Neural Coding of Visual Stimuli

National Institute Of Mental Health

Investigators

Linked publications & trials

Abstract

Primates, including old-world monkeys, can categorize images quickly based on similarity of visual features. We previously showed that two subregions of inferior temporal cortex, areas TEO and TE, contribute to visual categorization to differing extents. Monkeys with TE removals show a more sustained deficit than monkeys with TEO removals in the categorization of visual images. To learn more about how this difference in behavior arises, we recorded simultaneously from TE and TEO using implanted 64- or 96- channel Utah arrays while two old world monkeys were learning a visual categorization task. We found that the activity of neurons in TE, not TEO, becomes more strongly related to image category during learning; this increasing modulation relationship is driven primarily by stronger encoding of the visual category associated with a larger reward. Using a one-layer neural network (no hidden layer) for population-level decoding, we found that decoding accuracy of TE neurons, not TEO neurons, increases monotonically during learning, approximately paralleling the monkey's behavior. Furthermore, decoding accuracy of category in correct trials is consistently higher than error trials across days in TE, but not TEO. Consistent with the findings from lesions studies, our results suggest that area TE is a key brain region for visual category learning.