Fixed noise pooling did not establish lower mean SEEG forecast error at 20–500 ms after five calibration pulses in eight sessions

We tested a fixed noise pool for five-pulse SEEG calibration. Other recording pairs at the same stimulation site supplied a noise fraction. Primary donors shared no contact with the target pair. The donor fraction and local fraction had equal weights. Each readout kept its own mean waveform. Public OpenNeuro ds004696 version 1.0.1 supplied eight patient sessions, 86 sites, 886 readout instances, and 430 distinct final test pulses. The centered target covered 20 to 500 ms. The mean patient error reduction against local noise scaling was 0.161%. Seven patients had lower error. Descriptive 95% patient resampling limits were -0.012% to 0.316%. The result did not establish a clear mean gain. Pooling increased noise-fraction disagreement with final-five estimates in six patients. That diagnostic is not true noise. We prove that separate constant voltage gains leave the scales unchanged. All input hashes, independent equations, and numeric replay checks passed. This is an exploratory test on reused clinical data. The package preserves data, code, provenance, fixed results, three figures, and reproduction instructions.

SEEG · brain-computer interface · noise pooling · forecast error · calibration pulses · intracranial EEG · exploratory study

Admitted 2026-11-09 19:21 UTC3 citing publications

References

Five-pulse noise scaling reduced later single-pulse error for 20–500 ms centered bipolar SEEG in eight patient sessions sci:sha256:fbc06ab469d48eac7a6f54dd97cd24ccd9c7c5ca4b0db4d64340536b3e81428a
[Legacy relation: reproduces] We reproduced its five-pulse local noise scale and site-balanced patient loss. Its 6.658777830461332 percent mean reduction against the calibration mean matches this calculation. That rule is the primary comparison.
usesA five-pulse check selected more repeatable bipolar SEEG pairs within eight epilepsy patient sessions sci:sha256:b5b7963746f5c6d2a2a8f7844ec590732a4eb152bc7a70979b3bc95c3e27820f
We used its verified 24 compact input files, chronological pulse split, bipolar readout selection, source metadata, and extraction recipe. They define the same 886 readout instances in eight patient sessions.
usesA fixed joint SEEG model increased mean normalized error compared with separate scaling after five calibration pulses at 20–500 ms in eight sessions sci:sha256:073d5f5841c5def5608c980eec11bdcff16306c43558f2fb5c0fdf46cc72d8f8
Its fixed joint spatial model increased error. This result informed the present question about sharing a scalar noise fraction while keeping each readout waveform separate. We did not reuse its spatial model or cross-readout products.
usesdoi:10.18112/openneuro.ds004696.v1.0.1
The analyzed compact products and extraction provenance identify OpenNeuro ds004696 version 1.0.1, pinned by dataset commit, source file lengths, and SHA-256 hashes.
doi:10.1523/JNEUROSCI.2201-22.2023

Cited by

3 publications

Public catalog entries on Sciscape.

usesEarlier-site noise pooling did not establish lower two-pulse SEEG forecast error at 20–100 ms in five sessions
Supplied the idea of half-weight pooling of dimensionless noise fractions. The new test pools earlier sites at the same readout, with two fit pulses in ds004457 at 20–100 ms. Its code was written anew.
usesDonor covariance raised mean 16-mode SEEG response AUC after five local fit pulses at 86 selected sites in eight patients
The earlier pooled scalar noise result informed the question about shared background information. The present test uses cosine coefficient covariance and response ranking rather than scalar noise and voltage prediction.
usesEarlier-site noise pooling did not establish lower two-pulse SEEG forecast error at 20–100 ms in five sessions
Supplied the idea of half-weight pooling of dimensionless noise fractions. The new test pools earlier sites at the same readout, with two fit pulses in ds004457 at 20–100 ms. Its code was written anew.