Table 1.
Clinical data and implantation sites.
Figure 1.
HFO in a temporomesial recording.
(A) Raw iEEG 10 s epoch from channel HL1 in patient 1. (B) Raw iEEG at extended time scale of 500 ms. (C) Filtered iEEG of panel B with envelope (red line). The envelope satisfies the criteria for an EoI (Stage 1 of detection). The peak of the envelope is marked by dashed vertical lines in panels A, B and C. (D) Time frequency representation of the iEEG of panel B. The circle marks the peak of the envelope of the EoI. The “blob” represents the HFO. (E) Power spectral density (PSD, unit: 10log10μV2Hz−1) at the peak of the EoI. For the event illustrated here, there is a HFO peak at 116 Hz, a trough at 73 Hz and a low frequency peak at 47 Hz. The thin line shows the PSD of the same data calculated by the short-time Fast Fourier Transform for comparison.
Figure 2.
(A) Raw iEEG data 10 s epoch from a frontal channel in patient 6. (B) Raw iEEG at extended time scale. (C) Filtered data (blue line) with envelope (red line). The envelope satisfies the criteria for an EoI (Stage 1 of detection). While the high-frequency activity is separated by a trough (D, E), it is excluded from acceptance as HFO because the peak of the spectral power appears at frequencies above 500 Hz.
Table 2.
Spatial HFO distribution compared to seizure onset zone (SOZ).
Figure 3.
Interictal Epileptiform Spike (IES) without HFO.
(A) Raw iEEG 10 s epoch from channel HL6 in patient 1. (B) Raw iEEG at extended time scale of 500 ms. (C) Filtered iEEG of panel B with envelope (red line). The envelope satisfies the criteria for an EoI (Stage 1 of detection). (D, E) The peak of the high-frequency activity and the trough coincide at 60 Hz, i.e. the EoI is not separated from low-frequency activity by a trough and the EoI is therefore excluded from acceptance as HFO.
Figure 4.
HFO with frequency peak below 80 Hz.
(A) Raw iEEG data 10 s epoch from channel HL6 in patient 1. (B) Raw iEEG at extended time scale of 500 ms. (C) Filtered iEEG with envelope (red line). The envelope satisfies the criteria for an EoI (Stage 1 of detection). The EoI is salient enough to be separated from low-frequency activity by a trough (D, E). This EoI has the visual appearance of a HFO but the peak frequency is around 60 Hz. Therefore, in Stage 2 the lowest boundary for a HiFP was chosen at 60 Hz and this EoI was accepted as HFO by our detector.
Figure 5.
HFO in a neocortical recording.
(A) Raw iEEG data 10 s epoch recorded from a frontal channel in patient 6. (B) Raw iEEG at extended time scale of 500 ms. (C) Filtered data with envelope (red line). The envelope satisfies the criteria for an EoI. The peak of the envelope is marked by dashed vertical lines in panels A, B and C. (D) Time frequency representation of the iEEG of panel B. The circle marks the peak of the envelope of the EoI. (E) Power spectral density (PSD, unit: 10log10μV2Hz−1) at the peak of the envelope. The high frequency peak is at 82 Hz, the trough at 40 Hz and the low frequency peak at 30 Hz for this event, which was accepted as HFO in Stage 2 of the detection.
Table 3.
Temporal and spectral characteristics of HFOs in individual patients.
Figure 6.
Fast ripple in a temporomesial recording.
(A) Raw iEEG data 10 s epoch recorded from channel AR1 in patient 3. (B) Raw iEEG at extended time scale of 500 ms. (C) Filtered data with envelope (red line). The envelope satisfies the criteria for an EoI. The peak of the envelope is marked by dashed vertical lines in panels A, B and C. (D) Time frequency representation of the iEEG of panel B. The circle marks the peak of the envelope of the EoI. (E) Power spectral density (PSD, unit: 10log10μV2Hz−1) at the peak of the envelope. The high frequency peak is at 292 Hz, the trough at 192 Hz and the low frequency peak at 153 Hz for this event, which was accepted as a FR in Stage 2 of the detection.
Figure 7.
Channels with high HFO rate are assumed to mark the epileptogenic zone. Channels with HFO rate above the half maximum (black bars) constitute the HFO area. The channel locations of the HFO area in each patient is given in Table 3.
Figure 8.
HFO peak frequency distribution.
(A) Total of all HFOs recorded from the four patients with temporomesial electrodes (N = 1179). The sharp edge to low frequencies stems from the frequency threshold of 60 Hz of the detector. (B) Total of all HFOs recorded from the two patients with neocortical electrodes (N = 3561). The distributions are fitted with a lognormal function (red line). The frequency peaks of HFOs from temporomesial recordings are more widely distributed than those from neocortical recordings.
Table 4.
Comparison to existing detectors.