Contralateral SEEG channels did not improve key press detection in 8 subjects at equal channel count

Question: do SEEG channels in the hemisphere contralateral to the response hand detect a key press better than ipsilateral channels? Data: OpenNeuro ds004473 version 1.0.2, all 8 subjects, all with contacts in both hemispheres. Patients pressed arrow keys with the dominant hand (Rockhill et al. 2023). Features: archived 70–150 Hz (HG) and 13–30 Hz log power from sci 04683a2a, using adjacent-bipolar channels. Hemisphere was assigned from the first letter of each channel name. A shrinkage LDA was trained on trials 1–100 and tested on trials 101 onward (movement vs rest windows, AUC). The primary test used equal channel counts, with 200 random subsets from the larger hemisphere. The analysis plan was hashed before analysis. Contralateral minus ipsilateral HG AUC had a mean of +0.012, median +0.010, was positive in 4 of 8 subjects, and had an exact two-sided sign-flip p = 0.80. Per-subject differences ranged from −0.213 to +0.196. Results were also not significant with HG+B, all channels, or the strongest single channel. The strongest single channel was contralateral in 4 subjects and ipsilateral in 4; in sub-11 it was on the ipsilateral SMA. Both hemispheres together had higher AUC than contralateral channels alone (mean difference +0.058, 6 of 8, p = 0.047, not corrected for seven contrasts). In this clinical sample, sampled structure appeared to matter more than hemisphere. A SEEG movement detector should not drop ipsilateral contacts. Limits: 8 subjects and low power; the result does not establish equal performance. Regions were unmatched, hemisphere was inferred from names, analysis was offline, and one session was used per subject.

stereoelectroencephalography · SEEG · motor brain-computer interface · key press detection · hemispheric laterality · high gamma · shrinkage LDA · OpenNeuro ds004473