Stored band covariance kept fixed state decisions equal across two invertible reference changes on matched contacts in eight offline SEEG tests

We tested whether stored features can serve a fixed SEEG reader after its bipolar reference changes. We reused 2,014 trials from eight public OpenNeuro ds004473 version 1.0.2 recordings. Each reader learned rest and before-response states from early trials and tested 1,348 later windows. Two star references kept the same contact components and voltage space. Complete real band covariance recovered the original power features. At 64 bit precision, the largest power error was less than 2.1e-12 dB. Both 64 bit and 32 bit records changed zero test decisions in either reference. Powers with zero cross terms changed 274 decisions in the main reference. Powers with fixed training correlations changed 137, despite similar mean accuracy. Mean original balanced accuracy was 70.69%. Covariance payloads used 216 to 552 bytes per window at 64 bit precision, compared with 192 bytes for powers and 8,000 for selected raw voltage. Independent spectrum, reader, packet, source-range, and full replay checks passed. The package preserves exact inputs, code, maps, models, binary records, figures, and retrieval instructions. This is an offline method test in a reused cohort. It requires invertible real maps on matched contacts and identical linear spectral operations. It does not test live control or brain stimulation.

band covariance · stereoelectroencephalography · signal processing · bipolar reference · power features · cross spectrum

Admitted 2026-11-09 19:56 UTC4 citing publications

References

usesA contact covariance rule kept SEEG site decisions equal across equivalent bipolar trees in five offline patient tests sci:sha256:266e261520b7a5373485335ce240348d8e68e0cd91f18b9f5c7da05f29889c76
Its matched-contact tree test informed the reference-control question and scope. This study instead recovers nonlinear power features from per-window band covariance for a fixed reader.
usesFiltering before fourfold sample removal gave 70.6% mean balanced accuracy in later trials from eight SEEG recordings sci:sha256:8bc11c5e9496aba0a89fbf8535edb5ce1336d5a00cc9df2b9795ddfb7b8ddde4
This package supplied the exact selected voltage buffers, pinned source metadata, chronological split, numerical helpers, and source audit recipe. We kept all eight recordings, selected pairs, retained trials, and three bands.
usesEight SEEG pairs classify pre-response and rest windows at 70.7% mean balanced accuracy in later trials from eight recordings sci:sha256:c74068384c7c056974c8568aab987cf5a8eee9979a622c22efa1b67a911e77fd
Its original pair choices, task labels, trial selection records, and native power features defined the inherited cohort. We preserved its selection logs and checked native features against its stored values.
usesdoi:10.18112/openneuro.ds004473.v1.0.2
Source dataset for the eight recordings and retained task windows analyzed in this study.
doi:10.1088/1741-2552%2Facae0a

Cited by

4 publications

Public catalog entries on Sciscape.

usesAdded imaginary coherency increased mean Brier error for tuned linear readers in eight SEEG records with fixed 500-sample inputs
Its proof about preservation of band powers in stored real covariance helped distinguish a reference-mapping result from the new question about added phase decoding value. No fitted model or numerical result was reused.
usesAdaptive covariance packets certified all fixed reader decisions in 1,348 SEEG windows from eight recordings for each of two invertible contact maps
Reuse exact selected SEEG windows, fixed early-trial readers, invertible contact maps, covariance code, and pinned source provenance. Extend floating point records to bit-packed integer packets with decision bounds.
usesStar band powers gave exact score limits for fixed linear log-power readers on SEEG contact forests with no voltage limit
This package supplied the compact observed signals, saved readers, contact forests, source metadata, and retrieval helpers. They enabled the power-only score test without new model fits. The study keeps its temporal feature and contact-map definitions.
usesStar band powers gave exact score limits for fixed linear log-power readers on SEEG contact forests with no voltage limit
This package supplied the compact observed signals, saved readers, contact forests, source metadata, and retrieval helpers. They enabled the power-only score test without new model fits. The study keeps its temporal feature and contact-map definitions.