Abstracts

Rationale: To investigate high frequency oscillations (HFOs) in temporal lobe epilepsy (TLE). Methods: We studied patients who had temporal resection after intracranial recording with subdural grids/strips and posteroanteriorly-inserted hippocampal depths. EEG was acquired at 1000 Hz sampling rate. We marked seizure onset visually (bipolar montage, 50-333 Hz filter, 2-s window) based on earliest occurrence of HFOs. If no HFOs were seen, seizure onset was marked based on earliest rhythmic activity (1-70 Hz, 10-s). Spatial extent of seizure onset was determined by the HFO channels with subsequent evolution (i.e., HFO+, see Modur et. al., Epilepsia 2011). For this study, we defined HFO by 2 criteria: oscillation with ≥4 peaks; peak frequency ≥70 Hz or total power in 70-333 Hz band higher than power in 1-70 Hz band. Using MATLAB, we developed an automated method to detect HFOs based on root mean square (RMS) of the EEG signal which was bandpass filtered at 50-333 Hz and notch filtered. We set variable thresholds based on the mean and standard deviation of the channel's RMS to identify putative HFOs. Such oscillations were rectified to count the peaks, and subjected to FFT to determine frequency and power before classifying them as definite HFOs. Consecutive HFOs (<200-ms and <20-ms apart for ictal and interictal respectively) were combined into a single HFO. We analyzed ictal (5-s segment) and interictal (10-min per patient sampled over 2 days) HFOs in the seizure onset channels (SOC) and non-seizure onset channels (nSOC) in both amygdalo-hippocampal (AH) and neocortical temporal (NT) channels. Statistical analysis was done with GraphPad. Results: Among 8 patients, we analyzed 22 seizures. All had ictal HFOs in at least 1 seizure. Overall, 53 AH (16 SOC, 37 nSOC) and 478 NT (82 SOC, 396 nSOC) channels were analyzed. Interictal data were available for only 6 patients. Although HFOs ranged up to a max of 257 Hz, the peak frequencies varied only marginally from the 70 Hz cutoff. Ictal HFOs had lower peak frequencies than interictal HFOs in both AH (62 vs 76 Hz, p=0.001) and NT (68 vs 77 Hz, p=<0.0001) SOC. However, peak frequencies of interictal HFOs were similar regardless of channel type (SOC vs nSOC) or localization (AH vs NT). Peak frequencies of ictal HFOs in SOC were lower than those in nSOC in both AH (62 vs 77, p=0.002) and NT (68 vs 73, p<0.0001) onsets; in addition, peak frequencies were lower in AH than NT (62 vs 68, p=0.02) SOC. Compared to nSOC, the SOC showed lower interictal HFO firing rates in AH channels (22 vs 39 per min, p=0.32) and higher firing rates in NT channels (40 vs 20, p<0.001). However, among nSOC, the rate was higher in AH than NT channels (39 vs 20, p=0.008). HFO durations were similar in SOC and nSOC (~49-57 ms) regardless of localization (AH vs NT). Conclusions: Ictal and interictal HFOs appear to be generated by different network (or populations) of neurons in TLE and its subtypes. Interictal HFOs are more likely to be associated with neocortical temporal seizure foci although they seem to occur more frequently in the non-seizure generating areas of the mesial temporal structures.