Abstracts

Rationale: Unique high-frequency oscillations (HFOs) of 250-500 Hz, termed fast ripples (FR), have been identified in seizure generating limbic areas of kainic treated rats, and in patients with mesial temporal epilepsy using depth electrodes. Results from these studies have provided evidence supporting the view that fast ripples in the human brain reflect localized pathological events related to epileptogenesis. These fast ripples appear to reflect field oscillations composed of hypersynchronous action potentials of pathologically interconnected neuronal clusters. Several studies have demonstrated that HFOs may increase before and at the time of seizure onset in epileptic patients. Studies to date have relied mostly on invasive depth electrodes to detect fast ripples. We propose the use of MEG as a reasonable non-invasive method to detect fast ripples in epileptic patients.Methods: MEG data was recorded on an epileptic patient during the patient’s routine pre-surgical evaluation. Data was recorded on 248 MEG channels at a sample rate of 1017.25 Hz, high pass filter 0.2 Hz and bandwidth of 400 Hz. Data was then transferred to MATLAB for power spectral density analysis. Sequential periodograms were run on a moving 250 ms window over the 10 seconds of MEG data. Fast ripple frequencies of 250-500 Hz were divided into 17 frequency bins and the power calculated for each. The mean and standard deviation was calculated for each frequency band and the mean plus three standard deviations was set as the cut off limit for determining outliers. The outliers were then analyzed using a three dimensional matrix to determine the longest duration of each high frequency oscillation per frequency bin, and per MEG channel.Results: Outliers for a given frequency bin for the 250 ms windows, defined as those signals with a power greater than the mean plus three standard deviations, and for at least 3 consecutive windows, were detected in the MEG data at various frequency bins >250 Hz (see graph). The longest duration of any of the outliers was 203 ms (with a minimum duration of 1 ms). The bin frequency for the longest duration outlier was 365 Hz. The mean duration of all outliers was 3.2 ms.Conclusions: We believe these spectral outliers represent fast ripples previously described in the literature. The current methodology described above demonstrates by power spectral analysis HFOs in multiple channels and bandwidths in MEG data raising the possibility of detecting HFOs through non-invasive methods. That the outliers represent true high frequency bursts rather than simple statistical anomalies is supported by the effect of the 400 Hz filtering, seen on the graph as a precipitous drop off of signal power. Further studies are needed to determine the anatomic localization and distribution of these HFOs, and their potential clinical significance in the normal and epileptic brain. MEG may provide a unique method to explore and classify HFOs and determine whether they are truly related to epileptogenesis and localization.