Keynotes | K&Ts | GACs | Talks | Posters | Search

Poster E in Poster Session E: Thursday, August 6, 10:30 am – 12:15 pm, Kimmel Center, Shorin & Rosenthal Rooms

Diffusion-Based Stimulus Optimization Reveals Fine-Grained Feature Selectivity in Human Early Visual Cortex

Junru Zhao¹, Andrew F. Luo², Margaret M Henderson¹; ¹Carnegie Mellon University, ²University of Hong Kong

Presenter: Junru Zhao

Using generative image models to synthesize a most-exciting-input (MEI) for target neural populations provides a powerful tool for probing visual cortex tuning. While this approach has been successfully applied to human higher visual cortex using fMRI data, generating MEIs for early retinotopic visual areas requires additional modeling constraints, due to the small receptive fields in these areas. To address this challenge, we present a novel diffusion-based MEI generation framework for fMRI data, which incorporates a pRF-constrained voxelwise encoding model alongside a pretrained latent diffusion model, enabling optimization of single-voxel responses. Preliminary results suggest that this approach is promising for generating interpretable, localized MEIs in early visual cortex.

Topic Area: Computational Models of Vision & Visual Cortex