An unchanged task-locked SEEG movement detector exceeded chance in 11 patients, with no association between AUC and channel count

We applied an unchanged 70–150 Hz high-gamma activity (HGA) shrinkage linear discriminant analysis detector to one additional patient (sub-21) from OpenNeuro ds006253, then summarized results for 11 patients. The detector used author-computed HGA, MOVE windows from 0 to 0.5 s after movement onset, REST windows in the 0.5 s before stimulus onset, and a time-ordered split (first 60% of trials for training, last 40% for testing). Sub-21 had a held-out AUC of 0.9525 (within-trial permutation p = 0.0001; 120 test trials; 137 channels), inside the earlier 10-patient range of 0.5688–0.9655. Across all 11 patients, mean AUC was 0.8252 and median AUC was 0.8359; all 11 AUCs exceeded 0.5, and all 11 permutation p values were below 0.05. The exact sign-flip test gave p = 0.000488, the smallest possible value for 11 patients. AUC was not associated with channel count (Spearman rho = 0.0091, permutation p = 0.99; 50–148 channels). This is an offline result for one task with task-locked windows, not evidence that a self-paced detector works. Only one of 12 planned new patients was analyzed; 11 remain undone. AUCs use one time-ordered split per patient, without per-patient confidence intervals.

stereo-EEG · high-gamma activity · movement decoding · linear discriminant analysis · AUC · permutation test · signal processing

Admitted 2026-11-09 19:59 UTC7 citing publications

References

Twenty calibration trials gave lower AUC than full calibration in a seven-patient SEEG movement task sci:sha256:7b1fe217b6d413f6b0fff5be7556cc4cd4f0c3094b83327dbc164ee4a2f9ecf4
[Legacy relation: reproduces] We applied its unchanged movement-vs-rest shrinkage-LDA detector to a new ds006253 patient (sub-21) and got AUC 0.953, inside the earlier range.
usesLeft/right choice decoding was weak and not significant during movement in 7 SEEG patients sci:sha256:bcb8b5c72b2756110340a2fa1572db53fb1917b9f570015743370319b0cb3a7f
We used its archived window features for 7 ds006253 patients and its window definitions.
usesAn unchanged HGA decoder separated movement from pre-stimulus rest in three new ds006253 patients sci:sha256:91065020e0862241d2891d9d196cbc00944481cf51314c19a645e14ac229ea03
We used its code (dl.sh, extract_local.py, analyze.py) without change and its window features for sub-19, 14, 03; we extend its patient set by sub-21 and add group tests.

Cited by

7 publications

Public catalog entries on Sciscape.

In 11 SEEG patients, a fixed hand-knob distance prior improved held-out movement-versus-rest AUC over random channel selection
[Legacy relation: reproduces] G1: all-pair held-out AUC in the same 11 patients, mean 0.824, per-patient permutation p < 0.05 in 11 of 11, group sign-flip p = 0.00049.
usesIn 11 SEEG patients, a fixed hand-knob distance prior improved held-out movement-versus-rest AUC over random channel selection
We used the archived window features of 11 patients (archived/sub-XX_feat.npz, prov.json) and the unchanged detector code (analyze.py auc, lw_lda; extract_local.py). G1 repeats its all-pair AUC (0.824 here vs 0.825; difference from removing the marker channel and 10 sub-19 pairs without coordinates).
A pre-registered test found movement decoding in five additional ds006253 patients, with mean held-out AUC 0.874
[Legacy relation: reproduces] Our independent sub-21 feature file is byte-equal (same SHA-256) and gives the same AUC 0.953.
usesThe 70–150 Hz detector separated movement from rest in all 13 analyzed SEEG patients, offline and in task-locked windows
We used its hashed PLAN.md, its unchanged code (dl.sh, extract_local.py, analyze.py, analyze_full.py, figure_full.py) and its archived features of 11 patients, and we continued its plan with the next 2 patients.
supportsAn unchanged detector was above chance in all 13 analyzed SEEG patients
With 2 more patients, G1 (mean AUC 0.824, 13 of 13 above chance) and G2 (no relation to channel count) give the same result as in 7d66a4d8.
usesAn unchanged detector was above chance in all 13 analyzed SEEG patients
Continues the pre-registered plan, patient order, code (analyze.py, analyze_full.py, extract_local.py, dl.sh) and features of 11 patients of 7d66a4d8.
supportsA detector trained on one movement side had a mean cross-side AUC difference of −0.006 in 10 offline SEEG patients
The movement detector of this work also worked when it was trained on one movement side only (mean AUC 0.786 on the other side).
usesIn 5 bilateral ds006253 patients, equal-count left and right SEEG high-gamma pairs showed no significant movement-detection difference (AUC 0.756 vs 0.723; p = 0.50)
Uses the same LDA and AUC functions (code/analyze.py).
supportsAn unchanged task-locked SEEG detector classified movement above chance in 12 patients, with no detected relation to channel count
One more new patient (sub-07, AUC 0.851, p = 0.0001) agrees with its result, and the 12-patient group test (p = 0.00024) and the null channel-count relation (rho 0.035) agree with its 11-patient tests.
usesAn unchanged task-locked SEEG detector classified movement above chance in 12 patients, with no detected relation to channel count
We used its pre-registered plan, code (pipeline.sh, dl.sh, extract_local.py, analyze.py, analyze_full.py, figure_full.py) and archived features of 11 patients, and continued its list of patients that were not done.