In 112 RAM patients, white-matter SEEG pairs had lower broadband power than gray-matter pairs, but power alone poorly identified tissue

In archived stereo-EEG (SEEG) spectra from 112 patients in the RAM free-recall datasets, bipolar pairs with both contacts labeled cerebral white matter (WM) had 0.489 log10 (4.9 dB) lower median 2–45 Hz power than pairs with both contacts labeled neocortical gray matter (GM) in the same patient (95% CI 0.421–0.557; 102/112 patients). The difference persisted for pairs on the same shaft and after matching WM pairs to the nearest GM pair within 15 mm. A within-patient classifier based on power had mean AUC 0.710, while a leave-one-patient-out threshold had 0.667 balanced accuracy; WM pairs were not silent and power alone did not replace anatomical labels. GM pairs had a steeper 2–40 Hz aperiodic exponent and lower relative alpha and beta power; relative theta did not differ. Mixed pairs with one GM contact were near GM pairs in broadband power. The analysis used 60 s from the first word event, 1–45 Hz spectra, and automatic monopolar contact atlas labels. The planned label column was unavailable for the intended classification; the analysis used the whole-brain atlas column before calculating statistics. Four subjects whose electrode files no longer existed at their source URLs were excluded. Limits include atlas and contact localization error, selection of subjects with whole-brain labels, epilepsy patients, a task segment, no high-gamma analysis, and no distance-to-GM measure.

stereo-EEG · SEEG · white matter · gray matter · bipolar pairs · power spectrum · RAM · free recall · brain tissue · intracranial electrophysiology