Location-adjusted SEEG exponent and beta showed site effects mostly driven by one site

We tested for a recording-site effect on four unit-free spectral features of SEEG bipolar pairs (aperiodic exponent 2-40 Hz, relative alpha, theta and beta). We used the archived spectra of 256 RAM subjects (ds004789, ds004809). For each pair, we subtracted the mean of its 30 nearest MNI pairs from other sites. Then we took the median per subject. A pre-registered Freedman-Lane permutation F test (covariates age, sex, dataset and number of pairs; 8 sites with 5 or more subjects; n = 250) found a site effect for the exponent (F = 3.98, Holm p = 0.004, partial R2 = 0.105) and for relative beta (F = 3.80, Holm p = 0.004, partial R2 = 0.101). Age explained less than 1% of the variance of these features. The effect stayed when we added log sampling rate as a covariate and when we used only the 1000 Hz recordings. It was not significant for alpha or theta. An exploratory drop-one-site check showed that the effect was mostly from site J (exponent +0.26 above the location expectation). Without J, p = 0.33. Pooled SEEG normative maps must include a site term or a check of this site.

SEEG · recording site · spectral features · aperiodic exponent · relative beta power · RAM dataset