PhD defense of Margaux CALICE

Margaux CALICE

A theoretical multiscale examination of visual spatial attention in the mouse cerebral cortex

A major challenge in neuroscience is to understand how brain functions emerge from the interaction between cellular and network properties. Theoretical studies addressing this subject have often relied on abstract descriptions that neglected biophysical details or didn’t integrate them across different scales. To efficiently test mechanistic hypotheses regarding cognitive functions such as visual spatial attention (VSA), we need to develop new modeling strategies that integrate biological constraints at each level of neural organization: subcellular, cellular, population, and neuronal network at a large scale.

In this work, we revisit the Synchronous Matching Adaptive Resonance Theory (SMART; Grossberg \& Versace, 2008), which suggests that attentional processes depend on thalamocortical networks conducting spatial comparisons between top-down attentional and bottom-up sensory signals. We predict that the primary visual area (V1) serves as the site where a match or a mismatch between these signals occurs, respectively reinforcing tonic activity through specific gamma-range dynamics or inducing a reset of the ongoing activity through spiking bursts in a low frequency range, enabling the testing of new comparisons. To investigate this hypothesis, we start by developing a conceptual model of V1 and its top-down and bottom-up signals. We then develop a data-driven modeling approach to construct a large-scale, biophysically mimetic spiking neuron network model of the mouse cortex. It is specifically designed to support the study of VSA through a biologically plausible network architecture inspired by SMART, with a topographic organization that enables the spatial comparison of top-down and bottom-up signals. For this purpose, we contribute to the expanding ecosystem of tools for brain network modeling by developing methods that facilitate the integration of evolving biophysical and connectivity databases, and that enable efficient construction and evaluation of large-scale networks, allowing for hybrid networks with varying degrees of explicit detail. In summary, this thesis offers new insights and perspectives on the challenges of reconciling a multiscale implementation with robust data integration for theory testing. It also lays the foundation for refining future implementation strategies and guiding experimental work on VSA.

Direction

• Lyle J. Graham, Docteur, Directeur de Thèse

Jury 

• Laurent Perrinet, Docteur, Rapporteur
• Boris Gutkin, Docteur, Rapporteur
• Lorenzo Fontolan, Docteur, Examinateur
• María Victoria Sánchez-Vives, Professeur, Examinatrice