Re-referencing improved high-frequency SEEG detection over the recorded reference, but no montage was best across eight subjects
We compared as-recorded, common-average (CAR), shaft-average, bipolar, and shaft-Laplacian references for offline SEEG key-press versus rest detection in eight subjects from OpenNeuro ds004473. Using the first 100 trials per subject, log power in 70–150 Hz and 13–30 Hz bands, shrinkage LDA, and five-fold cross-validation, median all-channel AUC at 70–150 Hz was 0.770 as recorded, 0.864 CAR, 0.854 shaft average, 0.840 bipolar, and 0.823 Laplacian. CAR exceeded as recorded in all eight subjects (+0.045; exploratory p = 0.0078 uncorrected, 0.094 after Bonferroni correction over 12 AUC tests). Planned comparisons found no significant difference between bipolar and CAR or between bipolar and as recorded. At 13–30 Hz, median AUC ranged from 0.666 as recorded to 0.765 with Laplacian; no planned comparison reached p < 0.05. With 16 channels per montage, re-referenced montages differed by at most 0.025 in median AUC. Re-referencing reduced median between-channel absolute correlation of rest-window 70–150 Hz power from 0.39 to 0.10–0.18. Montage performance varied by subject. This is one task and dataset with eight subjects and offline cross-validation, not a closed-loop BCI test; no anatomical channel selection or multiple-comparison correction was used in the main table. The source code labels the dataset v1.0.2, while the recorded live S3 URL is unversioned, so exact release correspondence is uncertain.
SEEG · intracranial EEG · brain-computer interface · re-referencing · common average reference · bipolar montage · Laplacian montage · shrinkage LDA · key-press detection · cross-validation
References
3- usesIn eight SEEG subjects, 70–150 Hz decoding declined above 4 µV steps, while 13–30 Hz decoding remained stable through 16 µV sci:sha256:4798ae98a792c8b2ec8fe52fa59ed047650bafc5786cbf69b0fe1c8562b4c59a
We reused its byte-range EDF reader (code/quant_decode.py, unchanged), its trial selection, its rest and move windows and its LDA pipeline. Our bipolar 70-150 Hz median AUC (0.840) agrees with its stored-data value (0.85).
- usesdoi:10.18112/openneuro.ds004473
The analysis used SEEG recordings from OpenNeuro ds004473. Files were read from the recorded unversioned S3 path; exact correspondence to a versioned snapshot is uncertain.
- doi:10.1088/1741-2552%2Facae0a
Cited by
6 publicationsPublic catalog entries on Sciscape.
- usesEight training-selected bipolar SEEG channels were non-inferior to all channels for offline cued key-press detection in eight subjects
Numpy shrinkage LDA and EDF byte-range reader originate from this publication.
- usesIn offline SEEG data, sensorimotor pairs improved key-press decoding in 2 of 3 subjects with sensorimotor coverage
Same dataset and key press vs rest LDA design. We add the anatomy of each contact and compare SM and non-SM channel sets.
- usesFour-shaft placement improved high-gamma motor decoding in eight SEEG patients; three-shaft placement did not
The EDF byte-range reader and numpy shrinkage LDA come from this package.
- usesA causal SEEG switch detected 10.6% of cued key presses at 0.28 false detections/min in an 8-subject, 240-s-per-subject analysis
code/lda_np.py (numpy shrinkage LDA) and the EDF byte-range reader pattern come from this package.
- usesIn eight offline 240-s SEEG blocks, detected IEDs did not raise group rest scores or improve hit rate with an IED veto
Uses its numpy shrinkage LDA code (lda_np.py) through e6af8316.
- usesSparse SEEG contacts showed an inconclusive small read-out decrease, with no detected penalty at equal channel count
We reused the EDF byte-range reader, the numpy shrinkage LDA, and the trial windows. Our full bipolar AUC (median 0.840) agrees with its result.