P310 Feedback input to apical dendrites of L5 pyramidal cells leads to a shift towards a resonance state in V1 cortex
Francescangelo Vedele*1, Margaux Calice2, Simo Vanni1
1Department of Physiology, Medicum, University of Helsinki, Helsinki, Finland 2Centre Giovanni Borelli - CNRS UMR 9010, Université Paris Cité, France
*Email: francescangelo.vedele@helsinki.fi Introduction:To make sense of the abundance of visual information coming in from the outside world, cortical and subcortical structures operate on stored models of the environment that are constantly compared with new information[1]. The cortical structures for vision are tightly interconnected and rely on multiple subregions to capture different facets of information. The SMART model by Grossberg and Versace[2]aims to build a simulation framework to provide a circuit-level perspective on learning, expectation, and processing of visual information in the brain. While cellular details are well understood at the microscopic level, computations linking visual system states to higher-order processes are scarce.
Methods:The macaque was chosen as a biological model because of its close evolutionary relationship to humans[3]. Computer simulations of macaque cortical patches were implemented using CxSystem2[4,5], a cortical simulation framework based on Brian2[6]. The SMART model includes cells in V1 (layers L2/3, L4e, L5, and L6), dendrites of compartmental neurons reaching L1, and thalamic specific, nonspecific, and reticular nuclei. Simulations were run for a duration of 2 seconds. Spike times and cell membrane voltage were monitored. Power spectral density (PSD) spectra of membrane voltage were obtained using Welch’s method. A feedback current of 1.5x or 2.5x the rheobase was injected into the apical dendrite of L5 pyramidal cells (located in L1).
Results:The SMART model was first simulated with bottom-up sensory input and a weak feedback current. In this state, all layers output in short (~150ms) bursts followed by longer periods of oscillatory activity (~500ms). The PSD plots show a broad, low-frequency peak in the alpha/beta frequency bands (up to 30Hz) across layers. Upon injection of a stronger feedback current, the model shifts to a resonance mode characterized by higher firing rates and a broad PSD peak in the gamma range (20-70 Hz) across layers. Therefore, strong feedback input shifts the state of the system, from resting to a high-frequency resonance mode. This might be related to population synchrony, which may bind features in different parts of the visual field[7].
Discussion:The SMART model provides a flexible way to model cortical coordination and feedback. Our simulations show how a weak input from higher cortical areas leaves the system in a disengaged state, akin to a mismatch between expectation and reality. By injecting a strong current to mimic feedback from higher cortical areas, the simulated system enters a resonant state as in the biological brain, establishing a condition that supports learning and plasticity. While this model is informative when studying single-region cortical dynamics, we plan to integrate V2 and V5 with the current model of V1, aiming to simulate hierarchical cortical processing of visual information.
Acknowledgements This work was supported by Academy of Finland project grant 361816. References
