Search Papers | Poster Sessions | All Posters
Poster B154 in Poster Session B - Thursday, August 8, 2024, 1:30 – 3:30 pm, Johnson Ice Rink
Exploring Brain Responses to Memorability-Controlled Generated Images
Hyewon Willow Han1,2, Mansoure Jahanian1, Johann Cardenas1, Yalda Mohsenzadeh1,2 (); 1Western University, 2Vector Institute for Artificial Intelligence
Some images are more memorable than others, but the underlying neural mechanisms of memorability are not fully understood. This study investigates how the human brain processes images based on their memorability using fMRI responses to natural images by employing advanced deep neural network (DNN) and generative adversarial network (GAN) models. Our work successfully manipulated the memorability of images in both increasing and decreasing directions and discovered that brain regions associated with the face and body exhibited differential activity based on the memorability alterations, with specific activity patterns emerging in early visual areas. The amygdala responded to changes in both directions, whereas the hippocampus was primarily responsive when memorability decreased. Notably, place-associated areas showed reverse responses, being less active when images with increased memorability were presented. This investigation contributes to our understanding of the cognitive processes involved in visual memory, demonstrating the potential of integrating generative neural network models with neuroimaging to study brain function, and paving the way for formulation of experimentally testable hypotheses.
Keywords: Memorability Generative Adversarial Networks Visual Memory