Researchers at University Medical Center Utrecht's Brain Center in the Netherlands and collaborators exhaustively tested every rectangular subgrid inside 32-, 64- and 128-channel ECoG arrays recorded from nine people with epilepsy. Grid area could be cut by 75% to 94% without meaningful loss of hand-movement classification accuracy, as long as the remaining electrodes sat over informative cortex; below a critical area of about 60 mm², performance fell sharply. The study appeared in the journal Neuroinformatics on August 7, 2026.
