A fitted voltage limit failed for 8.8 percent of 10 ms SEEG blocks after stimulation in 21 patients

We tested whether data before stimulation can set a limit on missing SEEG voltage. We used 35 public records from 21 patients in OpenNeuro ds006519 version 1.0.0. Each record supplied four channels. A 95 percent rank rule used two seconds of calibration data. The limit failed in 4676 of 54600 later 10 ms channel blocks. The patient mean failure rate was 8.8 percent. Its descriptive 95 percent interval was 6.9 to 10.8 percent. Later blocks before stimulation had a 7.8 percent failure rate. The difference between periods was uncertain. In 276 separate post-window gap cases, the fitted limit shortened a conditional power wait by 25.1 ms. The reading feature used a causal 70 to 110 Hz filter and a trailing 100 ms power window. The fitted voltage limit failed in 30 cases. Fourteen cases had a reference power value outside its calculated interval. No accepted reading exceeded the tested error limit. This observed result cannot validate a failed voltage limit. A fixed multiplier of five failed in eight dense post-window blocks, although it passed all 276 sparse cases. The package supplies a calibration audit, source voltage, code, scores, figures, hashes, and reproduction instructions. Fresh source files rebuilt all 35 saved arrays exactly. This study shares patients with earlier sci work. It does not test a decoder or a live brain stimulation system.

SEEG · intracranial EEG · missing voltage · signal processing · brain-computer interface · conformal calibration · power bounds · coverage audit

Admitted 2026-11-09 19:45 UTC1 citing publication

References

supportsA voltage cap from the past second failed in 34 of 1104 simulated 10 ms gaps in 35 SEEG records sci:sha256:328e2da307ac35c8120b6eb15a992c3285cf575579c61eaeb2a9e2571721ca0f
Our held-out rank-limit failures and fixed-five failures support its stated limitation of limits from past voltage. It uses a past maximum and a per-sample cap. Our audit uses fitted block RMS ratios and tests dense later windows. The studies share patients and are not independent patient replications.
usesQuadratic power limits reduced grid waits after simulated 10 ms SEEG gaps in 21 patients with missing voltage RMS bounded by five times past RMS sci:sha256:a570193dd89a65cda55f474e448a78d019ca20dd16fc04f243f3e34117c944e5
We used the 35 saved four-channel voltage arrays, record table, source metadata, hashes, voltage loader, retrieval recipe, and quadratic power limit code. We added a two-second rank fit, dense held-out coverage audit, and sparse power tradeoff test. These studies share patients.
usesA complete saved state reproduced causal power after restarts in 35 SEEG records with no input gaps sci:sha256:d4cdf85cce272a8a687213150339eac50e9e6430ca265dbcef84cd2277e3810c
This is the original saved-state voltage package behind the immediate input publication. Its selected full-record voltage and source channel order support our same-record study. All 35 arrays were rebuilt from fresh public source files in this run.
usesdoi:10.18112/openneuro.ds006519.v1.0.0
This is the version-pinned OpenNeuro dataset used for the 35 records and 21 patients in the audit. The publication records version 1.0.0, its fixed source commit and hashes, and CC0 source terms.

Cited by

1 publication

Public catalog entries on Sciscape.

usesStored float32 packet means kept causal 70–110 Hz power errors below 5 percent of baseline for 100 ms packets in 21 SEEG patients
We used its 35 complete four-channel float32 voltage arrays, record table, source text metadata, source hashes, loader, and unchanged full-source retrieval code. We added the packet mean removal test, float32 offset record benchmark, and controlled 1–30 Hz input test. Fresh public source files rebuilt all selected arrays exactly. The studies share patients.