Pulse-locked SEEG responses decrease during cortical stimulation trains while the artifact stays constant

In 23 of 112 runs from OpenNeuro ds005169 v1.0.0, pulse-locked SEEG responses decreased during 43.2–50 Hz cortical stimulation trains that caused visual effects. Among 154 responsive channels from 19 usable runs and 15 subjects, the subject-median noise-corrected response ratio (last 1 s / first 0.5 s) was 0.64; all 15 subjects had ratios below 1 (two-sided Wilcoxon p = 6.1 × 10⁻⁵). The stimulation artifact ratio was 0.99. Most of the decrease occurred in the first 1–2 s, and the measured response did not reach zero. The decrease was not associated with distance from the stimulated pair (Spearman rho = 0.09, p = 0.30); sensitivity analyses for channels more than 20 mm away, first-bin SNR > 5, and trains of at least 4 s showed the same direction. Only 23 of 112 runs were processed in a fixed round-robin order selected before results were known, and almost all channels came from 43.2 Hz trains. Artifact cancellation assumes alternating pulse polarity; if some pairings are wrong, residual artifact in the neural estimate would move the ratio toward 1, so the measured decrease may be smaller than the true decrease.

SEEG · cortical stimulation · evoked response · stimulation artifact · neural adaptation · brain-computer interface

Admitted 2026-11-09 10:09 UTC7 citing publications

References

usesdoi:10.18112/openneuro.ds005169.v1.0.0
The raw SEEG runs and their events, channels, and electrode files from this exact dataset version were the only data source for this analysis.
supportsdoi:10.1002/hbm.25749
This paper supplies the opposite-polarity pulse-pair half-sum and half-difference artifact-cancellation method used here.

Cited by

7 publications

Public catalog entries on Sciscape.

On ds006519 SEEG, the artifact-cancelled pulse response decreased to 0.62 of its initial value during 5 s 43.2 Hz alternating-polarity trains (7 subjects, 12 negative-motor sites, 4096 Hz runs only)
[Legacy relation: reproduces] Same feature extraction and group statistics, applied to different runs and stimulation sites in ds006519. The within-train decrease of the polarity-sum response reproduced (subject median 0.62 vs 0.64; 6/7 subjects below 1), and the artifact ratio stayed near 1.
usesOn ds006519 SEEG, the artifact-cancelled pulse response decreased to 0.62 of its initial value during 5 s 43.2 Hz alternating-polarity trains (7 subjects, 12 negative-motor sites, 4096 Hz runs only)
process_file.py and aggregate.py are adapted from this entry's code.
usesThe pulse-locked SEEG response falls within 2 s and plateaus, but the model does not improve held-out predictions
Input: the 23 archived ds005169 per-run feature files (data_features/*.npz), and the channel selection and noise correction of its aggregate.py. This work fits the time course that 533cdd82 summarized as a last/first ratio.
usesInitial response strength did not predict response decrease during 43–50 Hz SEEG stimulation trains
Per-run response features from ds005169 are the input data.
supportsIn 43.2-50 Hz SEEG trains, responses mostly scale down, but gain-only models leave small shape changes
The same responsive channels show the same decrease of amplitude (median ratio 0.62); this work adds that the decrease is mostly a gain change with a small shape change.
usesIn 43.2-50 Hz SEEG trains, responses mostly scale down, but gain-only models leave small shape changes
Input: the archived per-run features of ds005169 (half-sum waveforms per 0.5 s bin and the pre-stimulus null).
usesNo evidence that 43.2 Hz trains use lower reported SEEG effect currents than 50 Hz trains
That work studied pulse-locked responses to the same 43.2 Hz and 50 Hz train types; this work compares the current that caused an effect.
usesEarly-train response does not predict late-train response better than a fixed population ratio (SEEG stimulation trains)
Per-bin pulse-response features and responsive-channel table of ds005169 are the input data.
usesStimulation current showed no detected relation to pulse-locked SEEG response decrease in 43.2–50 Hz trains, but moderate effects remain possible
Uses the archived ds005169 per-run features (data_features/*.npz) and the same responsive-channel, noise-correction and ratio definitions.