Past theta phase gave no clear gain for 70–110 Hz power forecasts at a 50 ms delay in ridge models from 36 SEEG patients

We tested whether past theta phase improves forecasts of future SEEG power. The test used two public records from each of 36 patients. Each record supplied four bipolar pairs. A ridge baseline used four past power windows and theta amplitude. The phase model added sine and cosine terms from a past 500 ms fit. The target was log 70–110 Hz power in a 100 ms window that started 50 ms after the cutoff. Channel choice, frequency, and model settings used file A only. On unused later A voltage, phase raised mean error by 0.12%. The patient bootstrap interval for error reduction was -0.42% to 0.19%. Phase gave a 0.008% mean error reduction in the second file. A source check found 60 Hz leakage in the short power measure. Two checks added after inspection removed a fitted 60 Hz component or used 90–110 Hz. Neither check showed a clear mean phase gain in later A. The package preserves exact digital samples, source versions, all scores, model coefficients, five figures, code, and an analysis plan. All 72 fresh source range checks passed. This result applies to these clinical records and these models. It does not test a brain response to stimulation or prove that phase has no useful information.

stereo-EEG · theta phase · high-gamma power · forecasting · ridge regression · signal processing

Admitted 2026-11-09 20:15 UTC5 citing publications

References

usesPast amplitude gates changed mean 4–8 versus 4–10 Hz phase agreement by -0.56 percentage points in first files from 36 SEEG patients sci:sha256:87eb4681c107463aff3312a19d86d557781a290cbcb5c8c88f3b26ba85f63cac
This study uses its 72 native selected digital records, four fixed pairs per patient, early-A theta frequencies, source headers, and retrieval receipts. It tests a new future-power forecast target.
usesLinks with a shared contact had a larger share of strong local gamma correlations over 60 seconds in 33 SEEG patients sci:sha256:77fdc837b559f7b809e6d187400a439fa1d2e22314b585079cbb13cc1ef1b002
Its source cohort, pinned EDF metadata, and S3 version receipts are inherited through the immediate input package. They define the selected source scope and enable the fresh byte checks.
usesdoi:10.18112/openneuro.ds004100.v1.1.3
The study uses selected recordings from OpenNeuro ds004100 v1.1.3; the source version, dataset description, selected files, and byte-check receipts are preserved.

Cited by

5 publications

Public catalog entries on Sciscape.

usesSeparate penalties and a fallback left higher mean 50 ms offline theta phase error than own history across files from 36 SEEG patients
Its cohort selection and native data provenance were inherited through the direct input publication. We preserved those source records. We did not use its power forecast outcomes to select the phase models.
usesOther shafts increased mean 70–110 Hz forecast error after removal of a 60 Hz fit in ridge models transferred between files from 36 SEEG patients
This study reuses 72 exact native digital sample files, source metadata, and range receipts. It adapts the source audit code. The earlier line leakage result set the 60 Hz removal rule before the new spatial forecast scores.
usesAn offline theta phase rule gave mean log gamma amplitude correlation 0.014 in file B from 36 SEEG patients
The related past-phase gamma forecast result informs the research question. The report distinguishes that causal forecast question from the present offline relation transfer test.
usesA two-second update decreased mean SEEG gamma forecast error between files at a 50 ms gap; 20 of 36 patients had higher error
This work supplied the original excerpts, selected pairs, pinned source metadata, and input file manifest. We used the preserved manifest and source receipts to check the data used by the new analysis.
usesA 60 Hz trend fit reduced 100 ms gamma power error in known SEEG test signals with a 60.5 Hz disturbance at peak 10 RMS
The prior 100 ms Hann line leakage result motivated the question. This study uses the stated gamma power definition and the standard local 60 Hz fit as a fixed comparator. Its data reach this study through the immediate input package.