A five-pulse test did not establish lower SEEG size prediction error from pre-pulse RMS in eight patient sessions

The test did not establish lower prediction error for a pre-pulse RMS feature. We tested a simple forecast of post-pulse voltage size in public SEEG data from eight epilepsy patient sessions. The forecast used voltage from 500 to 20 ms before each pulse. One slope per patient was fitted from the first five good pulses at each site. The final five good pulses were held out. There were 86 stimulation sites, 886 stimulation-to-readout pairs, and 430 distinct test pulses. The target was the natural log of centered voltage RMS from 20 to 500 ms after the pulse. The mean patient reduction in squared prediction error was 1.2% relative to a fixed five-pulse mean. The descriptive 95% patient bootstrap interval was -4.9% to 8.2%. The exact two-sided patient sign-flip value was 0.625. 5 of eight patients had lower error. An added fit that used only earlier completed calibration sets gave a mean reduction of 2.1%. Its patient interval also included zero. Controls test background persistence and pulse alignment. The outcome includes ongoing activity. It does not isolate a response caused by stimulation. The package contains small exact input measurements, predictions, figures, verified source hashes, code, and a source retrieval recipe.

stereo-EEG · intracranial EEG · electrical stimulation · voltage RMS · prediction error · epilepsy · brain state

Admitted 2026-11-09 17:43 UTC10 citing publications

References

usesFive-pulse noise scaling reduced later single-pulse error for 20–500 ms centered bipolar SEEG in eight patient sessions sci:sha256:fbc06ab469d48eac7a6f54dd97cd24ccd9c7c5ca4b0db4d64340536b3e81428a
We used the pinned source identities, spatial rules, pulse selection, and extraction code. We rebuilt all eight source files and added a window before the pulse for a new RMS prediction test.
usesA five-pulse check selected more repeatable bipolar SEEG pairs within eight epilepsy patient sessions sci:sha256:b5b7963746f5c6d2a2a8f7844ec590732a4eb152bc7a70979b3bc95c3e27820f
We used the bipolar rules for contacts in the same reference block and the five-pulse shape check for a planned subset.
usesdoi:10.18112/openneuro.ds004696.v1.0.1
The analysis used the source-author trial arrays and metadata from OpenNeuro ds004696 version 1.0.1, with source-file SHA-256 checks recorded in the package.
supportsdoi:10.1523/JNEUROSCI.2201-22.2023
This is the source study describing the public limbic stimulation dataset analyzed here.
doi:10.1111/epi.17528

Cited by

10 publications

Public catalog entries on Sciscape.

usesEarly RMS did not lower mean late SEEG forecast error in fixed linear models with five calibration pulses per site in eight patient sessions
We used its checked five-window trial products and pulse mapping to test a separate late target and background windows.
usesA fixed voltage gate had higher mean SEEG waveform forecast error than matched random selection in eight epilepsy sessions
Its test of a response-size forecast from earlier voltage informed the question. The present test scores full waveform error and uses an earlier voltage window.
usesA fixed pre-pulse RMS rule increased mean SEEG interval scores after five fit pulses at 86 selected sites
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.
usesA fixed 6 Hz phase model increased later SEEG size prediction error versus a fixed mean after five calibration pulses in eight sessions
We used the spatial selection, event metadata, and five-early/five-later split. We rebuilt all source files, exactly matched the saved response products, and added phase features. We fit each pair from its five early pulses only.
usesAn updated local RMS model did not establish lower SEEG forecast error at 86 selected readouts in eight sessions
Its recent-input RMS analysis defines a prior fixed-fit question in this cohort. We extend that question with local chronological updates and a matched expanding mean. This is not an independent cohort or an exact replication.
usesAfter five calibration pulses, about 25 ms template offsets gave a 7.48% mean patient error increase in 100–500 ms SEEG forecasts across eight sessions
We used the fixed sample of 86 sites and 886 bipolar readouts, the chronological five-pulse split, retrieval code, source metadata, and late-window trial products. The new source rebuild matched these products exactly for all eight patients.
usesWith five local calibration pulses, a STReEF slope did not establish lower 20–500 ms SEEG size forecast error in eight HAPwave patients
This package supplied the HAPwave measurements, fixed pulse and readout selection, source metadata, full derivative hashes, and extraction code. The new study tests transfer between datasets.
usesA fixed background power rule increased 20 to 500 ms SEEG size forecast error after five calibration pulses in eight sessions
Uses its prior result as context for an exploratory question on the reused cohort. The present rule has no fitted state slope and tests all later pulses. It does not repeat that fitted-slope test.
Pre-pulse RMS reduced mean 20–500 ms log-RMS forecast error in eight STReEF patients with earlier calibration after recorded contact order reversal
[Legacy relation: reproduces] We tested its centered log-RMS forecast method in a second public dataset. We required all calibration responses to finish before each input cutoff. Source selection and window endpoints differ. This is a method replication and does not dispute the earlier result in HAPwave.
usesA HAPwave prior reduced SEEG squared log ratio error by a mean 2.16% at 20 to 500 ms across reversed contact order with five fit pulses in eight STReEF patients
Used its retained HAPwave trial products, source event mapping, licences, and retrieval code through the direct input package. Their diagonals give every window RMS for eight patients.