Automatic detection of high-frequency-oscillations and their sub-groups co-occurring with interictal-epileptic-spikes
Daniel Lachner-Piza1,2, Julia Jacobs3, Jonas C Bruder3
1Epilepsy Center, Medical Center-University of Freiburg, Breisacher Str. 64, 79106 Freiburg, Germany.
Objective:
High-frequency-oscillations (HFO) and interictal-epileptic-spikes (IES) are spatial biomarkers of the epileptogenic-zone. Those HFO spatially and temporally co-occurring with IES (IES-HFO) are potentially superior biomarkers, their use is however challenged by the difficulty in detecting the low amplitude HFO riding the high-amplitude and steep-waveform of IES. We aim to develop an automatic HFO detector with an improved performance with respect to current methods and that also correctly distinguishes IES-HFO from IES occurring in isolation (isol-IES). We also aim to validate the biomarker-value of the automatic detections by utilizing them to localize a surrogate of the epileptogenic-zone.
Approach:
We developed automatic-detectors of HFO-Ripples (80-250 Hz), HFO-FastRipples (250-500 Hz) and IES based on kernelized support-vector-machines (SVM). The training of the HFO-detectors was based on visually marked HFO and the parameter optimization during this training-process promoted the correct discernment between IES-HFO and isol-IES. Both HFO-detectors were benchmarked against other published detectors using databases with both visually marked and simulated gold-standards. The IES-detector was trained with a publicly available simulated database and tested against both simulated and visually marked gold-standards. To validate the detections' biomarker-value, the detectors were run on intracranial-EEGs from 8 patients and each with durations of 2-3 days, the detections' cumulated per-channel occurrence-rate was then used to predict the seizure-onset-zone as a surrogate of the epileptogenic-zone.
Main Results:
The HFO-detectors obtained at least 21 more F1-score points than previously published algorithms at the lowest signal-to-noise-ratio. The success achieved when discerning between IES-HFO and isol-IES was comparable to that of other published algorithms. The automatically detected IES-HFO and IES-Ripples showed the best biomarker-value to localize the epileptogenic-zone.
Significance:
The developed detectors are publicly available and provide a reliable tool to further study HFO, IES-HFO and their value as biomarkers. The putative higher biomarker value from IES-HFO was validated on automatically-detected, long-term data.
Related Concept Videos
10:22Interictal High Frequency Oscillations Detected with Simultaneous Magnetoencephalography and Electroencephalography as Biomarker of Pediatric Epilepsy
09:35Automatic Detection of Highly Organized Theta Oscillations in the Murine EEG
13:14Multi-electrode Array Recordings of Human Epileptic Postoperative Cortical Tissue
07:01Electrophoretic Delivery of γ-aminobutyric Acid (GABA) into Epileptic Focus Prevents Seizures in Mice
Electrophoretic Delivery of GABA into the Epileptic Focus of an Anesthetized Mouse
02:39In Vitro Recording of Theta Oscillations in the Mouse Hippocampus


