Epileptic seizure detection in children and young adults in their home replacement environment
10 November 2016
UZA, Auditorium Kinsbergen (route 12) - Wilrijkstraat 10 - 2650 EDEGEM
Anouk Van de Vel
Prof B. Ceulemans & Prof P. Cras
PhD defence Anouk Van de Vel - Faculty of Medicine and Health Sciences
Epilepsy is a neurologic condition affecting 1% of the world population. The occurrence of unprovoked, unexpected seizures has a serious impact on quality of life, certainly for 30% of patients whose seizures cannot be controlled. An alarm system warning someone in case of a seizure could facilitate emergency treatment. The gold standard for seizure detection is brain activity measurement, but electrodes attached to the scalp are uncomfortable and signal analysis is not automated yet. Alarm systems therefore use methods measuring other body signals: heart rate, oxygen saturation, respiration, body movement, muscle tension, blood pressure, sweating and sound. There are some commercial devices, but experience with those is limited or negative (missed seizures, false alarms).
We created an overview of detection methods, state-of-art research and commercial devices, which proved to be helpful in guiding patients to find an appropriate system.
We investigated necessity of and requirements for seizure detection by questioning patients, caregivers and doctors. It seems acceptable to develop a device with 90% correct detections and one false alarm per seizure, focusing on substantial movement, falls and heart rate changes during seizures.
We also discuss difficulties in seizure detection research.
What are dangerous seizures? We focus on nocturnal hyperkinetic frontal lobe seizures and tonic-clonic seizures, possibly leading to injuries, status epilepticus (prolonged seizure state) or death. We also learned that next to seizure type, duration, intensity, pathophysiological changes and circumstances define danger.
How to take into account seizure variation between and within patients? We promote a combination of novelty detection (allowing device installation without prior knowledge on that patient’s seizures) and active learning (gradually replacing the generic algorithm with patient-specific data).
Which sensors should be used? It seems a combination of methods generates the best results, and we are convinced this should include movement and heart rate monitoring. Ideally, sensor selection should be made possible depending on the patient and on use for daytime versus nocturnal seizures.
Which options should a system incorporate? Patient comfort and user friendliness need to be taken into account.
A seizure detection system will never be 100% reliable but offers improved care, reassurance and independence.