Abstracts

Rationale: Intracranial high frequency oscillations (HFO) have been extensively investigated as a biomarker for the seizure onset zone (SOZ) in patients with drug-resistant epilepsy. Although the resection of brain tissues that generate high HFO rates positively correlates with successful surgical outcomes, the large variability of the HFO rate across patients and over time significantly limits the diagnostic utility of HFO. In this study, we characterize the temporal arrangement of the HFO events over multiple days of continuous intracranial EEG (iEEG) recording and investigate the presence of signature temporal patterns in the SOZ versus the remaining brain areas. Methods: Twenty patients from the iEEG Portal database were included (Table 1) based on the availability of an estimation of the SOZ, the post-surgery score (Engel or ILAE scale) at the 12-months follow-up, and at least 72h of continuous multichannel iEEG recording. The dataset included 10 patients with temporal-lobe epilepsy (TLE) and 10 patients with extra-temporal epilepsy (NTLE), see Table 1. A validated HFO detector [1] was applied and ripple HFO (80-150 Hz) were identified. For each channel, the multi-day iEEG time series was discretized into a binary 0-1 sequence (bin size: 1 second), with 1s indicating bins where a ripple was detected and 0s indicating bins without events. Each sequence was envisioned as the output of a time-varying point process and a history-dependent point process model (PPM) was fitted by using the statistical framework in [2] (Figure 1A). PPM parameters were then used to infer significant non-stationary patterns according to criteria in [3]. REFERENCES: [1] Gliske et al., Clin Neurophysiol, 127:1057-66, 2016. [2] Truccolo et al., J Neurophysiol, 93: 1074-89, 2005. [3] Vyas et al., IEEE Trans Neural Syst Rehabil Eng, 24:36-45, 2016. Results: All PPMs passed the goodness-of-fit test defined in [2] (P-value PP<0.05). In the TLE cases, a recurrent pattern with HFO occurring in small groups of 1-2 events at intervals of 30 seconds or 60 seconds was inferred (Z-test, P<0.05) in SOZ-related channels (Figure 1B). The pattern was identified in all SOZ channels of 8 out of 10 TLE cases and in 90% of the SOZ channels in the remaining 2 cases (average: 10.9±2.8 SOZ channels per patient, mean±SD). In the remaining channels, instead, the HFO occurred in groups of max 2 consecutive events distributed according to a history-independent Poisson process. In the NTLE patients, the HFO distributed as history-independent Poisson processes in all channels, regardless of the SOZ, and no significant difference between SOZ and non-SOZ channels was reported (Figure 1C). Conclusions: A distinctive temporal pattern of the HFO is reported for the SOZ-related channels of TLE patients and statistically validated against data from NTLE patients. The difference in temporal arrangement of HFO events between TLE and NTLE epilepsy may contribute to better characterize the dynamics of the epileptogenic network in the temporal lobe. Funding: Not applicable