Spreading SEEG bipolar pairs across shafts improved key press decoding at 4, 8, and 16 pair budgets in eight subjects

We tested whether the higher decoding performance previously observed when spreading stereoelectroencephalography (SEEG) bipolar pairs across shafts continues as the channel budget increases. Using archived bipolar log-power features (70–150 Hz and 13–30 Hz) from eight subjects in the cued key-press task, we compared 200 random dense placements (contiguous pairs on the fewest shafts) with 200 random spread placements (one pair per shaft, with a second round when needed) at budgets of 4, 8, and 16 pairs. Shrinkage linear discriminant analysis was trained on trials 1–100 and tested on trials 101 onward. Mean area under the ROC curve (AUC) was higher for spread placements at every budget: 0.629 versus 0.589 at 4 pairs, 0.684 versus 0.631 at 8 pairs, and 0.737 versus 0.677 at 16 pairs. The difference was positive in all eight subjects at each budget (exact two-sided sign-flip p = 0.0078; Holm-adjusted p = 0.023). It increased by 0.010 AUC per budget doubling (seven of eight subjects; sign-flip p = 0.023). These results show that, under random placement, more shafts provided more information than more contacts on one shaft, and the effect did not decrease up to 16 pairs. The study was offline and used one session, one task, and eight subjects; shafts are not a direct measure of anatomical distance. Placement was random: selected pairs can remove the spread benefit.

stereoelectroencephalography · SEEG · bipolar pairs · channel placement · key press decoding · shrinkage LDA · area under the ROC curve · high gamma · beta band