ePoster, COSYNE 2023
Identifying state-dependent interactions between brain regions during decision making
Orren Karniol-Tambour, E Mika Diamanti, E. Mika Diamanti, David Zoltowski, Lucas Pinto, Carlos Brody and 5 more
Princeton Neuroscience Institute; Stanford University; Statistics; Northwestern University; Department of Neuroscience
- Montreal, Canada
- Computational Neuroscience, Behavioral Neuroscience and
Abstract
Understanding how multiple brain regions interact to produce behavior is a major challenge in systems neuroscience, with many regions causally implicated in common tasks such as sensory processing and decision making. Evidence-based decision-making involves multiple brain regions across cortex, suggesting a distributed neural mechanism, but a precise description of interactions between regions remains an open problem. Addressing this problem requires developing new analysis methods that account for communication between areas, and pairing these with novel experimental approaches to capture neural activity across disparate brain regions during decision making. Here we developed MR-SDS, a multiregion, nonlinear switching state space model that decomposes global dynamics into local and cross-communication components in the latent space. Critically, the model allows us to estimate and quantify the directional “messages” communicated between regions at each timepoint. We first applied the model to a multiregion evidence accumulation simulation, finding the model accurately recovered latent states and dynamics. We then applied the model to multi-region, cellular-resolution two-photon mesoscope recordings. We simultaneously imaged excitatory neurons in 3 cortical areas, visual AM, premotor M2 and retrosplenial cortex, using a random-access two-photon microscope, while mice performed a sensory-accumulation decision making task. In preliminary analysis, our method revealed distinct local dynamics and communication profiles across regions, such as bi-directional communication between AM and M2, and feed-forward communication from AM to retrosplenial. Additionally, our analysis suggests sensory effects are weaker drivers of cortical dynamics than local recurrence and cross-region interactions, despite the sensory nature of the task. Thus, we introduce an important method for analyzing multiregion neural activity and communication, and showcase preliminary analysis of a novel multiregion dataset to help elucidate the role of different cortical regions in decision making.
Topics
- systems neuroscience, search World Wide
- Evidence-based decision-making, search World Wide
- MR-SDS, search World Wide
- nonlinear switching state space model, search World Wide
- multiregion neural activity, search World Wide
- two-photon mesoscope, search World Wide
- excitatory neurons, search World Wide
- premotor M2, search World Wide
Details
- Session
- Poster Session III