In 52 NREM-sleep subjects, transient events explain most of the longer high-gamma timescale at epileptic SEEG pairs

This study replicates and tests pilot 40883d0d. Data: OpenNeuro ds007118, ds007119 and ds007120 v1.0.0, NREM sleep, 30 s per subject, 1100 same-shaft bipolar SEEG pairs in 67 subjects. 52 subjects have both normal and abnormal (spiking or seizure onset zone) pairs. The metric is the 1/e lag (tau) of the autocorrelation of the 70-150 Hz envelope. With all samples, tau at abnormal pairs was longer than at normal pairs in the same subject (median ratio x1.35, 95% CI x1.21 to x1.60, 44/52 subjects, Wilcoxon p = 3.9e-6). Abnormal pairs also had more transient events (spikes or high-gamma bursts, robust z > 5). Then the samples within 100 ms of events were excluded from the autocorrelation. After this masking, the ratio decreased to x1.14 (38/52, p = 2.8e-4). With a lower threshold (z > 3.5), the ratio was x1.09 (p = 6.2e-4). The decrease was significant (paired p = 0.0017). Thus most of the pilot effect comes from transient epileptic events, and a small residual effect stays. This study cannot tell if the residual effect comes from the background signal or from events below the threshold. For BCI decoders: at contacts near epileptic tissue, a high-gamma timescale measured on raw data mostly measures the event rate. The package includes code, S3 keys and byte ranges for each segment, per-pair and per-subject tables, statistics and a figure.

SEEG · high-gamma activity · autocorrelation · epilepsy · transient events · event masking · NREM sleep