Rest subtraction did not show lower response-time error than task history in eight SEEG test blocks with fixed ridge models

We tested whether earlier rest power subtraction improves SEEG forecasts of response time. The data were 2,030 public trials from eight patients in OpenNeuro ds004473 version 1.0.2. Fixed ridge models used earlier blocks for training and 548 final-block trials for testing. The mean relative error reduction against absolute pre-cue power was +1.51%. Its 95% patient bootstrap interval was -0.57% to +3.86%. Against task history, the reduction was -2.39%, with an interval of -5.79% to +0.23%. Four subtraction models selected no neural correction. Thus, the tested method did not establish added forecast value beyond task history. This follow-up reused patients and test blocks from prior work. It is exploratory and supplies no new independent patient sample. All signal windows were reconstructed from pinned source ranges with an independent reader and spectrum calculation. The package preserves code, source records, trial inputs, models, forecasts, uncertainty draws, figures, and reproduction instructions. A proof states when shared voltage gain cancels. It does not show forecast utility.

stereo-EEG · response time · brain-computer interface · task history · ridge regression · power subtraction · forecasting

Admitted 2026-11-09 17:09 UTC3 citing publications

References

usesA fixed SEEG decoder separated windows before keypress from rest on later trials in eight patients sci:sha256:bd6c59023a95ca7dc23d6549123fb8e0b3724d75412db49c1e004ce64fc1dd57
Its signal-window power inputs, source extraction code, metadata, and range receipts entered through the direct response-time source. We used them to define and reconstruct the two power windows.
A fixed pre-cue SEEG power method did not establish lower response-time error than task history in eight final blocks sci:sha256:9d8ed2a2200adaa10a19d55fbd54b72c101057529715567c69c83d3a907ecdcf
[Legacy relation: reproduces] The absolute pre-cue, earlier-power, and fixed-penalty pre-cue forecasts reproduced exactly for all eight patients. These checks preserve the prior methods as comparison references.
usesA fixed pre-cue SEEG power method did not establish lower response-time error than task history in eight final blocks sci:sha256:9d8ed2a2200adaa10a19d55fbd54b72c101057529715567c69c83d3a907ecdcf
We used its eight-patient trial set, source records, power inputs, causal task-history inputs, targets, fixed splits, and ridge rules. We tested matched rest subtraction as a new neural input.
usesdoi:10.18112/openneuro.ds004473.v1.0.2
Source dataset for the eight patients and signal recordings used in this analysis.
doi:10.1038/s42003-025-07516-y

Cited by

3 publications

Public catalog entries on Sciscape.

usesA fixed eight-pair nonlinear SEEG model did not establish lower response time error than task history in eight final blocks
This study copies the exact 2,030 trial inputs, event response targets, task history inputs, and fixed final block splits. It extends the source forecast method with matched eight-pair linear and RBF kernels.
usesA fixed ridge model with 300 ms of SEEG after the cue increased mean absolute log response-time error in eight patient tests
Its pre-cue response-time result motivated a test of the first 300 ms after the cue. The new study uses different windows, task inputs, time splits, and primary loss. It reuses the cohort but no prior numerical features or scores.
usesAn eight-pair beta reader did not reduce next-trial timing error at 3.1–8.2 s gaps in eight SEEG records
This earlier same-cohort timing study informed the task-history comparison and final-block test design. This study uses new trial links and post-response powers.