In 5 new FR3 SEEG subjects, HFA power nearly recovered within 0.2 s after offset, while causal theta power differed on 28% of pairs at 0.5–1.5 s

This replication and extension analyzes five FR3 SEEG subjects not used in the source study, using 0.5 s closed-loop stimulation bursts and the same epoching, control matching, channel removal, zero-phase band-power features, and Welch tests. We added a causal pipeline with forward-only Butterworth filtering, squared signal, and 100 ms bin means. In all five subjects, at most 2.8% of non-stimulated pairs differed in 70–150 Hz power 100–200 ms after offset with either pipeline, near the 1–11% pre-stimulation range. With the causal pipeline, the median fraction differing in 3–8 Hz power was 87% at 100–200 ms and 28% at 0.5–1.5 s after offset; theta remained above 5% through 1.3–1.5 s in four subjects. Beta power recovered to 5% or fewer differing pairs by 0.3–0.4 s in four of five subjects with both pipelines. Zero-phase filtering showed pre-onset theta differences in 37% of pairs (up to 66%), compared with 0.1–1.3% using the causal pipeline, consistent with backward filter spread and earlier stimulation rather than a pre-trigger brain-state difference. The analysis includes five subjects from session 0; stimulation amplitudes vary and are pooled in three sessions; stimulation was closed-loop; tests used p<0.001 per pair without correction across bins; and the 5% recovery threshold is arbitrary. The causal squared-signal feature has group delay, especially for theta, and other causal designs can give different recovery times. The analysis does not interpret times after +2.0 s from onset because the next stimulation contaminates that window.

SEEG · closed-loop stimulation · band power · causal filtering · zero-phase filtering · brain-computer interface · theta · high-frequency activity · replication