Later pulse variation accounted for 72.1 percent of normalized 20–500 ms SEEG error for a scaled five-pulse forecast in eight patient sessions

We tested how much fixed response forecast error came from differences between later stimulation pulses. Public SEEG data from OpenNeuro ds004696 provided eight patient sessions, 86 stimulation sites, and 886 stimulation-to-readout pairs. The forecast used a scaled mean of five early pulses. Testing used five later pulses at each site. Within-test pulse variation accounted for 72.1 percent of normalized 20–500 ms square voltage error. The descriptive 95 percent patient bootstrap interval was 69.6 to 74.5 percent. The observed error split is exact and needs no independent-noise assumption. A five-pulse correction gave a noise-to-error ratio of 90.1 percent under a stable mean and independent test deviations. Those assumptions were not established. The raw voltage variation share was 66.0 percent. All eight public source files were checked. Their compact arrays and tables matched the archived inputs exactly. A fresh analysis reproduced all six result files. The package contains a proof, reusable code, compact data, all scores, and three figures. The result applies to fixed centered waveforms in these selected epilepsy sessions. It does not establish a biological noise floor or functional brain control.

stereoelectroencephalography · electrical stimulation · evoked potentials · forecast error · pulse variation · noise estimation · brain-computer interfaces · reproducibility

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 reused its five-pulse scaled forecast, selected pairs, and exact pulse products. We reproduced 288 patient forecast-loss comparisons. The new result splits its remaining single-pulse error into test-pulse variation and distance to the test mean.
usesA five-pulse check selected more repeatable bipolar SEEG pairs within eight epilepsy patient sessions sci:sha256:b5b7963746f5c6d2a2a8f7844ec590732a4eb152bc7a70979b3bc95c3e27820f
We reused its GPL extraction recipe, chronological first-five and final-five pulse rule, spatial pair definitions, and pinned source provenance. We freshly rebuilt all eight source CRP files and checked every compact array and table.
usesdoi:10.18112/openneuro.ds004696.v1.0.1
The analysis used processed SEEG source files from OpenNeuro ds004696 snapshot 1.0.1.
doi:10.1523/JNEUROSCI.2201-22.2023

Cited by

10 publications

Public catalog entries on Sciscape.

usesSeparate pair bounds covered 71.5 percent of individual SEEG responses and 17.0 percent of whole pulses at 20 to 500 ms in eight sessions
Uses exact compact waveform products, pulse maps, source hashes, and the prior source audit for all eight sessions. These inputs support the new analysis of individual and whole-pulse RMS error coverage.
usesA joint rank check accepted 83.2 percent of later SEEG responses at 20 to 500 ms in eight patients after five calibration pulses
Uses the exact 10-pulse waveform products, site and readout definitions, chronological pulse split, and pinned source metadata. The new endpoint is joint conformal response coverage. This is the same cohort, not an independent replication.
usesA pulse order test under joint exchangeability did not detect closer adjacent 20 to 500 ms SEEG responses in eight sessions
Reused and hash-checked its exact pulse product matrices, contact and event metadata, source provenance, and extraction code. The new analysis tests adjacency within the final five pulses.
usesTwo components from five early pulses captured 29.8 percent of later SEEG variation at 20 to 500 ms after removal of the early mean direction in eight patients
Uses its credited inherited extraction code for contact parsing, region definitions, and analysis windows. The code and source lineage were preserved through the all-pulse parent publication.
usesThree spread pulse records had higher mean normalized 20–100 ms SEEG forecast error than nearby records at equal mean times in eight patients
Uses the inherited GPL extraction functions for contact parsing, broad source region rules, and four window definitions. The new quadratic forecast tool makes no independent-noise interpretation of the descriptive calibration variance.
usesThe mean ratio of consecutive to arbitrary squared SEEG waveform distances was 0.992 at 20–500 ms in eight patients
Used its source extraction code and verified ten-record pulse products to check every inherited first/final-five product. Its noise assumptions motivated this order audit.
usesThree spread pulse records had higher mean normalized 20–100 ms SEEG forecast error than nearby records at equal mean times in eight patients
Uses the inherited GPL extraction functions for contact parsing, broad source region rules, and four window definitions. The new quadratic forecast tool makes no independent-noise interpretation of the descriptive calibration variance.
usesIdeal signed scale changes left 84.5 percent of 20–500 ms SEEG error for a fixed five-pulse waveform in eight patient sessions
Reuses its exact chronological ten-pulse products, archived source metadata, extraction code, and first-five noise-scaled mean forecast. The new analysis bounds scale-only improvement and splits errors by projection onto the early mean shape. Its within-test components reproduce the parent variation share.
usesA geometric median of five early pulses gave an uncertain 0.4 percent decrease in later SEEG prediction error at 20 to 500 ms in eight patients
This study reuses its public-source extraction code and compact data definition through publication 4cfd1859-0a96-479f-b45f-6c521cbe75ed. The new result tests common pulse-map weights. The extraction signal rules are unchanged.
usesA joint five-pulse test did not detect smaller adjacent SEEG waveform distances at 20 to 500 ms in eight patients
We use the source file hashes, source metadata, and processing code. We check our first and final block matrices against the archived matrices.