One ripple adds little short-term predictability beyond local rate in interictal NREM SEEG

In author-marked high-frequency oscillation events from 20 patients with epilepsy, we asked whether a detected ripple (80–250 Hz) predicts another ripple on the same bipolar SEEG channel within 0.1–1.0 s beyond a ±5 s local-rate null, and whether ripples co-occur across channels. In run-01, median same-channel probability was 0.71 versus 0.65 for the jitter null; median enrichment was 1.05 (range 1.02–1.22), above 1 in 20/20 subjects (exact Wilcoxon p = 8.8e-5). In run-02, the corresponding values were 0.70 versus 0.67 and 1.04 (0.79–1.12), above 1 in 17/18 (p = 0.0033). Fast-ripple enrichment was 1.06 in run-01 (15/20 above 1, p = 0.021) and 1.04 in run-02 (10/17, p = 0.38). Across channels in run-01, median co-occurrence probability within ±50 ms was 0.13 versus 0.10 for the null (p = 1.9e-6); the median fraction of channel pairs with more than 2x enrichment was 0.19, mostly among neighboring pairs on the same estimated shaft. These findings indicate modest added short-term predictability on the same channel and limited sharing across channels. They concern detector marks from 5-min interictal NREM runs only; raw signals were not analyzed, and detector dead time or event merging may explain the dip below 0.1 s. The exact release represented by the unversioned event-file URLs is uncertain.

SEEG · ripple · high-frequency oscillations · NREM sleep · closed-loop stimulation · event predictability · epilepsy