SEEG power before a response gives 57.7% mean balanced accuracy for two cued keys on later trials in eight recordings

We tested whether SEEG power before a key press can predict key choice in later trials. We used eight people from OpenNeuro ds004473, snapshot 1.0.1. A fixed ridge classifier learned from the first 70% of eligible trials and tested on the last 30%. Mean balanced accuracy was 57.7% before the response and 51.4% before fixation. The subject bootstrap interval for the first score was 52.7% to 63.3%. The exact one-sided paired comparison with baseline gave p=0.0586. A cue and key audit recovered 102 eligible trials with a zero source correctness value in person 1. The package preserves features, code, trial audits, figures, and a source retrieval recipe. Cue identity and key identity were coupled. Recorded response times set the windows. The evidence concerns offline key reading in these records. It does not establish continuous BCI control or brain stimulation.

stereoelectroencephalography · brain-computer interface · balanced accuracy · ridge classifier · within-recording decoding · label audit

Admitted 2026-11-09 08:10 UTC18 citing publications

References

usesdoi:10.18112/openneuro.ds004473.v1.0.1
Source recordings, task events, and channel metadata from OpenNeuro ds004473 snapshot 1.0.1 were used.
doi:10.1088/1741-2552%2Facae0a

Cited by

18 publications

Public catalog entries on Sciscape.

usesByte parity kept all six-bit SEEG decisions in eight offline records under one known packet loss and correct received bytes
Used inherited trial audits and upstream source records through the compact signal and graph parents. These records identify the retained samples and support signal retrieval.
usesSelection for recorded keys did not show lower mean offline Brier error in eight cued SEEG records with eight pairs and C=0.1
Its earlier recorded-key test supplied the related-work context. We tested a new matched comparison of state and key pair selection. We did not use its features or models.
usesRemoving contact means from 32 early baseline windows raised mean later balanced accuracy by 6.34 percentage points over a raw median in eight SEEG recordings
Its original source version records, SEEG contact lists, trial audits, EDF reader, and power calculation were inherited through publication b3bf91d3-4686-4a56-8964-7d495e2d724d. These fixed the source and offline eligibility scope.
usesA median reference reduced mean accuracy loss from simulated white noise at 100 times the training scale on one unused contact in eight SEEG recordings
The exact source versions, good SEEG contacts, trial audits, EDF reader, and power code enabled the trial-matched fault test. Saved power inputs enabled source extraction checks.
usesMedian reference power at 70–110 Hz gave 53.1% mean balanced accuracy on later trials from periodic 1–30 Hz SEEG inputs in eight people
Its audit code and eligibility rule were inherited through the direct parent package. The original code and exact trial rows are preserved for inspection.
usesOffline SEEG power before a response separated task and rest across cue durations in six patients
The source correctness audit identified a key and correctness-flag mismatch in patient 1. We used that finding to add a metadata audit after the main analysis. We confirmed 102 synchronized, source-kept, matching-key left trials within the deadline had a zero correctness flag. The six primary patients had no such mismatch. The planned trial filter was retained.
usesLocal baseline subtraction did not establish higher cued-key accuracy with eight SEEG pairs in later trials from eight people
This publication supplied the fixed power features, retained trial labels, split rule, source metadata, and raw retrieval hashes. The new study tests same-trial baseline subtraction with eight selected bipolar pairs. It also reproduces the source all-pair bipolar score.
usesPower from eight SEEG pairs before known response times gave 44.7 percent mean key agreement on 33 cue conflict trials in six recordings
Prior two-key decoding showed that ordinary cue and key labels were coupled. This limit supplied the question for our conflict test. We used a new fixed reader and test split rather than that study's feature inputs or reported score.
usesA fixed ridge model with 300 ms of SEEG after the cue increased mean absolute log response-time error in eight patient tests
Its source label audit informed the decision to leave the source correct flag out of eligibility and model inputs. We use the explicit source key map and preserve the unused flag for audit.
usesAn eight-contact common average reference changed fixed task decoder decisions in eight public SEEG records
The inherited trial rules, audits, source metadata, EDF reader, and power method came from this work through the cited reference fault study. We preserve the inherited trial method code. The original task audit enabled fixed eligible trials and source sample windows.
usesAn early score threshold reduced the mean command rate outside response guards in 4 later SEEG task streams
This study adapts EDF access, source retrieval, and spectral functions from code/prepare.py. It also uses the pinned source metadata and object version records as retrieval inputs. All source text and all four included full EDF files were checked again. No prior feature or result arrays enter this analysis. The studies use the same source patients.
usesLocal baseline subtraction did not establish higher cued-key accuracy with eight SEEG pairs in later trials from eight people
This publication supplied the fixed power features, retained trial labels, split rule, source metadata, and raw retrieval hashes. The new study tests same-trial baseline subtraction with eight selected bipolar pairs. It also reproduces the source all-pair bipolar score.
usesSeven connected difference channels preserved eight-contact SEEG decoder decisions without added error in eight public records
We used its source trial rules, metadata, and EDF methods through the retained upstream code.
usesEight public SEEG recordings contain 33 trials with opposite cue and key labels after fixed signal checks
The prior key-label audit and its stated cue/key ambiguity guided this conflict test. We used its explicit key-code and source correctness audit as a reference. Its feature arrays and test scores did not enter the new models.
usesTwo eight-pair SEEG selection rules gave 48.4% and 50.6% mean three-output recall in eight offline later-trial tests
Its separate cued-key test helped define the question about shared pair selection for task detection and key decoding. We used different source features and did not compare its scores directly with our scores.
usesTwo eight-pair SEEG selection rules gave 48.4% and 50.6% mean three-output recall in eight offline later-trial tests
Its separate cued-key test helped define the question about shared pair selection for task detection and key decoding. We used different source features and did not compare its scores directly with our scores.
usesLearned eight-contact references reduced reference error, but decoder agreement gains were uncertain in eight public SEEG records
We preserved its trial rules, inclusion audit, and source trial-mask code through the parent package. These rules define the retained source trials used for calibration and testing.
usesMedian reference power at 70–110 Hz gave 53.1% mean balanced accuracy on later trials from periodic 1–30 Hz SEEG inputs in eight people
Its audit code and eligibility rule were inherited through the direct parent package. The original code and exact trial rows are preserved for inspection.