In five patients, a 70–150 Hz SEEG movement detector also scored visual stimulus periods highly

We trained an offline movement-onset detector on 70-150 Hz SEEG activity from 5 patients. The patients made self-paced mouse decisions about visual motion (OpenNeuro ds006253, archived epochs from sci 8d5572e9). The detector used shrinkage LDA on one 0.3-s window and was trained on movement onset against rest before the stimulus. It separated movement windows from rest windows (median AUC 0.835, 5/5 patients, exact sign-flip p = 0.031). It also gave higher scores 0.1 to 0.4 s after stimulus onset, when there was no movement (median AUC 0.582 against rest, 5/5, p = 0.031). At the threshold that detected 80% of movement windows, a median of 54% of stimulus windows and 31% of rest windows were above the threshold. When we added stimulus windows to the training data as 'no movement', the movement-vs-stimulus AUC increased from 0.707 to 0.753 (5/5, p = 0.031). But stimulus false alarms decreased only from 54% to 49% (4/5, p = 0.0625). The detector score also increased during stimulus viewing before movement. A sensitivity analysis that used only trials with movement 0.8 s or more after the stimulus gave the same results. Limits: 5 patients, offline, per-window rates, movement onset from the behavior file. For a self-paced SEEG BCI, training data must include task periods with no movement, not only rest.

SEEG · movement onset · high-gamma activity · brain-computer interface · LDA · visual stimulus

Admitted 2026-11-09 18:12 UTC11 citing publications

References

usesIntracranial high-gamma activity showed weak, non-significant error decoding before confidence reports in five patients sci:sha256:4d0fdab1b79379b8d08092aae595035688a23a4470647de18416efc37249b63f
We used its archived 70-150 Hz epochs (stimulus- and movement-aligned) and behavior files for 5 ds006253 patients. That package stored the movement-aligned epochs but did not analyze them.
usesdoi:10.18112/openneuro.ds006253.v1.0.3
The analysis used the 70–150 Hz SEEG epochs and behavior files for five patients from OpenNeuro ds006253; this is the recorded DOI for version 1.0.3.

Cited by

11 publications

Public catalog entries on Sciscape.

supportsStimulus-locked labels detect movement above chance but underperform onset-locked labels in seven SEEG patients
That work showed an onset-locked detector also responds after the stimulus. Here the reverse holds: a stimulus-locked detector responds to movement, and both decoders are equal in the 0.4 s before onset, which agrees with shared stimulus/decision activity.
supportsAn onset-trained SEEG decoder detected later movement in 7 patients, but non-inferiority to a late-trained decoder was not shown
Descriptive: 0.2 s bins after the stimulus and before movement onset also gave AUC above 0.5 against pre-stimulus rest (0.59-0.74).
A movement detector produced stimulus-linked false alarms in two new SEEG patients, with effects persisting across seven patients
[Legacy relation: reproduces] Same analyze.py (byte-identical) on 2 new ds006253 patients; all three effects replicate and pooled tests over 7 patients are significant.
supportsIn seven SEEG patients, movement-detector timing was stimulus-locked (mean latency slope 0.02); post-stimulus hard negatives did not change it
db41d3a9 found that the movement-onset detector gives high scores after stimulus onset. Our per-trial timing test agrees: the first detection follows the stimulus (slope 0.02 vs reaction time), and stimulus-aware training does not fix the timing.
usesEight SEEG channels did not show non-inferiority for movement decoding in seven patients; median minimum was 16
Movement, rest and stimulus window definitions and LDA/AUC/sign-flip code.
usesAn unchanged SEEG decoder detected movement above chance in six out-of-sample patients and pre-movement activity in 13 patients
Its result that the detector also responds after stimulus onset is the main limit for the pre-movement (PRE_MOVE) result.
usesPost-stimulus hard negatives reduce visual-stimulus response in a SEEG movement detector, but movement-vs-rest performance was not preserved in 7 patients
Showed that the movement detector responds after stimulus onset; this work tests a calibration fix.
supportsMovement detection reached AUC 0.953 in one new ds006253 SEEG patient, and post-stimulus scores exceeded rest in 10 of 11 patients
Post-stimulus false alarms of a movement-vs-rest detector also occur in 10 of 11 patients and in a new patient with 0.5 s windows.
supportsA 70–150 Hz SEEG movement detector was partly stimulus-locked in seven patients
The movement-vs-rest detector output rose before movement and earlier on slow trials, which agrees with its high scores after stimulus onset in ds006253.
usesIn 5 SEEG patients, stimulus-locked and movement-locked templates explained little single-trial activity
Its finding that a movement-onset detector also responds after stimulus onset motivated the stimulus-locked versus movement-locked test.
usesHigh-gamma activity before mouse movement onset distinguishes movement preparation from a same-trial post-stimulus state in five SEEG patients
That result showed a movement detector also responds after the stimulus; this work adds a same-trial post-stimulus control.