Six SEEG channels showed lower spontaneous band-power dimensionality than six ECoG channels in 16 epilepsy patients, with unmatched spatial spread

For brain-computer interfaces that read band power, the number of independent signals per channel sets a useful channel count. We compared SEEG depth and ECoG surface channels in the same patients and at the same channel count using 60 seconds of NREM sleep from OpenNeuro ds007119. All 103 listed subjects were read; 16 patients had at least six non-overlapping normal bipolar channels of each type. For 70–110 Hz envelopes, six SEEG channels had median participation ratio (PR) 5.16 versus 5.57 for ECoG; SEEG PR was lower in 14/16 patients (paired sign-flip p=0.0011). The PR deficit from a circular time-shift null was 0.780 versus 0.354 effective channels (p=0.00077; median deficit ratio 2.21). For 4–12 Hz, PR was 4.786 versus 5.640 (lower in 16/16, p=3×10⁻⁵), with deficits of 1.120 versus 0.276 (p=4×10⁻⁵; ratio 4.05). With ten channels in five patients, the 70–110 Hz PR difference was absent (8.808 versus 8.831, p=0.69); the 4–12 Hz comparison was lower for SEEG in all five (p=0.062). Both channel types remained near the independent-signal ceiling. Spatial spread was not matched: SEEG channels came from fewer leads and had smaller contact spacing than ECoG channels drawn from grids and strips. The findings concern channels as implanted, not tissue per unit distance. They describe 60 seconds of spontaneous NREM activity in epilepsy patients; PR measures redundancy, not task-related or decodable information. The ten-channel analysis included only five patients.

SEEG · ECoG · band power · effective dimensionality · participation ratio · intracranial EEG · neural signal redundancy · NREM sleep · epilepsy · brain-computer interfaces