RAM SEEG power offsets track recording system, while EDF units remain uV

We examined whether EDF physical-dimension labels or per-bit gain explain a broadband 2–45 Hz power offset between sampling-rate classes in 250 patients from the RAM free-recall datasets. All SEEG channels in the 250 recordings were labeled uV, so a unit-label mismatch did not explain the offset; the mean within-center higher-minus-lower sampling-rate effect remained −0.43 log10 units (centers C, D, M, P: −0.31, +0.63, −0.98, −1.06; permutation p = 0.0016). Per-bit gain separated sampling-rate classes completely in centers C, M, and P (AUC 0) and overlapped in D (AUC 0.65). In an exploratory comparison among 1000 Hz recordings, Medtronic files had 0.83 log10 more power than Blackrock files on average across six centers (about 6.7-fold; permutation p = 0.0001); power correlated with log10 gain (r = 0.30, n = 212, p = 0.0001). These findings indicate that sampling-rate class marks recording-system differences, but do not establish their cause. Manufacturer labels are self-reported, and patient groups, recording time, and electrode types may also differ. Treat recording system as a batch variable in absolute-power analyses or use within-patient relative measures.

SEEG · RAM · EDF · sampling rate · recording system · power spectrum · unit calibration · intracranial EEG · free recall