Pitfalls of spikes filtering for detecting High Frequency Oscillations (HFOs)

Cerebral High Frequency Oscillations (HFOs) have recently been discovered in epileptic EEG recordings. HFOs have been defined as spontaneous rhythmic oscillations with short duration, operating approximately in the frequency range between 80 Hz and 500Hz. HFOs have been considered as reliable and precise biomarkers for delineating the epileptogenic tissue. Also, HFOs have a profound impact for understanding the cerebral mechanisms involved in the generation of epileptic seizures. Therefore, several algorithms for HFOs detection with different performance and computational complexity have been proposed over the last few years. One of the major issues associated with HFOs detection algorithms applied on filtered EEG signals is how to differentiate spurious oscillations from true HFOs. The objective of this study is to highlight the original phenomena of spurious oscillations resulting from the filtering of simulated spikes. Our results are then validated on real spikes. In our study, three filtering methods are considered: the Finite Impulse Response (FIR), the Complex Morlet Wavelet (CMOR) and the Matching Pursuit based technique (MP).