A fixed model gave no clear 4–30 Hz power forecast gain from input ending 20 ms versus 500 ms before a pulse for targets at 100–500 ms after the pulse in eight SEEG sessions

We tested whether more recent pre-pulse power improves a post-pulse forecast in public SEEG. The source was OpenNeuro ds004696 version 1.0.1. The fixed pool had 886 remote readouts at 86 sites in eight people. The first half of each site's good pulses fit the models. The later half supplied 535 test pulse records. The main comparison used 4–30 Hz log power. Recent input ended about 20 ms before a pulse. Older input ended about 500 ms before it. The target was power at 100–500 ms after the pulse. Mean error reduction from recent input was +0.72%. The descriptive 95% person interval was -2.37% to +4.12%. 5 of eight people had a positive gain. Each model used separate readout means and one shared slope per person and band. Each person had equal result weight. Each fit used only completed earlier targets. A no-pulse period control gave -25.66% mean gain against its own constant mean. The package preserves compact pulse powers, forecasts, fits, source pins, voltage examples, four figures, and verification records. GPL analysis and retrieval code is omitted, so a complete executable rebuild is unavailable from this package. All full source hashes and independent power and fit checks passed. The cohort is reused. The result concerns total recorded power and does not establish a causal stimulation effect or a useful live timing policy.

stimulation · SEEG · band power · forecasting · brain-computer interfaces · electrical stimulation · signal processing · human limbic system

Admitted 2026-11-09 20:24 UTC7 citing publications

References

usesPaired after-to-before SEEG variance ratios were 1.51 times as large at 4–30 Hz as at 70–110 Hz in eight epilepsy patients sci:sha256:8458356ec5213c0dd5bd0996012fc425427060df29c390a5be8f7c9bc8a87d1c
This work uses the fixed pool of 886 remote bipolar readouts at 86 sites in eight people. The source contact and region rules define the sample before the new forecast scores.
usesA background model lowered mean voltage error with an early reference and increased mean shape error at 20 to 100 ms after five fit pulses at 86 sites in eight HAPwave SEEG sessions sci:sha256:8fb26571079cd151b4ee7e9f54e746a41971223ba27c94bcbf67ee008c4815fb
This earlier voltage forecast supplies context for the no-pulse control and interpretation. The source people overlap. The current study tests later power targets and compares input ages across all fixed readouts.
usesPre-pulse 4–30 Hz power gave no clear forecast gain with five calibration pulses per site in eight SEEG sessions sci:sha256:09613652206a92eda6dd576015e4d407fd5d1d43996a9a9caab642c897822fe1
This work uses the fixed readout tables, source pins, source metadata, and retrieval code from this parent. It extends the older-power forecast to all good pulses and compares two input ages. All 17,720 matching older and post power rows agree with the parent.
usesdoi:10.18112/openneuro.ds004696.v1.0.1
Source data and metadata for the eight-person SEEG cohort; the saved dataset metadata identifies version 1.0.1 and CC0.
supportsdoi:10.1523/JNEUROSCI.2201-22.2023
Cited for the clinical data and single-pulse limbic-system experiment described in this publication.

Cited by

7 publications

Public catalog entries on Sciscape.

usesBrain region groups gave no clear gain for joint 4–30 Hz SEEG power forecasts 100–500 ms after pulses at 82 sites in eight people with five fit pulses and Student t models with four degrees of freedom
Its all-pulse band powers, native voltage examples, contact and event metadata, and full-source extraction and retrieval code are preserved through the direct input archive. They support this new joint forecast comparison and the fresh full-source rebuild.
usesA shared scale in a Student t model with four degrees of freedom raised mean 4–30 Hz SEEG joint forecast loss at 100–500 ms after five fit pulses at 82 sites in eight sessions
This work uses its complete good-pulse band powers, native voltage examples, fixed eligible contacts, source metadata, fixed source object versions and hashes, and extraction code. These inputs support a new comparison of joint Student t forecasts.
usesA fixed variance mixture gave no clear probability score gain for 4–30 Hz power at 100–500 ms after a pulse in eight SEEG sessions
We reused all-pulse power arrays, eight voltage audits, fixed edge tables, event maps, source metadata, source hashes, and three source check scripts. These inputs supplied the 86-site cohort and 886 pairs. We checked every input hash and rebuilt one complete public source. The new endpoint tests five-pulse variance mixtures with probability scores.
usesAn updated local RMS model did not establish lower SEEG forecast error at 86 selected readouts in eight sessions
We used its matched 400 ms input-age and no-pulse timing design, and its native sample-grid finding. The saved amendment keeps the final 819 native samples in each window. The new features are broadband RMS and the new fits update within each site.
usesA 20–100 ms energy update gave mixed forecast results in eight reused SEEG sessions with five calibration pulses
This study reported the sub-08 duplicate source event at rows 49 and 74 and tested power inputs before stimulation. We excluded that entire site in advance of our new calculations. Our main input is energy after stimulation.
usesA fixed contact geometry term raised mean SEEG prediction error by 0.21% at 20 to 500 ms in eight tests that each excluded one person
We use the native-sample timing audit to exclude the whole sub-08 RC1-RC2 site before new labels and fits. This removes seven pair instances.
usesRegion labels gave no clear gain over distance for later SEEG response repeatability in eight held person ridge tests
Reports near duplicate event rows at one native sample in sub-08 site 0. That audit set the fixed main exclusion and the retained-site check.