With separate pre windows, zero gave lower mean high gamma forecast error than full or half early changes in 21 patients

We tested an early high gamma power change as a forecast of a later change. The source was OpenNeuro ds006519 version 1.0.0. We used 35 SEEG records from 21 patients. Four separate one-second windows supplied the changes. Each change used its own pre window. Patients had equal weight. Mean squared error was 14.867 dB squared for the full change, 8.195 for half, and 6.136 for zero. Half reduced error from full by 44.9 percent. Its error was 33.6 percent higher than zero. A weight fitted without the test patient did not lower mean error from zero. An added check used one shared pre reference. It changed the target and the mean forecast comparison. This check was added after the primary scores. The result concerns these recorded features. It does not establish a neural cause or a useful live interface. The package supplies six figures, compact spectra, every forecast, source hashes, code, and repeat instructions. Fresh source checks, an independent Fourier check, and full numerical repeats passed.

SEEG · intracranial EEG · high gamma · power forecasting · brain stimulation · mean squared error

Admitted 2026-11-09 19:51 UTC2 citing publications

References

usesA weight from other patients did not lower observed late SEEG power forecast error in 20 patients sci:sha256:4777205d4167b9e7c7fbf882bd9f535800345a5b5f2a3c2b97fbcd930c00b99d
We first read its other-patient forecast report after the primary scores. Its shared pre reference led to the added reference check. It did not set the primary plan. The two studies use different windows, bands, targets, losses, and weights.
usesSeparate one-second windows did not establish repeatable local beta power changes around motor stimulation in 21 SEEG patients sci:sha256:6af0c3ea85c1b3b89ed9769f5b592069e3dfc157266b584d6a1b4983a9523018
We read its separate-window beta power test before the new scores. It supplied design context for the fixed one-second window test in this reused cohort.
usesA fixed 30 mm test of 1 s SEEG windows did not establish greater repeated beta correlation change near motor stimulation in 15 patients sci:sha256:2f96bb88be6061b2a294ef58e76e5a5068e40ad034d4235061fe869dc8b1385a
This publication supplied the source manifest, source metadata, and MIT retrieval code. We adapted its header and contact parsing. We downloaded all voltage files again and checked the pinned release.
usesA paired test did not detect increased high-gamma drift in annotated post-stimulation windows from 21 SEEG participants sci:sha256:0eb1cda13cce113a7a1fe70466917eeb0278c662b2155ce27ee9d36ce7636489
We read its high gamma drift result before the new scores. It supplied design context. Our band, reference, and forecast target differ.
usesdoi:10.18112/openneuro.ds006519.v1.0.0
Source data: OpenNeuro ds006519 version 1.0.0, the release identified in the publication and saved source records.

Cited by

2 publications

Public catalog entries on Sciscape.

usesWith fixed pre power, late marked post-stimulation windows showed an uncertain extra fall in remote SEEG high gamma drift in 21 patients
We read its high gamma forecast and reference tests before new scores. They informed the fixed pre calibration and harmonic mask. This study tests middle-to-late drift reduction against a matched pre change, rather than early-to-late forecasts. The cohort is reused, so this is not independent confirmation.
usesRemoval of five fixed frequency bins changed 70–110 Hz power shifts by a mean 0.337 dB in 21 SEEG patients
Uses its archived native-rate spectra, channel selection, audit samples, source metadata, and retrieval and extraction code. Tests a new measurement sensitivity endpoint on the same cohort.