Abstracts

Rationale: Given the known interaction between high frequency oscillations (HFOs) and sleep, reflected in rate dependences with different sleep stages and sleep microstructure, we hypothesize that the rate of HFOs is further influenced by the circadian rhythm. Methods: A series of 16 patients with refractory focal epilepsy undergoing intracranial investigation (SEEG) were recruited. Scalp EEG, EOG, and EMG leads were used for sleep staging during the SEEG study. An automatic detector was used to identify HFOs during the complete night. Since the amount of data precluded a semi-automated approach, wake and N1 stages were excluded given the higher rate of false positive detections in these stages due to movement artifacts. HFOs were grouped into (i) physiological ripples (80-250Hz) recorded from channels showing no epileptic or non-epileptic abnormalities, outside of the seizure onset zone (SOZ), and outside lesions, (ii) pathological ripples (80-250Hz) and (iii) fast ripples (250-500Hz) recorded from channels in the irritative zone and SOZ. The occurrence of each HFO type was modeled as a Poisson process, with a time varying rate. The rate of HFOs was defined in 30s intervals and models including the sleep stage, average amplitude of slow waves, delta band power, and time were built. A single set of parameters was estimated for the whole patient population as variations with respect to the individual HFO rates. We performed model comparisons using the Akaike Information Criterion, which can be interpreted as an F-test for non-normally distributed data. Results: We found different patterns of variations of HFO rates for physiological ripples and pathological ripples/fast ripples during the night. For physiological ripples the most important variation was an increase in rate of 8%/per hour during rapid eye movement (REM) sleep. For pathological ripples and fast ripples the rate decreased with time during non REM (NREM) sleep, reaching variations of -8%/per hour and -15%/per hour respectively during stage N3. The variations cannot be explained completely as a linear effect of the slow wave amplitude or delta band power, since a model comparison indicated that the inclusion of accumulated time resulted in significant improvements compared to models including sleep stage, slow wave amplitude, and delta power (p< < 0.001 in all cases). Conclusions: This investigation of the circadian rhythmicity of HFO expression over a full night of sleep suggests that HFO rates are under circadian influence. Our findings further point to the existence of different mechanisms in the generation of physiological and pathological ripples. For pathological ripples and fast ripples the decrease of the rate during NREM stages as the night advances suggests an association with sleep pressure, possibly through sleep slow waves, given the known association between the occurrence of HFOs and slow wave amplitude. The increase of the rate of physiological ripples during REM sleep as the night progresses could indicate that physiological ripples are associated with cortical activation mediated by cholinergic drive, since phasic REMs ?" going along with an increased cholinergic drive - are more frequent during REM sleep in later sleep cycles. Funding: CIHR grant FDN-143208.