Assessing Performance of Detectors of High Frequency Oscillations in EEG Signals
Abstract
Attempts to perform epileptic seizure prediction have been made for decades, but there is no solution yet that is generally effective, though some progress in this area has been reported. Interictal epileptic spikes were considered as the only biomarkers of seizure-generating brain tissue until the recent discovery of high-frequency oscillations (HFOs) in electroencephalographic (EEG) signals, with frequency contents in the range from 80 Hz to 800 Hz. HFOs are now considered as biomarkers of epileptogenic tissue and the seizure onset zone. However, there are challenges in the definition and detection of HFO events which complicate this perspective. In studies about automatic HFO detection in EEG recordings, visual markings of HFOs by experts are considered as gold standards to compare the performance of detection algorithms. However different authors define their gold standards in variable manners and HFO detection methods are used to produce candidate HFOs which must be further classified visually to declare them as true HFO events. Likewise, when comparing automatic detectors, only HFO rates per unit time are noted without a detailed analysis of candidate vs true HFO events. Also, there is no consistent method as to whether all or some part of such events must be considered for a given analysis. This work correlates events detected with different automatic detectors independently of any detection by experts and analyzes the effect and potential benefit of performance differences in automatic detectors. It also briefly analyzes the possibilities and options for using multiple HFO detectors together. Results show that combination of detectors gives better performance than the individual detectors.
Subject Area
Electrical engineering
Recommended Citation
Bhat, Deeksha Seetharama, "Assessing Performance of Detectors of High Frequency Oscillations in EEG Signals" (2018). ETD Collection for University of Texas, El Paso. AAI13424605.
https://scholarworks.utep.edu/dissertations/AAI13424605