Abstracts

Rationale: Some seizures cause injury and even death, but there are no reliable detection methods that can alert a caregiver when a seizure occurs. We measured physiological responses caused by changes in autonomic nervous system activity in children with refractory epilepsy to identify patterns that correlate with seizures but not with non-seizure behavior. This pattern recognition algorithm will be used to develop a non-invasive wearable seizure alert system that detects generalized seizures with high sensitivity and precision. Methods: Six children with epilepsy, aged 3-18 and undergoing video EEG have been enrolled into the study with an enrollment target of 30 patients. During monitoring, the patients wore unobtrusive sensors that continuously recorded heart rate, respiration, torso orientation, surface electromyography (EMG), and skin conductance. A preliminary seizure detection algorithm based solely upon cardiac changes (% maximum heart rate, rate of change in heart rate) was developed and tested to detect generalized seizures and minimize false positives. Algorithms incorporating the full sensor array will be developed from the first 10 patients enrolled and tested on the data from the second set of 10 patients enrolled. Results: Fifteen seizures (3 generalized tonic clonic [GTC], 12 partial) were observed during 242 hours of monitoring. Significant changes in cardiac activity, respiration, muscle activity, and sweat gland activity were observed during the seizures, the most prominent being a rapid increase in heart rate (Figure 1). Maximum heart rate measured during GTC seizures was 197±1 bpm, and 154±35 bpm during partial seizures. For all seizures, heart rate elevation was observed within 5 to 20 seconds of electrographic onset, and it preceded or occurred simultaneous with the first clinical sign. Using the most stringent detection criterion, 3 out of 3 GTC seizures were detected, along with 2 false positives (FPs) or on average one every 5 days. Other physiological changes consistently observed during seizure events included a rapid change in respiration rate and amplitude, a characteristic rhythmic increase in muscle activity (EMG), and a rapid increase in sweat gland activity (skin conductivity). These findings confirm the physiologic changes we reported previously (Kroner et al, AES 2010) from seizures observed in the residential setting. Conclusions: Results support the hypothesis that generalized and some partial seizures can be detected by autonomic physiological changes. Detection of these changes using nonobtrusive sensors will form the basis of a seizure monitoring device for daily, non-clinical use. Use of multiple sensors, in addition to cardiac, will improve seizure detection sensitivity and specificity. Such a device could have impact on the potential prevention of SUDEP, status epilepticus, and seizure-related injury, as well as improvement in quality of life and increased independence for both caregivers and persons with epilepsy. Source of funding was RTI International.