An approach for reliably identifying high-frequency oscillations and reducing false-positive detections.

Area of Science:

• Neuroscience
• Epilepsy Research
• Signal Processing

Background:

• High-frequency oscillations (HFOs) are crucial biomarkers in epilepsy research.
• Accurate detection of HFOs is challenging due to false positives from motion and background noise.
• Translational studies require reliable and feasible HFO detection methods.

Purpose of the Study:

• To develop an improved automatic detector for high-frequency oscillations (HFOs).
• To enhance the feasibility and reliability of HFO detection for translational epilepsy studies.
• To specifically address and reject common sources of false positives in HFO detection.

Main Methods:

• An integrated, multi-layered pipeline was developed for automatic HFO detection and false-positive rejection.
• The method employs a time-frequency contour approach with peak constraints, power thresholds, and morphological identification.
• Validation involved evaluation by four experts on HFO events from posttraumatic epilepsy (PTE) animal models.

Main Results:

• The algorithm processed 768 hours of intracranial recordings from 48 PTE animals.
• Out of 453,917 initially detected HFOs, 203,531 events were retained after refinement.
• The HFO detection achieved high accuracy, precision, recall, and F1 scores, with expert agreement demonstrating reliability.

Conclusions:

• A segregated pipeline design focused on false-positive rejection significantly improves HFO detection efficiency and reliability.
• The developed pipeline uses fixed parameters, requiring no customization.
• This approach is highly feasible and translatable for both basic research and clinical applications in epilepsy.