher risk for injuries and death, including SUDEP (sudden unexpected death in epilepsy). There is an unmet need for developing reliable wearable sensors for detecting convulsive seizures, especially for use in the community.

Patients admitted to the epilepsy monitoring unit, whose semiology included motor seizures were consented for inclusion. Four Neblina triaxial accelerometer sensors on nordic semiconductor chips were placed in wrist and ankle bands connected to all four extremities. The chip sets communicated with a base station using a bluetooth low energy (BLE) platform. Time series of acceleration were acquired and processed. Seizure detection was performed using thresholding for values of cross-correlation between sensors, as well as amplitude and variance in individual signals.

We demonstrate the ability of the system to detect convulsive seizures. Initial funding and grant cycle allowed inclusion of six patients and two convulsive seizures were recorded. Both of these seizures were detected autonomously by the algorithm before being confirmed by an epileptologist.  Sensitivity exceeded 0.7. Furthermore, false positives were quite low, at  less than 0.3 per day. While BLE was convenient, about 7% of the data was lost from one of the sensors due to BLE range limitations and battery failures.

This small pilot study shows proof of principle for detection of seizures by patient-friendly sensors, using accelerometers and highly efficient signal processing techniques. We report the design and initial data, with future work to include further testing and development of this prototype.