Gaze Decoding with Sensory and Motor Cortical Activity

Eye movements require neuronal processing of visual stimuli and transmission of motor commands to the eye muscles. The brain circuitry and cortical regions responsible for sensory-motor transformations are generally understood, but deciphering the specific neuronal patterns underlying particular eye movements remains challenging. We used high-resolution eye tracking, invasive neuronal recordings, and machine learning to reveal, for the first time, the information encoded within neuronal populations in sensory and motor cortical areas that governs gaze behavior. We achieved high prediction accuracy when decoding continuous eye position using spiking activity. Decoders trained with only sensory or motor activity showed roughly equal performance, and we explored other methods of combining activity into a single model to enhance performance. Our results estimate the upper bound on decoding accuracy using brain recordings, shed light on sensory and motor cortical signals for gaze decoding, and offer insights into harnessing spatially broader brain signals for brain-computer interfaces.