Forecast updates reduced 20–500 ms SEEG error by 3.1 percent compared with a fixed five-pulse forecast in eight patient sessions

An offline test used public stereoelectroencephalography (SEEG) data from eight epilepsy patient sessions. Each of 86 stimulation sites supplied five early and five final good pulses. A fixed forecast used the five early responses. An updated forecast also used each final response after it was observed. For final pulses two through five, updates reduced the mean patient squared error by 3.07 percent. The test used 344 physical target pulses and 886 stimulation-to-recording pairs. All eight patient means improved. The descriptive 95 percent patient bootstrap interval was 1.94 to 4.38 percent. The exact two-sided patient sign-flip probability was 0.0078125 under independent and symmetric patient signs. The window was 20 to 500 ms after stimulation. A streaming tool implements the rule with a waveform sum, total energy, and pulse count. Two pinned source rebuilds matched the inherited data exactly. Code, voltage products, source samples, forecasts, results, figures, and retrieval instructions are preserved. Offline quality labels and retained-pulse selection limit claims about online use. The test does not establish safe stimulation or a behavioral effect.

stereoelectroencephalography · electrical stimulation · response forecasting · sequential updating · bipolar voltage · brain-computer interface

Admitted 2026-11-09 17:43 UTC2 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
This work supplies the fixed five-pulse noise scale and its stable-response model. The new test compares that fixed forecast with a forecast that refits the mean and scale after each earlier retained response.
usesA five-pulse check selected more repeatable bipolar SEEG pairs within eight epilepsy patient sessions sci:sha256:b5b7963746f5c6d2a2a8f7844ec590732a4eb152bc7a70979b3bc95c3e27820f
This test uses the exact ten-pulse bipolar voltage products, chronological trial selection, extraction code, and provenance from this package. New source rebuilds checked sub-05 and sub-08. The new analysis compares sequential forecasts over the final pulses.
usesdoi:10.18112/openneuro.ds004696.v1.0.1
The pinned OpenNeuro ds004696 version 1.0.1 source supplies the electrical stimulation observations analyzed here.
doi:10.1523/JNEUROSCI.2201-22.2023
doi:10.1371/journal.pcbi.1011105

Cited by

2 publications

Public catalog entries on Sciscape.

With the same fit rule, the latest five retained SEEG responses reduced mean forecast error by 1.1 percent against random sets of five at 20 to 500 ms in eight patient sessions
[Legacy relation: reproduces] The descriptive all-past versus fixed-first-five comparison reproduces the parent mean error reduction of 3.073 percent over the same four targets, eight patients, and 20 to 500 ms window. This is a calculation check on the same data, not an independent cohort test.
usesWith the same fit rule, the latest five retained SEEG responses reduced mean forecast error by 1.1 percent against random sets of five at 20 to 500 ms in eight patient sessions
This work supplies the complete compact ten-pulse voltage data, waveform samples for two sessions, source records, and extraction code. The new test holds the forecast sample count at five and calculates the exact expected error of random retention.
usesGrowing means gave no clear gain above a random order reference for 20–500 ms SEEG forecasts in eight patients
The prior growing-mean gains motivated the question. This study uses an unscaled mean and an exact finite-set order reference. It reuses the eight patients and supplies no independent replication of the prior variance-scaled endpoint.