University of Twente Student Theses


Using seizure detection algorithms in the Epilepsy Monitoring Unit to improve staff response to seizures

Rommens, Nicole (2017) Using seizure detection algorithms in the Epilepsy Monitoring Unit to improve staff response to seizures.

[img] PDF
Abstract:Background: At an Epilepsy Monitoring Unit, people with epilepsy can be admitted to answer different diagnostic questions. EEG, ECG and video are recorded and monitored by staff. When a seizure is recognized, patient safety is secured and standardized tests are performed. These standardized tests assess consciousness, cognitive and neurological functioning. Objective: Around 33% of seizures are missed by staff at the Epilepsy Monitoring Unit of SEIN. A response time over 60 seconds is seen in 19% of responses. Research has shown that this can be improved with EEG seizure detection algorithms. However, specificity was not researched. It remains unclear if the algorithm requires improvement before it can be used online. The goal for this project is to investigate which changes, in terms of sensitivity, specificity, latency and alarming, could make usage feasible. Methods: EEG and ECG recordings of seizures were included in a seizure database and non-stop recordings in a 24-hour database. False positive rate for EEG algorithms was researched. An ECG seizure detection algorithm was created to improve detection latency, which consists of the Pan Tompkins algorithm, processing of the heart rate and a classifier with self-adjusting thresholds. The performance and added value on top of the EEG seizure detection algorithm was researched. Lastly, a literature study was performed on alarm systems, to evaluate what kind of system would be suitable. Results: The seizure database included 188 seizures and the 24-hour database 1235 hours of 60 patients. The EEG algorithms showed a false positive rate of 4.9 (Encevis EpiScan) and 2.1 (BESA Epilepsy) per 24 hours. The ECG algorithm showed a sensitivity of 47.8% with a median latency of 36.5 seconds and a median false positive rate of 0 per 24 hours. There was added value in terms of extra detected seizures when it would be used as an addition to the EEG algorithm. However, no faster latency was seen. Lastly, the response to alarms depends on different behavioural mechanisms, which have to be taken into consideration when designing an alarm system. Discussion: Results show that it is feasible to use the EEG seizure detection algorithm in terms of a low false positive rate. The ECG seizure detection algorithm could be used to detect extra seizures. Further work should focus on improvement and optimization. For the alarm system, it is recommended to combine auditory and visual alarms. More research on novel alarm types is necessary.
Item Type:Essay (Master)
Faculty:TNW: Science and Technology
Subject:44 medicine
Programme:Technical Medicine MSc (60033)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page