High Frequency Oscillations and spikes: Separating real HFOs from false oscillations

Area of Science:

• Neuroscience
• Signal Processing
• Epilepsy Research

Background:

• High Frequency Oscillations (HFOs) are crucial biomarkers in epilepsy research, often detected using filtered EEG signals.
• Filtering processes can inadvertently generate spurious HFOs, complicating accurate detection and interpretation.
• Distinguishing true HFOs from artifactual signals is essential for reliable HFO analysis.

Purpose of the Study:

• To quantify the occurrence of false High Frequency Oscillations (HFOs) arising from signal filtering.
• To develop and validate a method for distinguishing genuine HFOs from spurious ones by analyzing raw EEG signals.
• To improve the accuracy and reliability of HFO detection in epilepsy research.

Main Methods:

• A novel method was developed to detect oscillations in the raw EEG signal concurrent with sharp events.
• Support vector machines were employed to classify sharp events with and without HFOs based on temporal features in ripple and fast ripple bands.
• The traditional time-frequency representation was used as a benchmark for validating real versus false HFOs.

Main Results:

• A substantial percentage of detected HFOs were identified as false positives: 44% of ripples and 43% of fast ripples associated with sharp events.
• Sharp events accompanied by true HFOs exhibited significantly more oscillations in the raw signal compared to those without.
• The new method achieved classification accuracies of 76.6% for ripples and 72.6% for fast ripples in distinguishing real from false HFOs.

Conclusions:

• Detecting HFOs by analyzing oscillations in both raw and filtered EEG signals is recommended for increased accuracy.
• The classical time-frequency display for HFO identification should be used cautiously due to potential masking effects from broadband activities.
• This improved HFO detection methodology enhances the validity of findings and supports future research in epilepsy.