A Device to Detect and Quantify Seizures Using Noncerebral Sensor Modalities
Abstract
Over 3.5 million people in the United States and 65 million people worldwide suffer from epilepsy. Traumatic brain injury (TBI) is a known cause in up to 20% of epilepsy diagnoses in the general population. Hundreds of thousands of service members have suffered TBI since 2000, and depending on the severity, 4% to 53% have or will develop post-traumatic epilepsy. As with the civilian population, there is no reliable method to determine a patient s seizure occurrence rate. For most epilepsy suffers there is a long trial and error process to select the anti-epileptic drug(s) that minimize the number of seizures while maximizing quality of life. This process relies on self-reporting of seizures by patients. Unfortunately, many patients are unable to determine or remember that they have had a seizure, leading to underreporting of seizure rates by 50% or more and hindering efforts to assess the efficacy of therapeutic regimens. Currently, epileptic seizures are identified by analyzing clinically obtained videoelectroencephalogram (Video-EEG) data. Electrodes thatdetect the change in voltage at the surface of the scalp are placed on the scalp and the patient is videotaped. The Video-EEG is not portable; it requires an expensive hospital stay and an epileptologist to interpret the results. This limits its use to hospitalized patients, although most patients seizures are experienced outside this setting. Recent studies, however, suggest that seizures can be statistically separated from normal activities using signal sources other than EEG. Variations in heart rate, breathing rate, and specific movements have all been shown to have positive correlations with seizures. We aim to develop a low-cost, continuously wearable device that accurately detects and records seizures using non-EEG signals. Our approach is to use multiple physiological signals combined with machine learning to detect diverse seizure manifestations. While seizure types vary, distinct changes in motor function, vocalization, heart rate, blood pressure, responsiveness to external input, and other readily observed behaviors provide abundant opportunity to sense and classify seizures relative to normal activity. Our non-portable prototype shows promise, but data from more patients will validate the approach across a wide variety of seizure types to determine what combination of physiological measures are most useful for detecting seizures, including distinguishing non-epileptic (psychogenic) spells from epileptic seizures. Having preliminarily detected a seizure, our device will ask the patient "Are you having a seizure?" Lack of a response will confirm the seizure. This "real-world" measure of seizure frequency and type will help doctors to tailor therapeutic regimens and to evaluate the efficacy of new anti-epileptic drugs taken by their patients. It will also improve patient safety by notifying bystanders and caregivers that a person is experiencing a seizure and needs protection from harm. In a broader context, the device will vastly increase the database of recorded seizures, information that might help identify pathophysiologic pathways leading to status epilepticus (a state of continuous seizures) and sudden unexpected death in epilepsy.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Oct 29, 2018
- Source ID
- W81XWH1810712
Entities
People
- Erik Kobylarz
Organizations
- Dartmouth–Hitchcock Medical Center
- United States Army