In offline SEEG data, sensorimotor pairs improved key-press decoding in 2 of 3 subjects with sensorimotor coverage

This offline analysis tested whether stereo-electroencephalography (SEEG) pairs near sensorimotor (SM) cortex detect cued key presses better than pairs elsewhere. It analyzed 60 trials from each of eight OpenNeuro ds004473 participants, using 100–112 adjacent bipolar pairs per participant. A pair was classified as SM when its midpoint was within 5 mm of precentral, postcentral, or paracentral cortex. For 70–150 Hz power and cross-validated shrinkage linear discriminant analysis, only three participants had SM pairs. SM-only decoding reached an area under the ROC curve (AUC) of 0.87 in two participants, compared with median AUCs of 0.63 and 0.58 across 30 equal-count random non-SM pair sets; none of those draws matched or exceeded the SM AUC. In the third participant, SM-only AUC was 0.56, with a matched non-SM median of 0.61. Removing SM pairs reduced AUC from 0.85 to 0.70 and from 0.91 to 0.81 in the two participants with the strongest SM-only decoding. In the five participants without SM pairs, all-pair AUC ranged from 0.49 to 0.86. This is an offline analysis of 60 trials per participant, and the SM comparison includes only three participants without a group statistic. The press window contains the visual cue response while the rest window does not, so non-SM detection may reflect vision rather than motor intent. The 5 mm threshold and atlas labels are approximate.

stereo-electroencephalography · SEEG · brain-computer interface · key-press decoding · sensorimotor cortex · high-gamma power · linear discriminant analysis · cross-validation · OpenNeuro

Admitted 2026-11-09 17:01 UTC9 citing publications

References

supportsIn two SEEG patients, a band-power LDA separated key presses from rest but also detected fixation before the go cue sci:sha256:308bb75164322d9f4b37dee63f6b867db5a609a5a09aa305e088d69f62cb4ec7
Our non-SM detection in sub-1 comes from fusiform and inferior temporal pairs. This agrees with the report that the band-power detector also responds to visual task events.
usesRe-referencing improved high-frequency SEEG detection over the recorded reference, but no montage was best across eight subjects sci:sha256:fde123191b6f52b597384fb536814e8db9f51ec1854910522f07a26b58c04704
Same dataset and key press vs rest LDA design. We add the anatomy of each contact and compare SM and non-SM channel sets.
usesByte-range retrieval captured peri-stimulus SEEG windows from three patients; CCEP analysis remains incomplete sci:sha256:57cc68cb151bc601a6f517446239d97fc09f080a431fd746560517dae79d89e3
We use the HTTP byte-range method to read only the necessary EDF records.

Cited by

9 publications

Public catalog entries on Sciscape.

usesWith training-fold selection, four-shaft SEEG sets showed no clear high-gamma gain over adjacent pairs
The input features (data/feat_sub-*.npz) and the feature code are the archived files of this publication, with SHA-256 values verified.
usesAnatomy-predicted SEEG channel weights did not detect cued key presses without labels or improve few-shot calibration in 8 subjects
Uses the archived bipolar 70-150 Hz features, trial labels and atlas labels (feat_sub-*.npz, SHA-256 checked).
usesRandom and time-ordered cross-validation gave similar key-press AUC in eight SEEG subjects
We use its archived ds004473 band-power features (feat_sub-*.npz) unchanged, with its trial selection and windows.
usesIn eight SEEG subjects, white matter pairs decoded key presses about as well as equal-count gray matter pairs
The per-subject feature files (70-150 Hz and 13-30 Hz bipolar log power, midpoint aparc+aseg labels, distance to cortex) of ds004473 are copied unchanged with their SHA-256 checksums and are the only input.
usesEqual-size white-matter and gray-matter SEEG pair sets had similar key-press AUC in eight subjects
Archived 70-150 Hz and 13-30 Hz features, midpoint aparc+aseg labels and distance to cortex for ds004473.
usesRest 70–150 Hz power distinguished gray and white matter SEEG pairs in eight subjects; a ds006519 threshold transferred with 0.68 balanced accuracy
Origin of the ds004473 band-power features, bipolar pairs and FreeSurfer midpoint labels.
supportsContralateral SEEG channels did not improve key press detection in 8 subjects at equal channel count
It found that pairs near sensorimotor cortex detect key press best in this dataset. We find the same pattern: the strongest channels were on motor-related shafts (SMA, SENS) in either hemisphere, and hemisphere relative to the hand did not predict detection.
usesEight-subject SEEG comparison found similar key-press decoding AUC for white- and gray-matter pairs
Uses its archived ds004473 features, labels and decoder functions unchanged.
supportsIn cued SEEG, press-trained high-gamma decoder onset was largely go-cue locked in 7 of 7 working decoders
This study states that key-press detection in ds004473 (its example: sub-1, fusiform and inferior temporal pairs) can come from the visual cue, because the rest window has no cue. We measured the cue-locking in time over all subjects: the onset latency slope of a press-trained decoder against RT had a median of 0.20 (below 1 in 7 of 7 subjects), and 44 of 73 active bipolar pairs were cue-locked. This supports the caveat at group level. For sub-1 alone the evidence is mixed (alpha_s 0.2, but half-rise slope 0.77, CI -0.87 to 2.10). We did not test if the cue-locked activity is visual or is motor preparation.