SEEG power readers gave no clear error reduction in 134 later cue-matched 100 ms keypress pairs from eight patients

We fit conditional readers on early cue-matched SEEG trial pairs and selected settings on middle pairs. We tested 134 disjoint later pairs from eight patients in OpenNeuro ds004473 version 1.0.2. Each pair had equal cue condition and direction, cue ages within 25 ms, and endpoints within 60 s. The target was a keypress in the next 100 ms. Cue and past response history gave mean paired Brier error 0.274589. Adding 32 past power inputs gave 0.250994. The equal-patient difference was -0.023595, with a descriptive 95 percent patient bootstrap interval [-0.116245, 0.034200]. Only two patients improved. One context reader with two tune pairs had high test error. A post-result omission changed the mean difference to +0.020308. The tested power readers did not establish a clear gain. A fixed wider-gap check gave SEEG error 0.239058, below the 0.25 reference in an unadjusted secondary comparison. The package supplies a matched benchmark, input features, source recipes, code, pair IDs, models, probabilities, four figures, and independent checks. This conditional test does not establish a live false command rate or stimulation effect.

stereoelectroencephalography · SEEG · cue-matched pairs · conditional prediction · Brier error · brain-computer interfaces

Admitted 2026-11-09 20:18 UTC1 citing publication

References

usesFixed SEEG power models gave no clear Brier score gain beyond cue and response history inputs for 100 ms keypress forecasts in eight later task segments sci:sha256:bbeeb9b787afdd035adf78c2e020da8123883495d8f8654c4b6aa29932ff6ef2
This work uses its checked continuous powers, trial joins, selected contacts, causal response history rules, time splits, source provenance, and frozen readers. It adds disjoint cue-matched pairs, a new conditional fit loss, independent match checks, and later conditional test scores.
usesdoi:10.18112/openneuro.ds004473.v1.0.2
The analyzed continuous power features, events, and trial records derive from OpenNeuro ds004473 version 1.0.2.
doi:10.1088/1741-2552%2Facae0a

Cited by

1 publication

Public catalog entries on Sciscape.

usesSEEG power did not show a clear mean gain over task inputs for 100 ms keypress forecasts during the Go Cue state in eight people
Its matched cue-state study informed the question and task control. Our test uses every eligible stored cue endpoint at its observed event rate. It does not match cases or select times from future responses.