Modeling Localized Synaptic Degeneration and Neural Plasticity in Visual Cortex

Yash Shah, Kevin Tran, Daniel LK Yamins

Abstract: A patient diagnosed with a neurodegenerative disease of the visual system like Posterior Cortical Atrophy (PCA) or Visual Agnosia can have difficulty in recognizing faces, perceiving boundary shapes, discriminating color, and reading fragmented images. Computationally modeling the effects of neurodegeneration on different areas in the ventral visual stream (V1, V2, V4, and IT), while maintaining the mechanism of neural plasticity in the brain, can help inform effective rehabilitation strategies for the patient. To this end, we simulate localized synaptic degeneration and neural plasticity using CORnet-S—a deep artificial model of the brain. We observe that (1) different areas in the visual cortex have functionally different re- sponses to tasks such as stimuli processing, shape discrimination, and face perception; and (2) even when a certain area in the visual cortex is ∼99.85% progressively lesioned, the cortex has a remarkable ability to recover task performance through perceptual learning.