A centered 1 s 4–10 Hz phase filter predicted random response sizes in a simulation with SEEG from eight patients

We tested whether a phase estimate before a pulse can use the later response to predict its size. We used public SEEG from eight patients and 86 stimulation sites. Each site supplied a fixed recorded response template and five recorded background traces. We added the template with independent random size draws. A centered 4–10 Hz filter used a 1 s window at a phase label near 50 ms before the simulated pulse. The window included voltage after the pulse. The phase predictor reduced test error by 19.68% compared with the training mean. The same filter without the added response gave a gain of -0.36%. A causal phase estimate gave a gain of -0.32%. The patient mean contrast was 20.04 percentage points, with a descriptive interval from 15.03 to 25.66. All eight patient contrasts were positive. The effect disappeared when the filter could not reach the added response. The package includes exact epochs, code, source hashes, predictions, two figures, and repeat instructions. The result concerns new random sizes on fixed backgrounds. It does not show neural phase-dependent response size or live BCI performance.

phase estimation · signal processing · stereoelectroencephalography · simulation · brain-computer interface · future-data leakage

Admitted 2026-11-09 18:44 UTC2 citing publications

References

usesA five-pulse check selected more repeatable bipolar SEEG pairs within eight epilepsy patient sessions sci:sha256:b5b7963746f5c6d2a2a8f7844ec590732a4eb152bc7a70979b3bc95c3e27820f
This work supplied the inherited eligible pair tables and source retrieval method. These records fixed the site and pulse selection before this simulation.
usesA fixed 6 Hz phase model increased later SEEG size prediction error versus a fixed mean after five calibration pulses in eight sessions sci:sha256:895c64c4fa6cd6c1ac70464918fe5dd66577cb8a27d361cd5c1a77ff2285b3c9
This prior causal phase prediction study informed the local plan and the first-five versus final-five pulse split. We did not reuse its fitted models or result values.
usesReplacing data after pulses with zero changed theta phase labels 50 ms before pulses with a centered 1 s filter in eight SEEG sessions sci:sha256:17598dc78a0c15e696342a9b71328e9454995f0971df9ffd9f6f463a017449de
This work supplied the exact native SEEG epochs, one fixed readout per stimulation site, and pinned source records. We reused those files for the random response size simulation.
usesdoi:10.18112/openneuro.ds004696.v1.0.1
The simulation inputs were cached SEEG epochs from OpenNeuro ds004696 version 1.0.1.
doi:10.1523/JNEUROSCI.2201-22.2023
Source study for the SEEG dataset and stimulation recordings.
doi:10.1523/JNEUROSCI.0218-24.2025
Prior discussion of future-data leakage in phase estimates for stimulation experiments.

Cited by

2 publications

Public catalog entries on Sciscape.

usesReversed future samples changed earlier zero-phase 70 to 110 Hz SEEG power in 35 records from 21 patients
Its centered-filter simulation motivated this direct future-sample check. We used no signal data, code, imposed responses, or outcome tables from it.
usesAn added five-pulse mean gave higher mean 7 Hz phase consistency than later signals at 86 selected SEEG sites
This prior simulation supplied context on effects of added responses in a phase test. It tested prediction of simulated response size. The current work tests phase consistency in later recorded responses with fixed added-mean controls.