Universal automated high frequency oscillation detector for real-time, long term EEG
Stephen V Gliske1, Zachary T Irwin2, Kathryn A Davis3
1Department of Neurology, University of Michigan, USA.
Summary
Automated detection of high-frequency oscillations (HFOs) in epilepsy EEG is improved by a new algorithm that reduces false positives. This enhances the precision of HFO biomarkers for clinical use.
Area of Science:
- Neuroscience
- Biomedical Engineering
- Epilepsy Research
Background:
- Interictal high-frequency oscillations (HFOs) detected via intracranial EEG are promising epilepsy biomarkers.
- Current automated HFO detection methods are limited by artifactual detections requiring manual review.
- Reducing false positives is crucial for the clinical utility of HFOs.
Purpose of the Study:
- To develop an automated method for redacting false HFO detections in intracranial EEG.
- To improve the precision of HFOs as a biomarker for epilepsy.
- To facilitate the clinical application of interictal HFO analysis.
Main Methods:
- Intracranial EEG data from 23 patients were analyzed using automated HFO and artifact detectors.
- High-frequency oscillations (HFOs) not coinciding with artifacts were classified as quality HFOs (qHFOs).
- The correlation between qHFO rates and the seizure onset zone (SOZ) was evaluated using retrospective and quasi-prospective algorithms.
Main Results:
- Human review indicated <12% of qHFOs were artifacts, compared to 78.5% of redacted HFOs.
- qHFO rates showed stronger correlation with SOZ (p=0.020) and resected volume (p=0.0037) than baseline detections.
- The algorithm identified SOZ in 60% of ILAE Class I patients using qHFOs, with all SOZs within resected areas.
Conclusions:
- The developed algorithm significantly reduces false-positive HFO detections, enhancing biomarker precision.
- This method enables feasible real-time, continuous EEG monitoring with minimal human oversight for data quality.
