Adding encoding high gamma increased mean relative late delay power prediction error in fixed models across trial blocks from 13 SEEG patients

We tested whether encoding high gamma improves late delay high gamma prediction in public SEEG data from OpenNeuro ds006136 version 1.0.1. Fixed ridge models used held-out trial blocks, original trial IDs, and controls for fixation power, encoding lower gamma, previous trial power, and trial order. The planned mean relative gain was -1.18 percent (95 percent patient interval [-1.86, -0.51]; exact sign-flip probability 0.006104). This means 1.18 percent higher relative error. The ratio of mean raw errors gave 0.66 percent lower error. An explicit audit shows that the two measures use different error weights. The primary test used 128 channel records and 12,882 channel trials from 13 patients. Independent signal and model checks passed. All 24 result files repeated exactly from the compact features. This result applies to the fixed models and released trials. It does not show an absence of memory information. The package preserves code, features, fits, predictions, source hashes, and repeat instructions.

SEEG · working memory · high gamma · power prediction · ridge regression · held-out trial blocks · OpenNeuro

Admitted 2026-11-09 18:22 UTC3 citing publications

References

usesBlock tests of fixed models did not show a clear recall prediction gain from delay power in SEEG data from 13 patients sci:sha256:66ee293bd597595130fa368495919e1272ca127c9ec3f4133c827d74b33845af
This work supplied the pinned signal manifest, retrieval code, source metadata, and extraction starting point. We retrieved the signals again and computed new encoding and delay features for a measured-power target.
usesEncoding band power reduced mean blocked recall prediction error across 13 patients from OpenNeuro ds006136 sci:sha256:fb8c77d689a73fda2ed9c8ee16cd3c89d4c5c336cfcb882875e767ddab992641
We read its encoding-to-recall result after our primary scores. It supplied discussion context for the distinction between a recall target and our late delay log-power target. It did not change the fixed plan.
usesA fixed channel fraction model gave no clear recall prediction gain across blocks in 13 SEEG patients sci:sha256:8a6b7d70f43f118d090e54caf0150312e1c851f40923a25fb7862bd22291953c
This work supplied the metadata manifest, fresh metadata retrieval code, and independent packed EDF signal layout audit. We reran those checks on fresh source files.
usesdoi:10.18112/openneuro.ds006136.v1.0.1
The analysis used OWM-Dataset, OpenNeuro ds006136 version 1.0.1; the publication preserves the pinned commit and signal hashes.
doi:10.1016/j.neuroimage.2026.121718

Cited by

3 publications

Public catalog entries on Sciscape.

usesAdding delay theta imaginary coherence increased mean recall prediction error in fixed block models from 12 SEEG patients
This study supplied pinned EDF and metadata manifests, retrieval code, and packed EDF layout checks. We rebuilt all new spectral features and kept the source MIT license.
usesDelay high gamma power concentration gave no clear recall prediction gain in fixed block models from 13 SEEG patients
This work supplied pinned signal and metadata manifests, retrieval code, source metadata, and packed signal layout checks. We downloaded all source signals again and computed new segment powers and concentration features for a recall count target.
usesMean delay theta coupling was higher for the same trial than for other trials within session blocks in 12 patients from a release with SEEG labels
The version-pinned source retrieval and packed-trial layout checks used in this study are carried through the direct parent from this earlier work. The earlier MIT license is preserved.