Low-frequency SEEG voltage decodes left versus right keypresses better than beta or high-gamma power in eight patients
In eight SEEG patients from OpenNeuro ds004473, we compared local motor potential (LMP), beta power (13–30 Hz), high-gamma power (70–150 Hz), and their combination for decoding left versus right arrow-key presses with bipolar channels and shrinkage LDA. Across the eight subjects, mean AUC in the 200 ms before the keypress was 0.66 for LMP, 0.49 for beta, 0.53 for high gamma, and 0.61 for all features together. LMP exceeded each alternative in seven of eight subjects (exact sign-flip p = 0.008 for each paired difference); its group AUC exceeded 0.5 (p = 0.004). The combined decoder performed worse than LMP alone. These findings are limited by the small cohort, clinical and variable electrode coverage, uncorrected per-subject tests, and the design confound: correct key side matched visual target side, so visual and motor information cannot be separated. Sub-2 decoding was already high at −600 to −400 ms, which can precede the go cue for short reaction times and need not reflect motor planning. The exact immutable dataset release is not established; the retrieved S3 files are recorded by size and ETag in the archived provenance manifests.
stereo-EEG · SEEG · brain-computer interface · motor potential · beta power · high-gamma power · decoding · linear discriminant analysis · OpenNeuro
References
2- supportsPre-press SEEG power decoded cued key side in one of two patients (AUC 0.83); cue and key sides were confounded sci:sha256:1bb869883fbc3bc8741692188f0bcea8727678515326b4421c70c91847bbc8b3
Same sub-2 result (high AUC early before the key press) and same cue/key confound; this work adds 6 subjects and shows that LMP decodes sub-1, where beta and high gamma did not
- doi:10.1088/1741-2552%2Facae0a
Publication describing the OpenNeuro SEEG dataset used in this analysis.
Cited by
4 publicationsPublic catalog entries on Sciscape.
- supportsA detector trained on one arrow key detects the other key in SEEG, with non-inferior 70–150 Hz press-vs-rest decoding in eight subjects
Its result that beta and high-gamma power give near-chance key-side decoding (mean AUC 0.49 and 0.53) agrees with our finding that a band-power press-vs-rest detector trained on one key side transfers to the other side without loss. Its better LMP result is a limit of our band-power-only test.
- usesIn 8 SEEG patients, ERP covariance decoders had lower AUC than voltage decoders, while high-gamma covariance added no gain
We reuse its byte-range EDF fetcher, its trial matching, its bipolar montage and its mean-voltage LDA baseline. The baseline AUCs reproduced exactly.
- usesCausal 5 Hz low-pass filtering delays SEEG key-press decoding by 100 ms in three subjects, consistent with filter group delay
Fetch code, trial selection, bipolar montage and artifact rule are copied from this package.
- usesIn eight SEEG patients, only sub-2 showed significant cue-locked key decoding
We adapted its HTTP byte-range EDF fetch code and its LMP + shrinkage-LDA decoder. We test the cue/key confound that it states as a limit, and its observation of early (-600 ms) decoding in sub-2.