A Comparaison of Methods for Detection of High Frequency Oscillations (HFOs) in Human Intacerberal EEG Recordings

High Frequency Oscillations (HFOs) in the range of 80-500Hz seem to be a reliable bio marker of tissue capable of producing seizures. Visual marking of HFOs related to long duration/multi channel EEG data is extremely tedious, highly time-consuming and inevitably subjective. Therefore, automated and reliable detection of HFOs is much more efficient. The purpose of the present study is to improve in the first stage three existing HFOs detectors (CM V, MP, BMT), and subsequently compare them on the same database. Our main findings are summarized as follows: The efficiency of methods depends on the required sensitivity and the False Discovery Rate (FDR). First, if the required sensitivity below 87% is sufficient for the intended application, CM V method could perform well in terms of low false detection rate (FDR<14%). Secondly, if the application requires a sensitivity between 87% and 92%, the three methods could perform in a similar way in terms of performance, wh ich appro ximately corresponds to an FDR in the range of 14-19%. Finally, if a high sensitivity is required (92% up to 98%), BMT based method can be considered the most efficient and leads to significantly lower FDR values (19% to 23%) co mpared to other methods.