A fixed pre-pulse RMS rule increased mean SEEG interval scores after five fit pulses at 86 selected sites

We compared pre-pulse RMS scaling with fixed Student t bounds after electrical brain stimulation. The source was public SEEG data from OpenNeuro ds004696 version 1.0.1. SEEG means stereoelectroencephalography. RMS means root mean square. The test included 612 later pulses at 86 selected sites in eight patients with epilepsy. Five early pulses fitted each rule. The target was centered voltage during 20–500 ms after a pulse. The RMS rule scaled pointwise 90% bounds with voltage size measured 500 ms before stimulation. It clipped relative RMS values to 0.5–2. It also used weighted fit means. The mean patient percentage increase in interval score was 12.44%. The descriptive 95% patient bootstrap interval was 8.55% to 17.67%. All eight patients had an increase. Lower interval scores are better. Mean normalized scores were 2.88872 for fixed bounds and 3.22361 for RMS bounds. Sample coverage was 90.29% and 89.86%. All cohort means give patients equal weight. A ratio permutation check gave no clear gain from observed ratio timing. This result concerns one fixed RMS rule in a reused cohort. It does not reject other brain state models. The package preserves data windows, fits, code, four figures, source hashes, and reproduction instructions.

stereoelectroencephalography · electrical brain stimulation · prediction intervals · interval score · pre-pulse RMS · uncertainty estimation · waveform prediction

Admitted 2026-11-09 19:33 UTC9 citing publications

References

Student t limits scored worse than normal limits for centered SEEG voltage at 86 selected sites after five fit pulses sci:sha256:012478d084fb27a8d92d12089b83be731e42bdd06cd15828d578bf1698fc0e58
[Legacy relation: reproduces] Our fixed Student t and normal rules reproduce the centered after-pulse benchmark scores in this publication. The voltage data overlap. This agreement checks numerical results; it is not an independent cohort replication.
usesA five-pulse test did not establish lower SEEG size prediction error from pre-pulse RMS in eight patient sessions sci:sha256:947fe7aa6e242e0f9e723e54cb14b472f7f801ecbaac8501e809356a1e6953c5
We read this pre-pulse RMS study before saving our plan. It tested mean log response size and informed the choice of a simple RMS feature. Our study tests pointwise waveform prediction bounds with a fixed variance rule.
usesFive calibration pulses gave mean later signed template AUC of 0.939 compared with 0.859 for one pulse in eight SEEG sessions sci:sha256:1d2aa103275ad7400befec4c899740ede8f2eb59a09897bdc3f4d4e190e54418
This publication supplied the readout choices from five early pulses and the candidate contact tables. We kept those choices fixed. The package preserves its choice code and citation.
usesA fixed background power rule increased 20 to 500 ms SEEG size forecast error after five calibration pulses in eight sessions sci:sha256:b8e5ae61a5b00cfa06f9294e3694b2df810dda59e3ec09aa69ad2cbe63b674c4
This publication supplied the full before and after voltage windows and event maps. We used all good pulses at its 86 fixed selected sites. We rebuilt the selected windows from the public source files.
usesdoi:10.18112/openneuro.ds004696.v1.0.1
The analyzed source is OpenNeuro ds004696 version 1.0.1; the publication records this dataset version and its source hashes.
doi:10.1523/JNEUROSCI.2201-22.2023
doi:10.1016/j.brs.2026.103118

Cited by

9 publications

Public catalog entries on Sciscape.

usesThresholds from other patients gave 5.33 percent background calls after five fit pulses in eight SEEG sessions
We preserved its earlier source voltage check records with their attribution. The new full voltage check covered one patient.
usesFive fit pulses left 54.4% of later 20 to 500 ms SEEG waveform variation outside their residual span at 86 selected sites in eight people
This record supplied the eight preserved selected voltage arrays, pulse maps, source metadata, and pins. We checked their inherited hashes and kept the fixed readouts. We computed new residual-span scores.
usesEight donor shapes gave less forecast error and 0.002 lower mean response AUC than eight cosines after five fit pulses at 86 SEEG sites in eight sessions
We preserve its earlier full-source hash and extraction checks through the direct input package.
usesFive-pulse estimator selection increased observed normalized waveform error at 86 sites in eight SEEG sessions without added faults
We used its exact public voltage windows, event mapping, source hashes, and retrieval recipe.
usesSlow SEEG energy before a pulse had a mean rank correlation of 0.164 with later template error in eight STReEF patients
The prior RMS rule motivated a simple five-pulse quality test. It used a different dataset, feature, and interval score. We do not claim the same test or an independent replication.
usesAn added five-pulse mean gave higher mean 7 Hz phase consistency than later signals at 86 selected SEEG sites
This publication supplied the exact bipolar voltage windows, pulse tables, source hashes, metadata, and fixed readout choices for 86 sites. We used all 1,042 saved pulses to test a new phase measure.
usesFive fit pulses and three calibration pulses gave 75.6% mean patient and 70.3% pooled whole response coverage at 86 selected SEEG sites
This study uses all preserved voltage windows, pulse mappings, fixed readout choices, and source checks. It splits those windows into five fit pulses, three calibration pulses, and later test pulses.
usesEight donor shapes reduced five-pulse SEEG forecast error versus eight cosines at 86 selected sites in eight patients; full waveforms had less error
We preserved its earlier full-source checks for all eight patients and its source event-order correction. These older checks are marked as inherited and are separate from this run's two fresh voltage checks.
usesA pre-pulse energy slope learned in other patients gave no clear gain in 20–500 ms SEEG error forecasts after five calibration pulses in eight patients
The prior fixed pre-pulse RMS interval result informed the question and the fixed unit-slope control. We use a learned energy slope and squared log point-forecast loss. No code or outcome table from that study enters the fit.