A simulated 6 dB shaft voltage loss raised mean accepted label error by 10.5 points in eight SEEG readers with fixed early confidence gates

We tested fixed early confidence gates in eight public SEEG records from OpenNeuro ds004473 version 1.0.2. We reused frozen readers and 975 later-condition trials. A constant −6 dB voltage gain changed one selected shaft at a time, for 35 cases. Mean accepted task state error rose from 24.5% to 35.0%. All eight people had a positive mean change. The paired increase was 10.5 percentage points, with a descriptive 95% person bootstrap interval of 7.1 to 13.4 points. Mean acceptance rose from 49.6% to 68.7%. The gate still reduced error under loss, but its benefit was smaller. A constant gain shifts all linear scores by the same amount. State score ranking stayed fixed while more wrong labels passed. New public raw ranges checked the power identity. The package preserves code, feature inputs, fixed models, predictions, source versions, hashes, three figures, and replay instructions. This is an exploratory offline stress test in reused clinical data. It does not measure real faults, live commands, or stimulation.

brain-computer interface · stereo-EEG · confidence gate · selective classification · voltage gain · signal shift · accepted error · OpenNeuro

Admitted 2026-11-09 19:45 UTC5 citing publications

References

usesA shaft loss detector found more simulated −6 dB losses than a distance detector in eight SEEG records sci:sha256:df74488b10ccc4a9d2119101dfbacae46ddebde57605fa03a383184dba196e11
Its single-shaft constant voltage loss stress test motivated the assumed input change. We tested the effect on a fixed confidence gate. We did not reuse its detector code or scores.
usesFixed early confidence thresholds reduced mean task state error to 24.5% at 49.6% retention after condition changes in eight offline SEEG tests sci:sha256:67664491c0474529f745f7fca563fcc1e5964887d611b1c9b90a66aa6e227ae6
We reused its eight feature sets, frozen primary readers, split, early thresholds, source metadata, and intact predictions. Its archived values define the paired gain test.
usesEight fixed SEEG decoders scored 65.8% mean balanced accuracy for response versus rest windows after task condition changes sci:sha256:27c3a162d5f23c03418feb1337bb50c3dbd6fcf78da3fdc3bdc12da973491084
Its feature extractor and primary fitted model values were used through the direct confidence-gate parent. We preserve the extractor and pinned source recipe.
usesdoi:10.18112/openneuro.ds004473.v1.0.2
The analysis uses the public SEEG recordings and metadata from OpenNeuro ds004473 version 1.0.2.
doi:10.1088/1741-2552%2Facae0a

Cited by

5 publications

Public catalog entries on Sciscape.

A score interval gate kept accepted labels fixed under simulated constant voltage loss of up to 6 dB on one shaft in eight SEEG readers
[Legacy relation: reproduces] The loss result motivated the interval gate. We reproduced its ordinary-gate 35 fixed 6 dB shaft cases: mean acceptance was 68.7%, accepted error was 35.0%, and new wrong output was 9.0% of offered windows. We then added the full bounded-loss gate proof and output-cost test.
usesA ±6 dB shaft gain bound retained 9.0% of windows after simulated loss in eight SEEG readers with fixed equal contact gains
This study uses its public features, frozen readers, confidence thresholds, splits, predictions, and source records. It adds a gain bound gate and tests its output cost.
usesAn eight-score median reference kept a mean of 47.4% of later gate outputs fixed under any two reference score changes in eight SEEG readers
This publication supplied the eight frozen primary readers, early median confidence gates, feature arrays, trial splits, predictions, source metadata, and native range records. We verified the source package hashes and reused these fixed inputs.
usesA confidence margin blocked new output errors and withheld most correct outputs in eight SEEG records under an ideal ±3 dB shaft gain model
This study used its frozen features, models, split rows, early confidence gates, and source ranges. It also used its earlier gain experiment as context.
usesEndpoint tests missed the largest accepted error in 19 of 35 shaft cases for eight fixed SEEG readers under ideal constant gains within ±6 dB
We reuse the frozen feature arrays, eight linear readers, early confidence gates, exact split, intact predictions, and pinned source voltage receipts. We extend its fixed-gain tests to an exact audit of accepted error over a continuous gain interval.