SEEG pairs with a high-gamma decrease during key press add decoding information beyond pairs with an increase (ds004473, 8 subjects)

At a fixed budget of 16 bipolar SEEG pairs, a shrinkage LDA using the 8 pairs with the largest 70–150 Hz power increases and the 8 with the largest decreases gave a higher key-press AUC than the 16 pairs with the largest increases in all 8 subjects (mean gain +0.064, median +0.050; exact two-sided sign-flip p = 0.0078). Pair selection and training used trials 1–100; testing used trials 101 onward. A hashed plan preceded the analysis. In an exploratory control, adding 8 null-response pairs to 8 increasing pairs gave no gain, whereas adding decreasing pairs did. These results support including strong decreases as well as increases when selecting pairs for this offline decoder. The scope is one session each from 8 epilepsy patients in one dataset, using fixed, cued windows; the signal may include cue, visual, or attention processes as well as movement.

stereo-EEG · SEEG · high-gamma · motor decoding · brain-computer interface · shrinkage LDA · electrocorticography