First-trial SEEG channel maps agree moderately with later maps; beta was not more stable than high-gamma in eight subjects

We asked whether the first 10–100 trials of an offline cued key-press session provide stable bipolar SEEG channel maps, and whether 13–30 Hz (beta) maps are more stable than 70–150 Hz (high-gamma, HG) maps. We used archived per-trial log mean band-power features for rest (−0.75 to −0.25 s) and movement (−0.25 to +0.25 s) from eight subjects in OpenNeuro ds004473 snapshot 1.0.2. For each channel, we calculated the paired effect size of movement minus rest. We compared each map from the first N trials (N = 10, 20, 40, 60, 100) with a reference map from trials 101 onward using Spearman rho, and measured overlap between the eight strongest channels by absolute effect size. The preregistered primary contrast was beta minus HG rho at N = 20. Beta was not more stable: mean difference +0.041, median +0.060, positive in 5 of 8 subjects, exact two-sided sign-flip p = 0.61. No band difference was significant at any N (p = 0.15–0.93; not corrected). Agreement increased with N but remained moderate: at N = 100, median rho was 0.45 for HG and 0.52 for beta. At N = 20, median overlap of the reference top eight was 44% for HG and 25% for beta, versus 7–8% chance. The study has low power with eight subjects. Slow session drift and sampling noise are not separated. These offline results come from one session per subject, epilepsy patients with clinical implants, an adjacent-bipolar montage, and one movement window at the key press.

stereo-EEG · SEEG · brain-computer interface · channel selection · calibration trials · beta band · high-gamma · map stability · OpenNeuro