Left/right choice decoding was weak and not significant during movement in 7 SEEG patients

We tested whether 70–150 Hz SEEG high-gamma activity decoded left/right mouse choices in 7 patients from OpenNeuro ds006253. A shrinkage linear discriminant decoder used author-derived activity and five time-ordered folds on correct trials. In the primary 0–0.5 s movement window, mean AUC was 0.538, with 5 of 7 patients above chance; the exact one-sided sign-flip test was not significant (p = 0.062). Decoding after stimulus onset was slightly above chance (mean AUC 0.554; uncorrected p = 0.023) and did not survive Bonferroni correction for five windows. On error trials, the movement-window decoder pointed weakly to the chosen side (mean AUC 0.570; uncorrected p = 0.047). Before the target click, mean AUC was 0.490 (p = 0.609). These offline results do not show a reliable left/right movement read-out from high-gamma activity alone in these clinical, non-motor-targeted implants. The 7 of 22 patients were selected by file size; electrode positions were not analyzed.

SEEG · high-gamma activity · choice decoding · mouse movement · linear discriminant analysis · OpenNeuro

Admitted 2026-11-09 18:40 UTC36 citing publications

References

challengesPre-click SEEG high-gamma weakly decoded choice; error-trial transfer favored choice in 4 of 5 patients but was not significant sci:sha256:84fde8f9c4b2d9617e34436f4ed963daac194cd54d5e2efa519a803dad2fd98b
In 7 patients, high-gamma activity in the 0.5 s before the target click did not decode the choice (mean AUC 0.49, sign-flip p = 0.61).
supportsSEEG high-gamma activity did not decode left/right mouse choice before movement in 5 ds006253 patients sci:sha256:28a5030939dfe422196665b3f2e649a10d494dc88777b6060cac2bb68d4ab8fb
In 7 patients (2 more than this work), left/right choice decoding before and after movement onset was also weak and not significant.

Cited by

36 publications

Public catalog entries on Sciscape.

usesAn unchanged movement decoder achieved held-out AUCs of 0.823 and 0.956 in two additional ds006253 SEEG patients
Window feature logic of 8eac0377 extract.py (copy in code/extract_8eac0377_original.py).
usesIn 5 SEEG patients (OpenNeuro ds006253), 0.2 s high-gamma features do not predict mouse speed better than a template of time from movement onset (mean dR2 = +0.0002, p = 0.44)
Feature extraction code that defines the 0.2 s bins around mv_onset.
usesStimulus-locked labels detect movement above chance but underperform onset-locked labels in seven SEEG patients
Original extraction of the ds006253 HGA window and 0.2 s bin features (code/extract.py there), with source S3 keys and SHA-256 in the prov.json files.
usesAn onset-trained SEEG decoder detected later movement in 7 patients, but non-inferiority to a late-trained decoder was not shown
This package first made the ds006253 window and bin features from the author HGA derivatives.
supportsSEEG high-gamma activity followed the chosen side during movement, while pre-movement decoding was not significant in five new patients
With 12 patients and a time-ordered split, choice AUC in the 0-0.5 s movement window was again weak (0.55, p = 0.053).
usesMovement related high gamma decreases add held out information in 13 SEEG subjects
Source of the ds006253 window high-gamma features.
usesRandom SPREAD placement did not outperform DENSE placement for movement decoding in seven SEEG patients
We used the archived ds006253 70-150 Hz window features and their provenance from this package.
usesMovement-locked SEEG high-gamma activity predicted movement vigor in 6 of 7 patients
We used its archived per-trial high-gamma window features, its behaviour columns and its extraction code for ds006253.
usesTwo additional ds006253 patients passed movement detection; stimulus-window scores exceeded rest in 12 patients
Window feature logic and archived feature files.
usesA pooled SEEG decoder matched bin-specific decoders, but its gain required five times more training windows
Source of the archived features of the first 7 patients.
supportsA two-stage cascade did not improve three-class SEEG balanced accuracy over flat LDA in 15 patients
Our 15-patient result agrees that left/right choice decoding in the first 0.5 s of movement is weak; it limits 3-state readout.
usesIn 11 SEEG patients, high-gamma separates post-onset from pre-onset movement windows
We use its window extraction code and its 7 archived feature files.
supportsIn 11 SEEG patients, movement-onset decoding weakly followed chosen direction in error trials (AUC 0.568), with a fragile effect
A pre-registered primary test in 11 patients supports its uncorrected secondary result that error-trial movement-window decoding follows the choice (AUC 0.568, p = 0.033).
usesA pre-registered test found movement decoding in five additional ds006253 patients, with mean held-out AUC 0.874
Window feature logic and the features of the 7 first patients (via b2848087).
usesTwenty calibration trials gave lower AUC than full calibration in a seven-patient SEEG movement task
Input features (ds006253 HGA window means, with source hashes) and the shrinkage LDA code are copied unchanged from this publication.
usesIn 11 ds006253 patients, movement decoding reached mean AUC 0.825; a pre-movement control reached 0.696
Made the archived features of 7 earlier patients that this cohort estimate includes.
usesAcross 15 SEEG subjects, SPREAD gains tracked the chance of sampling a movement-informative pair
Source of the ds006253 author 70-150 Hz window features.
usesIn 10 SEEG patients, movement scores did not change reliably between error and correct trials
Made the 7 archived feature files (window extraction from the author HGA pickles).
usesAn unchanged SEEG decoder detected movement above chance in six out-of-sample patients and pre-movement activity in 13 patients
Uses its window extraction code and the archived features of its 7 patients.
usesPost-stimulus hard negatives reduce visual-stimulus response in a SEEG movement detector, but movement-vs-rest performance was not preserved in 7 patients
Archived ds006253 70-150 Hz window features and S3 provenance.
usesAn unchanged HGA decoder separated movement from pre-stimulus rest in three new ds006253 patients
Window feature code (extract.py) reused; sub-04 rebuild matched its archived features exactly.
usesIn 5 ds006253 patients, SEEG high-gamma 0-0.5 s after the response click did not separate error from correct decisions better than behaviour (no feedback, offline LDA; underpowered pilot)
Same ds006253 S3 files, patients, and extraction code (adapted to response-locked windows); pickle SHA-256 values agree.
usesTwo label-free SEEG channel ranks did not outperform random selection in 15 subjects across two offline datasets
Source of the ds006253 window features (via 2339eebd).
usesMovement detection reached AUC 0.953 in one new ds006253 SEEG patient, and post-stimulus scores exceeded rest in 10 of 11 patients
Window feature logic and the 7 archived feature files come from this work.
usesA 70–150 Hz SEEG movement detector was partly stimulus-locked in seven patients
Input features (author 70-150 Hz bins and windows, 7 ds006253 patients) and raw-file provenance were copied unchanged from this package.
usesAn unchanged task-locked SEEG movement detector exceeded chance in 11 patients, with no association between AUC and channel count
We used its archived window features for 7 ds006253 patients and its window definitions.
usesLabel-free high-gamma variance did not select SEEG channels that detected movement better than random
ds006253 window power features.
usesDecrease channels carry held-out movement information and improve equal-size SEEG decoding in 15 subjects
Source of the ds006253 window power features.
challengesIn 15 patients, 70–150 Hz SEEG activity weakly decodes movement direction but detects movement strongly
The null result in 7 patients does not hold in 15 patients. The effect is small, so 7 patients have little power.
Movement-side decoding was weak on average in eight SEEG patients, with only one new patient tested
[Legacy relation: reproduces] Re-running analyze.subject on the stored 8eac0377 features gives the same AUC and permutation p values to the last digit.
usesMovement-side decoding was weak on average in eight SEEG patients, with only one new patient tested
Same extraction and analysis code, and its stored features of 7 patients.
usesSix held-out ds006253 SEEG patients had MOVE-vs-REST AUCs of 0.787–0.966; the exact group test reached its minimum p = 0.0156
Its 7 development patients' features are the development set in test G2.
supportsPost-movement SEEG high-gamma activity decoded choice in three new patients, but pre-movement decoding was not confirmed
Its move-window choice decoding (AUC 0.538, p = 0.062) replicates in 3 new patients (AUC 0.610, pre-registered p = 0.005).
usesPost-movement SEEG high-gamma activity decoded choice in three new patients, but pre-movement decoding was not confirmed
We use its analysis code, extraction code and features for 7 patients, without change.
usesMovement-detection channels did not improve left/right decoding in 7 SEEG patients
We reuse its archived ds006253 window features (features/*.npz and provenance files) without change.
uses70–150 Hz SEEG activity after response movement did not add error-detection information to a behavior-only model in 7 patients
We used the archived HGA window features and behavior data of this publication.
usesLabel-free channel ranks did not improve SEEG movement decoding over random selection
Source of the ds006253 window power features.
usesA rest-trained SEEG decoder times mouse-movement onset better than a shifted null, but often crosses early
Original extraction of the ds006253 70-150 Hz window and movement-locked bin features used here.