A linear background fit changed the mean alpha contrast from 0.29 to minus 0.23 dB in 21 SEEG patients

We tested whether alpha power change remained above a fitted spectral background in public SEEG data. The study used 35 recordings from 21 patients with epilepsy in OpenNeuro ds006519, version 1.0.0. Two-second windows supplied changes within the pre annotation and from late pre to post. The contrast was post change minus pre change. A line in log frequency at 4 through 30 Hz, excluding alpha, supplied the background estimate. Equal patient weights gave a mean raw contrast of 0.293 dB and a mean adjusted contrast of minus 0.229 dB. The adjusted 95% patient interval was minus 0.742 to 0.234 dB. Both exact primary tests gave a Holm-adjusted p value of 0.789. The tests require independent patients and null contrasts that are symmetric about zero. The data did not establish a mean increase, a mean decrease, or equivalence. Eight planned checks also gave uncertain negative adjusted estimates. All 35 source files passed content-hash checks. All 4,417,600 rebuilt epoch samples matched exactly. The package preserves code, epochs, spectra, results, provenance, and reproduction instructions. The reused cohort, unequal time gaps, and absent matched controls limit interpretation.

SEEG · alpha power · spectral background · Welch power spectral density · cortical stimulation · epilepsy