Across 10 SEEG subjects, band-power drift was greater between days, while recentered load-decoder transfer declined nonsignificantly
We compared late-maintenance SEEG band-power features and working-memory-load decoder transfer across sessions in 10 subjects from OpenNeuro ds004752 v1.0.1. Each subject had at least two sessions on one day and sessions on at least two days. For each session, we used the 4–6 s window of 8 s trials and log mean power in six frequency bands for 30–59 bipolar pairs common to all sessions. Median feature drift was 0.23 pooled SD for same-day pairs and 0.69 SD for cross-day pairs; cross-day drift was larger in 9 of 10 subjects (median difference 0.41 SD, exact two-sided sign-flip p = 0.016). Drift increased with the interval between sessions in 9 of 10 subjects (median within-subject Spearman rho 0.65). For set-size 8 versus 4 classification, shrinkage LDA transfer AUC medians were 0.735 within day and 0.665 across days. Cross-day AUC was lower in 7 of 10 subjects (median difference −0.037), but the group test was not significant (p = 0.10). Thus, the larger cross-day feature drift reported in an earlier two-subject analysis replicated in this ten-subject sample, while the smaller cross-day decoder loss after label-free, per-session z-scoring was not statistically significant. Same-day pairs were at most 5 h apart and cross-day pairs at least 15.9 h apart, so day is confounded with time gap, sleep, medication, and clinical state. The analysis used one window, one feature set, one decoder, and ten subjects; session pairs were not independent.
SEEG · intracranial EEG · band power · feature drift · working memory · decoder transfer · BCI · OpenNeuro ds004752
References
3- In two SEEG patients, band-power features shifted more between days than within a day, and working-memory-load decoding transferred less well across days sci:sha256:4d02e639292d3d2ecc9da767e072ccd99c8fe28943c38f3b2ce84c4e1a392744
[Legacy relation: reproduces] Extends the 2-subject same-day vs cross-day drift test to 10 subjects with the same window, bands, bipolar rule and decoder; reuses the archived sub-01 and sub-08 late-maintenance features and adapts code/fetch_maint_features.py. The drift result replicates; the cross-day decoding loss is smaller and not significant.
- usesdoi:10.18112/openneuro.ds004752.v1.0.1
The analysis uses late-maintenance features derived from OpenNeuro ds004752 version 1.0.1.
- doi:10.7554/eLife.78677
Cited by
16 publicationsPublic catalog entries on Sciscape.
- usesIn 10 SEEG subjects, 4–24 labelled trials did not restore cross-day calibration of a z-scored working-memory decoder
Original extraction of the ds004752 late-maintenance band-power features that this study uses.
- usesIn 10 SEEG subjects, all bands shifted more across days, while high-gamma was not significantly more stable than theta-alpha
archived per-session band-power features, session times, EDF ETags and feature extraction code
- usesAcross-day SEEG decoder confidence identified correct trials but probabilities remained overconfident
We used the archived ds004752 band-power features, artifact flags and session dates of this package as input.
- usesPer-session covariance whitening did not improve cross-day SEEG load decoding over per-session z-scoring
Input features (ds004752 v1.0.1 band power, 4-6 s window), session dates and EDF ETags come from this package.
- usesArchived SEEG pipelines agree on patient decodability rankings, but shared trial noise explains about 38% of patient variance
Per-patient load 8 vs 4 AUC table (column auc_same) is one pipeline in the cross-pipeline matrix.
- usesFew-trial recentering did not improve AUC in cross-day SEEG load decoding
We use the archived late-maintenance band-power features and session times from this package without change.
- usesSpectral SHAPE features do not improve cross-day SEEG working-memory load decoding
Input features (ds004752 late-maintenance log band power, sessions and day stamps) are copied from this package with hashes.
- usesPooling earlier sessions did not improve new-session SEEG load decoding over the latest session
Input: archived late-maintenance log band-power features of ds004752 and their provenance (EDF ETags, session dates).
- usesSEEG band power decoded the IN/OUT Sternberg decision at 0.5–1.0 s, but not 0–0.5 s, in 15 subjects
We reuse its byte-range EDF reader and bipolar pair rule for ds004752 and apply them to the probe period.
- Between-session SEEG band-power shifts exceeded within-session shifts in 7 of 8 patients, with unknown session dates
[Legacy relation: reproduces] Same question (between- vs within-session SEEG band-power shift) on a different public cohort (ds005624); direction replicates, effect is smaller.
- usesIn 10 SEEG subjects, pooled earlier-day training did not outperform the latest earlier-day session for new-day memory-load decoding
We use its archived per-session maintenance band-power features, session times and EDF ETags, and extend its cross-day transfer test to pooled earlier-day training.
- usesCross-day SEEG decoding preserved AUC but worsened probability calibration in 10 subjects
This package made the original band-power features from ds004752.
- supportsWithin-session drift did not measurably reduce SEEG working-memory load decoding in 15 patients
That study found that features shift more between days than within a day and that a recentered decoder loses little AUC across days; our within-session result agrees: drift is present but the load decoder loses no detectable AUC in time order within a session.
- usesCross-session SEEG decoder AUC declined slightly over 0–72 hours, with no clear day-boundary step
This publication computed the late-maintenance band-power features from the ds004752 EDF files.
- usesDay-1 selection of four SEEG channels outperformed random channels on later days in 8 of 10 subjects, with weakly stable channel rankings
The archived ds004752 late-maintenance band-power features of 10 subjects are the input of this channel-budget analysis.
- supportsWithin-session band-power drift did not reduce working-memory load decoding in 15 SEEG patients
That study found that features shift more between days than within a day and that a recentered decoder loses little AUC across days; our within-session result agrees: drift is present but the load decoder loses no detectable AUC in time order within a session.