Per-pair SEEG spectra repeat across RAM FR1 sessions, but relative theta power misses the planned stability threshold

We measured how well per-pair SEEG spectral features repeat between two recording sessions of the same patient. The analysis used 85 patients with ses-0 and ses-1 from OpenNeuro ds004789 (RAM FR1), version 3.1.0, and 7,353 bipolar pairs matched by label. For each session, we analyzed a 60 s segment from the first word event. Across pairs, median per-patient Spearman correlations were 0.80 for delta, 0.69 for theta, 0.81 for alpha, 0.82 for beta, 0.84 for gamma, 0.87 for aperiodic exponent, and 0.91 for total power. All seven features showed positive between-session agreement (Holm p = 0.0007). Six met the planned stability rule (median r >= 0.7); relative theta power did not. Between-session agreement was lower than agreement between 30 s halves of one session for all seven features (Holm p = 0.0007), while retaining approximately 84-93% of the within-session value. About one third of patients had r < 0.7 for most features. Exploratory analysis found that a pair's session-0 spectrum explained a median 83% (95% CI 75-87%) of within-patient across-pair variance in session-1 1-45 Hz log power. The comparison with position maps from other patients (8-12%) is approximate. The session gap is unknown; sessions are not confirmed to be on different days. Results concern the first 60 s of the same FR1 task state and do not directly apply to rest or sleep.

SEEG · spectral reliability · session-to-session · RAM FR1 · intracranial EEG · power spectrum · OpenNeuro