Dr. Tomas Knapen | Neural computation based in sensory topographies (hybrid meeting)

Neighboring locations in the sensorium are often represented at neighboring locations in the brain. These topographic maps reflect a fundamental mode of neural organization: for vision alone, over 30 retinotopic maps mirroring the retinal input array have been found and charted in the human brain. But how is this fundamental neural architecture used to facilitate behaviorally relevant computation? In this talk, I will first detail how the combination of UHF fMRI and computational modeling facilitated the discovery of several of these visual maps, in cerebellum, default mode network, and hippocampus. I will then show how so-called “population receptive field” models explain highly diverse response properties by invoking the canonical computational mechanism of divisive normalization. These detailed modeling results then serve as a bridge between human 7T fMRI and high-density non-human primate electrophysiology data from a virtually identical experiment. Lastly, if the brain analyzes a sensory stimulus in parallel across a multitude of maps that share a sensory reference frame, then this must shape the detailed structure of functional connectivity. We can use this notion to investigate how the brain performs higher-level cognitive operations such as multisensory integration in naturalistic contexts. I will show that by simultaneously tracking topographic connectivity from both V1 and S1 during movie-watching, high-level visual BOLD responses are predominantly explained by somatosensory and not visual connectivity. Moreover, the detailed somatosensory tuning in these distinct regions shows that they mirror the full somatosensory homunculus inside visual cortex. Together, these findings show that the topographic organization of the human brain provides a productive framework to understand its detailed computational properties, and an avenue into understanding more high-level cognition from sensory foundations.