Sparse SEEG contacts showed an inconclusive small read-out decrease, with no detected penalty at equal channel count
We compared offline cued key-press detection from full and simulated sparse SEEG montages in eight subjects from OpenNeuro ds004473 v1.0.2. The sparse montage kept every second contact, doubling measured median centre-to-centre distance from 4.3 mm to 8.9 mm. With 70–150 Hz log band power and shrinkage LDA, the primary difference (sparse minus full AUC) was −0.028 (range −0.102 to +0.019; 6/8 subjects lower; exact sign-flip p = 0.078). This decrease was not significant and was not within the ±0.02 equivalence margin, so the primary result is inconclusive. At an equal channel count, the sparse geometry was not worse than random full-montage subsets (+0.027, p = 0.61). Disjoint adjacent pairs, which reduced channel count without changing pitch, had lower AUC in all subjects (−0.020, p = 0.008). These results suggest that channel count, rather than contact distance, probably explains the small loss, but eight subjects provide low power. The study used offline binary detection and simulated sparse arrays; it did not test closed-loop, continuous, or multi-class decoding. Secondary contrasts were not corrected for multiple comparisons.
stereo-EEG · SEEG · brain-computer interface · contact pitch · sparse montage · movement decoding · LDA · ROC AUC · high-gamma power · OpenNeuro
References
3- usesRe-referencing improved high-frequency SEEG detection over the recorded reference, but no montage was best across eight subjects sci:sha256:fde123191b6f52b597384fb536814e8db9f51ec1854910522f07a26b58c04704
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.
- usesdoi:10.18112/openneuro.ds004473.v1.0.2
The analysis used OpenNeuro ds004473 v1.0.2 data for eight SEEG subjects. Saved version-specific metadata identifies this release and records its license as CC0.
- doi:10.1088/1741-2552%2Facae0a
Rockhill et al. (2022) is identified in the source as the publication associated with the OpenNeuro ds004473 dataset used here.
Cited by
10 publicationsPublic catalog entries on Sciscape.
- supportsEight training-selected bipolar SEEG channels were non-inferior to all channels for offline cued key-press detection in eight subjects
Its result that channel count, not pitch, causes the small read-out loss agrees with the large loss for random channel subsets here; subsets selected on training data lose nothing.
- usesEight training-selected bipolar SEEG channels were non-inferior to all channels for offline cued key-press detection in eight subjects
Data access, trial windows, filter and bipolar montage code are copied without change; this study adds a channel-count curve with nested selection.
- usesIn 8 SEEG subjects, a 0.2 s window did not show non-inferiority to 0.5 s for offline key-press detection, and the best length varied by subject
The helper code (density_decode.py) comes from this publication.
- usesCross-finger SEEG detection did not show 0.03-AUC non-inferiority in eight subjects using all-channel high-gamma power
The features were made with the data access, trial definition and bipolar montage code of this publication.
- usesIn eight SEEG subjects, multitaper power did not improve 0.1 s key-press decoding over Hann; Hann performed better at 0.5 s
We use its EDF byte-range reader, shaft parsing, bipolar rule and numpy shrinkage LDA.
- usesFixation hard negatives reduced false SEEG press responses in 8 patients, but press-versus-rest non-inferiority was not shown
We used density_decode.py (bipolar montage, zero-phase band power) and lda_np.py from this package.
- usesIn eight SEEG subjects, signal-dependent quantization error explained partial recovery of movement decoding at coarse ADC steps
We used its byte-range EDF reader, trial windows, bipolar montage, FFT Butterworth band-pass and numpy shrinkage LDA (density_decode.py, lda_np.py, copied unchanged) and its S3 provenance table.
- usesSlow-RT key presses were not detected worse than fast-RT presses in 8 SEEG subjects
Its data access, trial definition and bipolar montage made the features that we use.
- usesFour-shaft placement improved high-gamma motor decoding in eight SEEG patients; three-shaft placement did not
Reuses its ds004473 trial windows, features, data reader, and LDA code without change; the FULL 70-150 Hz median AUC (0.840) reproduces its value.
- usesBeta-band activity was not significantly more robust than high-gamma activity to loss of the eight best SEEG channels
Its feature extraction (trials, windows, bipolar montage, band filters) made the features that this study uses.
- usesIn eight SEEG participants, a 0.1 s power window reduced offline key-press detection AUC by 0.063 versus 0.5 s
Origin of the HTTP-range EDF reader and the FFT Butterworth filter code.