In 7 SEEG subjects, white-matter channels had 2–3× less broadband power than gray-matter channels, but the 1/f exponent did not separate them

We asked whether label-free resting features of bipolar stereo-electroencephalography (SEEG) channels distinguish gray matter (GM) from white matter (WM). We used 5-s prestimulation windows from 35 recordings of 21 subjects in OpenNeuro ds006519. In the strict GM-versus-WM contrast, 7 subjects had at least two channels in each class. Within-recording z-scored power at 1–30 Hz and 70–145 Hz had median subject-level AUCs of 0.79 and 0.77, respectively; both were above 0.5 in all 7 subjects (Wilcoxon p = 0.016, the smallest possible exact p for n = 7). Variance had median AUC 0.74. The 1/f exponent did not separate the classes (2–40 Hz: AUC 0.61, p = 0.58; 20–45 Hz: AUC 0.47, p = 0.22), and contact position had AUC 0.55. The spectra showed an approximately 0.3–0.5 log10-unit (2–3×) broadband power offset from 1 to 200 Hz, not a slope change. In leave-one-subject-out evaluation, 70–145 Hz power alone had median held-out AUC 0.77, matching the all-spectral model; the top 25% of ranked channels were 96.8% GM versus a base rate of 89.2%. When mixed GM/WM channels were included with WM, 70–145 Hz power gave AUC 0.63 across 13 subjects (11/13 above 0.5, p = 0.04). Interpretation is limited by the 7-subject strict contrast, 5 s of rest per session, author-supplied tissue labels that were not independently checked, bipolar montage, and the rarity of WM channels (103/1,754).

stereo-electroencephalography · SEEG · gray matter · white matter · broadband power · 1/f exponent · spectral features · tissue classification · OpenNeuro · BCI