A connectivity-constrained computational account of topographic organization in primate high-level visual cortex

Inferotemporal (IT) cortex in humans and other primates is topographically organized, containing multiple hierarchically organized areas selective for particular domains, such as faces and scenes. This organization is commonly viewed in terms of evolved domain-specific visual mechanisms. Here, we develop an alternative, domain-general and developmental account of IT cortical organization. The account is instantiated in interactive topographic networks (ITNs), a class of computational models in which a hierarchy of model IT areas, subject to biologically plausible connectivity-based constraints, learns high-level visual representations optimized for multiple domains. We find that minimizing a wiring cost on spatially organized feedforward and lateral connections, alongside realistic constraints on the sign of neuronal connectivity within model IT, results in a hierarchical, topographic organization. This organization replicates a number of key properties of primate IT cortex, including the presence of domain-selective spatial clusters preferentially involved in the representation of faces, objects, and scenes; columnar responses across separate excitatory and inhibitory units; and generic spatial organization whereby the response correlation of pairs of units falls off with their distance. We thus argue that topographic domain selectivity is an emergent property of a visual system optimized to maximize behavioral performance under generic connectivity-based constraints.

Differential Expression of DSCAM Guides the Patterning of Retinal Axons Along Their Path and at Their Target in the Developing Xenopus Visual System

Background The Xenopus retinotectal circuit is organized topographically, where the dorsal-ventral axis of the retina maps respectively on to the ventral-dorsal axis of the tectum; axons from the nasal-temporal axis of the retina project respectively to the caudal-rostral axis of the tectum. Studies throughout the last two decades have shown that mechanisms involving molecular recognition of proper termination domains are at work guiding topographic organization. Such studies have shown that graded distribution of molecular cues is important for topographic mapping. However, the molecular cues organizing topography along the developing optic nerve, and as retinal axons cross the chiasm and navigate towards their target in the tectum, remain unknown. Down syndrome cell adhesion molecule (DSCAM) has been characterized as a key molecule in axon guidance, making it a strong candidate involved in the topographic organization of retinal fibers along the optic path.Methods Using a combination of whole-brain clearing and immunohistochemistry staining techniques we characterized DSCAM expression and the projection of ventral and dorsal retinal fibers starting from the eye, followed to the optic nerve into the chiasm, and into the terminal target in the optic tectum in Xenopus laevis tadpoles. We also assessed the effects of DSCAM on the establishment of retinotopic maps through spatially and temporally targeted DSCAM knockdown on retinal ganglion cells (RGCs) with axons innervating the optic tectum. Results Highest expression of DSCAM was localized to the ventral posterior region of the optic nerve and chiasm; this expression pattern coincides with ventral fibers derived from ventral RGCs. Downregulating DSCAM levels affected the segregation and proper sorting of medial axon fibers, derived from ventral RGCs, within the tectal neuropil, indicating that DSCAM plays a role in retinotopic organization. ConclusionThese findings together with the observation that DSCAM immunoreactivity accumulates on the primary dendrites of tectal neurons indicates that DSCAM exerts multiple roles in coordinating retinotopic order and connectivity in the developing vertebrate visual system.

Reset Networks: Emergent Topography in Networks of Convolutional Neural Networks

We introduce Reset networks, which are compositions of several neural networks - typically several levels of CNNs - where possibly non-spatial outputs at one level are reshaped into spatial inputs for the next level. We demonstrate that Reset networks exhibit emergent topographic organization for numbers, as well as for visual categories taken from CIFAR-100. We outline the implications of this model for theories of the cortex and developmental neuroscience.

Mapping Pontocerebellar Connectivity with Diffusion MRI

The cerebellum's involvement in cognitive, affective and motor functions is mediated by connections to different regions of the cerebral cortex. A distinctive feature of cortico-cerebellar loops that has been demonstrated in the animal work is a topographic organization that is preserved across its different components. Here we used diffusion imaging tractography to characterize the connections between the pons and the individual lobules of the cerebellum, and generate a classification of the pons based on its pattern of connectivity. We identified a rostral to caudal gradient in the pons, similar to that observed in the animal work, such that rostral regions were preferentially connected to cerebellar lobules involved in non-motor, and caudal regions with motor regions. These findings advance our fundamental understanding of the cerebellum, and the classifications we generated provide context for future research into the pontocerebellar tract's involvement in health and disease.

Naturalistic Audiovisual Stimulation Reveals the Topographic Organization of Human Auditory Cortex

Despite the importance of audition in spatial, semantic, and social function, there is no consensus regarding the detailed organisation of human auditory cortex. Using a novel computational model to analyse a high-powered naturalistic audiovisual movie-watching dataset, we simultaneously estimate spectral tuning properties and category selectivity to reveal the modes of organisation and computational motifs that characterise human auditory cortex. We find that regions more remote from the auditory core exhibit more compressive, non-linear response properties and finely-tuned, speech-selective receptive fields in low frequency portions of the tonotopic map. These patterns of organisation mirror aspects of the visual cortical hierarchy, wherein tuning properties progress from a stimulus category-agnostic front end towards more advanced regions increasingly optimised for behaviorally relevant stimulus categories.

A connectivity-constrained computational account of topographic organization in high-level visual cortex

Inferotemporal cortex (IT) in humans and other primates is topographically organized, with multiple domain-selective areas and other general patterns of functional organization. What factors underlie this organization, and what can this neural arrangement tell us about the mechanisms of high level vision? Here, we present an account of topographic organization involving a computational model with two components: 1) a feature-extracting encoder model of early visual processes, followed by 2) a model of high-level hierarchical visual processing in IT subject to specific biological constraints. In particular, minimizing the wiring cost on spatially organized feedforward and lateral connections within IT, combined with constraining the feedforward processing to be strictly excitatory, results in a hierarchical, topographic organization. This organization replicates a number of key properties of primate IT cortex, including the presence of domain-selective spatial clusters preferentially involved in the representation of faces, objects, and scenes, within-domain topographic organization such as animacy and indoor/outdoor distinctions, and generic spatial organization whereby the response correlation of pairs of units falls off with their distance. The model supports a view in which both domain-specific and domain-general topographic organization arise in the visual system from an optimization process that maximizes behavioral performance while minimizing wiring costs.

Continuity within somatosensory cortical map shapes the integration of optogenetic input

The topographic organization of sensory cortices is a prominent feature, but its functional role remains unclear. Particularly, how activity is integrated within a cortical area depending on its topography is unknown. Here, we trained mice expressing channelrhodopsin in cortical excitatory neurons to track a bar photostimulation that rotated smoothly over the primary somatosensory cortex (S1). When photostimulation was aimed at vS1, the area which contains a contiguous representation of the whisker array at the periphery, mice could learn to discriminate angular positions of the bar to obtain a reward. In contrast, they could not learn the task when the photostimulation was aimed at the representation of the trunk and legs in S1, where neighboring zones represent distant peripheral body parts, introducing discontinuities. Mice demonstrated anticipation of reward availability, specifically when cortical topography enabled to predict future sensory activation. These results are particularly helpful for designing efficient cortical sensory neuroprostheses.

Topographic Organization of Correlation Along the Longitudinal and Transverse Axes in Rat Hippocampal CA3 Due to Excitatory Afferents

The topographic organization of afferents to the hippocampal CA3 subfield are well-studied, but their role in influencing the spatiotemporal dynamics of population activity is not understood. Using a large-scale, computational neuronal network model of the entorhinal-dentate-CA3 system, the effects of the perforant path, mossy fibers, and associational system on the propagation and transformation of network spiking patterns were investigated. A correlation map was constructed to characterize the spatial structure and temporal evolution of pairwise correlations which underlie the emergent patterns found in the population activity. The topographic organization of the associational system gave rise to changes in the spatial correlation structure along the longitudinal and transverse axes of the CA3. The resulting gradients may provide a basis for the known functional organization observed in hippocampus.

Topographic heterogeneity of intrinsic excitability in mouse hippocampal CA3 pyramidal neurons

Area CA3 is a major hippocampal region that is classically thought to act as a homogeneous neural network vital for spatial navigation and episodic memories. Here, we report that CA3 pyramidal neurons exhibit marked heterogeneity of somatodendritic morphology and cellular electrical properties along both proximodistal and dorsoventral axes. These new results uncover a complex, yet orderly, pattern of topographic organization of CA3 neuronal features that may contribute to its in vivo functional diversity.