Surface circumferential spinal cord recording in freely moving rodents.

Spinal cord injury affects over 2.5 million people worldwide, yet current neuroprosthetic strategies remain fragmented, addressing motor, sensory, or autonomic function in isolation. Here we show that a single ultrathin circumferential electrode array, conforming to the spinal cord without penetrating neural tissue, can simultaneously decode motor intent, classify sensory inputs, and discriminate visceral sensory inputs. In freely moving rats during short-term implantation (up to three days), deep learning decoders achieved robust motor intent decoding (R² = 0.97) by exploiting low-frequency spinal oscillations aligned with central pattern generator rhythms. The same interface classified eight sensory modalities with 94.4% accuracy. In acutely anaesthetized pigs, cross-species validation confirmed translational scalability and reliably distinguished visceral sensory inputs. Uniquely, the two-row electrode configuration resolved directional propagation within spinal tracts while electrode-dense one-row devices enabled high-precision intraspinal source localization. By consolidating motor, sensory, and visceral afferent decoding within a single conformal interface, this approach positions the spinal cord as a target for multifunctional neuroprosthetic interfacing, offering a path toward integrated restoration of physiological function after neurological injury.