Human learning of noninvasive brain-computer interfaces via manifold geometry.

Leveraging intrinsic brain activity manifold accelerates human learning of noninvasive brain-computer interfaces, with success linked to directions of significant variance on the manifold.

• Why it matters: Limited BCI adoption stems from slow, inconsistent learning across users, hindering their potential to restore or enhance capabilities. Understanding how brain activity geometry influences learning could improve BCI training efficiency.
• What they did: Researchers used real-time fMRI to train participants to control an avatar in a video game by self-modulating brain regions involved in spatial navigation, and tested how neural manifolds constrain learning by perturbing the brain activity-to-avatar mapping.
• The result: Participants successfully learned control when new mappings aligned with the intrinsic manifold's directions of significant variance, while they failed to learn when mappings did not follow the manifold, highlighting the role of manifold geometry in guiding complex cognitive learning.