Mammalian Visual System Research Articles

Dysregulation of excitatory neural firing replicates physiological and functional changes in aging visual cortex.

The mammalian visual system has been the focus of countless experimental and theoretical studies designed to elucidate principles of neural computation and sensory coding. Most theoretical work has focused on networks intended to reflect developing or mature neural circuitry, in both health and disease. Few computational studies have attempted to model changes that occur in neural circuitry as an organism ages non-pathologically. In this work we contribute to closing this gap, studying how physiological changes correlated with advanced age impact the computational performance of a spiking network model of primary visual cortex (V1). Our results demonstrate that deterioration of homeostatic regulation of excitatory firing, coupled with long-term synaptic plasticity, is a sufficient mechanism to reproduce features of observed physiological and functional changes in neural activity data, specifically declines in inhibition and in selectivity to oriented stimuli. This suggests a potential causality between dysregulation of neuron firing and age-induced changes in brain physiology and functional performance. While this does not rule out deeper underlying causes or other mechanisms that could give rise to these changes, our approach opens new avenues for exploring these underlying mechanisms in greater depth and making predictions for future experiments.

Survey of spiking in the mouse visual system reveals functional hierarchy.

The anatomy of the mammalian visual system, from the retina to the neocortex, is organized hierarchically1. However, direct observation of cellular-level functional interactions across this hierarchy is lacking due to the challenge of simultaneously recording activity across numerous regions. Here we describe a large, open dataset-part of the Allen Brain Observatory2-that surveys spiking from tens of thousands of units in six cortical and two thalamic regions in the brains of mice responding to a battery of visual stimuli. Using cross-correlation analysis, we reveal that the organization of inter-area functional connectivity during visual stimulation mirrors the anatomical hierarchy from the Allen Mouse Brain Connectivity Atlas3. We find that four classical hierarchical measures-response latency, receptive-field size, phase-locking to drifting gratings and response decay timescale-are all correlated with the hierarchy. Moreover, recordings obtained during a visual task reveal that the correlation between neural activity and behavioural choice also increases along the hierarchy. Our study provides a foundation for understanding coding and signal propagation across hierarchically organized cortical and thalamic visual areas.

A Trainable Monogenic ConvNet Layer Robust in Front of Large Contrast Changes in Image Classification

Convolutional Neural Networks (ConvNets) at present achieve remarkable performance in image classification tasks. However, current ConvNets cannot guarantee the capabilities of the mammalian visual systems such as invariance to contrast and illumination changes. Some ideas to overcome the illumination and contrast variations usually have to be tuned manually and tend to fail when tested with other types of data degradation. In this context, we present a new bio-inspired entry layer, M6, which detects low-level geometric features (lines, edges, and orientations) which are similar to patterns detected by the V1 visual cortex. This new trainable layer is capable of coping with image classification even with large contrast variations. The explanation for this behavior is the monogenic signal geometry, which represents each pixel value in a 3D space using quaternions, a fact that confers a degree of explainability to the networks. We compare M6 with a conventional convolutional layer (C) and a deterministic quaternion local phase layer (Q9). The experimental setup is designed to evaluate the robustness of our M6 enriched ConvNet model and includes three architectures, four datasets, three types of contrast degradation (including non-uniform haze degradations). The numerical results reveal that the models with M6 are the most robust in front of any kind of contrast variations. This amounts to a significant enhancement of the C models, which usually have reasonably good performance only when the same training and test degradation are used, except for the case of maximum degradation. Moreover, the Structural Similarity Index Measure (SSIM) is used to analyze and explain the robustness effect of the M6 feature maps under any kind of contrast degradations.

Retinal ganglion cell interactions shape the developing mammalian visual system.

ABSTRACTRetinal ganglion cells (RGCs) serve as a crucial communication channel from the retina to the brain. In the adult, these cells receive input from defined sets of presynaptic partners and communicate with postsynaptic brain regions to convey features of the visual scene. However, in the developing visual system, RGC interactions extend beyond their synaptic partners such that they guide development before the onset of vision. In this Review, we summarize our current understanding of how interactions between RGCs and their environment influence cellular targeting, migration and circuit maturation during visual system development. We describe the roles of RGC subclasses in shaping unique developmental responses within the retina and at central targets. Finally, we highlight the utility of RNA sequencing and genetic tools in uncovering RGC type-specific roles during the development of the visual system.

Neuroanatomy and Sampling of Central Projections for the Visual System in Mammals Used in Toxicity Testing.

Visual system toxicity may manifest anywhere in the visual system, from the eye proper to the visual brain. Therefore, effective screening for visual system toxicity must evaluate not only ocular structures (ie, eye and optic nerve) but also multiple key brain regions involved in vision (eg, optic tract, subcortical relay nuclei, and primary and secondary visual cortices). Despite a generally comparable pattern across species, the neuroanatomic organization and function of the visual brain in rodents and rabbits exhibit appreciable differences relative to nonrodents. Currently recognized sampling practices for general toxicity studies in animals, which are based on easily discerned external neuroanatomic landmarks and guided by extant stereotaxic brain atlases, typically will permit histopathologic evaluation of many brain centers involved in visual sensation (eg, optic chiasm, optic tract, dorsal lateral geniculate nucleus, primary and secondary visual cortices) and often some subcortical brain nuclei involved in light-modulated nonvisual activities needed for visual attention and orientation (eg, rostral colliculus in quadrupeds, termed the superior colliculus in bipeds; several cranial nerve nuclei). Pathologic findings induced by toxicants in the visual brain centers are similar to those that are produced in other brain regions.

Developmental distribution of primary cilia in the retinofugal visual pathway

The mammalian visual system is composed of circuitry connecting sensory input from the retina to the processing core of the visual cortex. The two main retinorecipient brain targets, the superior colliculus (SC) and dorsal lateral geniculate nucleus (dLGN), bridge retinal input and visual output. The primary cilium is a conserved organelle increasingly viewed as a critical sensor for the regulation of developmental and homeostatic pathways in most mammalian cell types. Moreover, cilia have been described as crucial for neurogenesis, neuronal maturation, and survival in the cortex and retina. However, cilia in the visual relay center remain to be fully described. In this study, we characterized the ciliation profile of the SC and dLGN and found that the overall number of ciliated cells declined during development. Interestingly, shorter ciliated cells in both regions were identified as neurons, whose numbers remained stable over time, suggesting that cilia retention is a critical feature for optimal neuronal function in SC and dLGN. Our study suggests that primary cilia are important for neuronal maturation and function in cells of the SC and dLGN.

Comparing biological and artificial vision systems: Network measures of functional connectivity

Advances in Deep Convolutional Neural Networks (DCNN) provide new opportunities for computational neuroscience to pose novel questions regarding the function of biological visual systems. Some attempts have been made to utilize advances in machine learning to answer neuroscientific questions, but how to appropriately make comparisons between the biological systems and artificial neural network structure is an open question. This analysis quantifies network properties of the mouse visual system and a common DCNN model (VGG16), to determine if this comparison is appropriate. Utilizing weighted graph-theoretic measures of node density (weighted node-degree), path length, local clustering coefficient, and betweenness, differences in functional connectivity patterns in the modern artificial computer vision system and the biological vision system are quantified. Results show that the mouse exhibits network measure distributions more similar to Poisson than normal, whereas the VGG16 exhibits network measure distributions with a more Gaussian shape than the sampled biological network. The artificial network shows higher density measures and shorter path lengths in comparison to the biological network. These results show that training a VGG16 for an object recognition task is unlikely to produce a network whose functional connectivity is similar to the mammalian visual system.

Cortical Inactivation Does Not Block Response Enhancement in the Superior Colliculus.

Repetitive visual stimulation is successfully used in a study on the visual evoked potential (VEP) plasticity in the visual system in mammals. Practicing visual tasks or repeated exposure to sensory stimuli can induce neuronal network changes in the cortical circuits and improve the perception of these stimuli. However, little is known about the effect of visual training at the subcortical level. In the present study, we extend the knowledge showing positive results of this training in the rat’s Superior colliculus (SC). In electrophysiological experiments, we showed that a single training session lasting several hours induces a response enhancement both in the primary visual cortex (V1) and in the SC. Further, we tested if collicular responses will be enhanced without V1 input. For this reason, we inactivated the V1 by applying xylocaine solution onto the cortical surface during visual training. Our results revealed that SC’s response enhancement was present even without V1 inputs and showed no difference in amplitude comparing to VEPs enhancement while the V1 was active. These data suggest that the visual system plasticity and facilitation can develop independently but simultaneously in different parts of the visual system.

Improved object recognition using neural networks trained to mimic the brain’s statistical properties

The current state-of-the-art object recognition algorithms, deep convolutional neural networks (DCNNs), are inspired by the architecture of the mammalian visual system, and are capable of human-level performance on many tasks. As they are trained for object recognition tasks, it has been shown that DCNNs develop hidden representations that resemble those observed in the mammalian visual system (Razavi and Kriegeskorte, 2014; Yamins and Dicarlo, 2016; Gu and van Gerven, 2015; Mcclure and Kriegeskorte, 2016). Moreover, DCNNs trained on object recognition tasks are currently among the best models we have of the mammalian visual system. This led us to hypothesize that teaching DCNNs to achieve even more brain-like representations could improve their performance. To test this, we trained DCNNs on a composite task, wherein networks were trained to: (a) classify images of objects; while (b) having intermediate representations that resemble those observed in neural recordings from monkey visual cortex. Compared with DCNNs trained purely for object categorization, DCNNs trained on the composite task had better object recognition performance and are more robust to label corruption. Interestingly, we found that neural data was not required for this process, but randomized data with the same statistical properties as neural data also boosted performance. While the performance gains we observed when training on the composite task vs the “pure” object recognition task were modest, they were remarkably robust. Notably, we observed these performance gains across all network variations we studied, including: smaller (CORNet-Z) vs larger (VGG-16) architectures; variations in optimizers (Adam vs gradient descent); variations in activation function (ReLU vs ELU); and variations in network initialization. Our results demonstrate the potential utility of a new approach to training object recognition networks, using strategies in which the brain – or at least the statistical properties of its activation patterns – serves as a teacher signal for training DCNNs.

Precise Targeting of Single Microelectrodes to Orientation Pinwheel Centers.

In the mammalian visual system, early stages of visual form perception begin with orientation selective neurons in primary visual cortex (V1). In many species (including humans, monkeys, tree shrews, cats, and ferrets), these neurons are organized in pinwheel-like orientation columns. To study the functional organization within orientation pinwheels, it is important to target pinwheel subdomains precisely. We therefore developed a technique to provide a quantitative determination of the location of pinwheel centers (PCs). Previous studies relied solely on blood vessel images of the cortical surface to guide electrode penetrations to PCs in orientation maps. However, considerable spatial error remained using this method. In the present study, we improved the accuracy of targeting PCs by ensuring perpendicularity of electrodes and by utilizing the orientation tuning of local field potentials (LFP) recorded at or near the optically determined positions.

A Differential Circuit via Retino-Colliculo-Pulvinar Pathway Enhances Feature Selectivity in Visual Cortex through Surround Suppression

In the mammalian visual system, information from the retina streams into parallel bottom-up pathways. It remains unclear how these pathways interact to contribute to contextual modulation of visual cortical processing. By optogenetic inactivation and activation of mouse lateral posterior nucleus (LP) of thalamus, a homolog of pulvinar, or its projection to primary visual cortex (V1), we found that LP contributes to surround suppression of layer (L) 2/3 responses in V1 by driving L1 inhibitory neurons. This results in subtractive suppression of visual responses and an overall enhancement of orientation, direction, spatial, and size selectivity. Neurons in V1-projecting LP regions receive bottom-up input from the superior colliculus (SC) and respond preferably to non-patterned visual noise. The noise-dependent LP activity allows V1 to "cancel" noise effects and maintain its orientation selectivity under varyingnoise background. Thus, the retina-SC-LP-V1 pathway forms a differential circuit with the canonical retino-geniculate pathway to achieve context-dependent sharpening of visual representations.

A bio-inspired quaternion local phase CNN layer with contrast invariance and linear sensitivity to rotation angles

Deep learning models have been particularly successful with image recognition using Convolutional Neural Networks (CNN). However, the learning of a contrast invariance and rotation equivariance response may fail even with very deep CNNs or by large data augmentations in training.We were inspired by the V1 visual features of the mammalian visual system to emulate as much as possible the early visual system and add more invariant capacities to the CNN. We present a new quaternion local phase convolutional neural network layer encoding three local phases. We present two experimental setups: An image classification task with four contrast levels, and a linear regression task that predicts the rotation angle of an image. In sum, we obtain new patterns and feature representations for deep learning, which capture illumination invariance and a linear response to rotation angles.

Biologically-Inspired Computational Neural Mechanism for Human Action/activity Recognition: A Review

Theoretical neuroscience investigation shows valuable information on the mechanism for recognizing the biological movements in the mammalian visual system. This involves many different fields of researches such as psychological, neurophysiology, neuro-psychological, computer vision, and artificial intelligence (AI). The research on these areas provided massive information and plausible computational models. Here, a review on this subject is presented. This paper describes different perspective to look at this task including action perception, computational and knowledge based modeling, psychological, and neuroscience approaches.

Towards a state-space geometry of neural responses to natural scenes: A steady-state approach

Our understanding of information processing by the mammalian visual system has come through a variety of techniques ranging from psychophysics and fMRI to single unit recording and EEG. Each technique provides unique insights into the processing framework of the early visual system. Here, we focus on the nature of the information that is carried by steady state visual evoked potentials (SSVEPs). To study the information provided by SSVEPs, we presented human participants with a population of natural scenes and measured the relative SSVEP response. Rather than focus on particular features of this signal, we focused on the full state-space of possible responses and investigated how the evoked responses are mapped onto this space. Our results show that it is possible to map the relatively high-dimensional signal carried by SSVEPs onto a 2-dimensional space with little loss. We also show that a simple biologically plausible model can account for a high proportion of the explainable variance (~73%) in that space. Finally, we describe a technique for measuring the mutual information that is available about images from SSVEPs. The techniques introduced here represent a new approach to understanding the nature of the information carried by SSVEPs. Crucially, this approach is general and can provide a means of comparing results across different neural recording methods. Altogether, our study sheds light on the encoding principles of early vision and provides a much needed reference point for understanding subsequent transformations of the early visual response space to deeper knowledge structures that link different visual environments.

Effects of dietary taurine level on visual function in European sea bass (Dicentrarchus labrax).

Dietary insufficiencies have been well documented to decrease growth rates and survival (and therefore overall production) in fish aquaculture. By contrast, the effects of dietary insufficiencies on the sensory biology of cultured fish remains largely unstudied. Diets based solely on plant protein sources could have advantages over fish-based diets because of the cost and ecological effects of the latter, but plant proteins lack the amino acid taurine. Adequate levels of taurine are, however, necessary for the development of a fully functional visual system in mammals. As part of ongoing studies to determine the suitability of plant-based diets, we investigated the effects of normal and reduced taurine dietary levels on retinal anatomy and function in European sea bass (Dicentrarchus labrax). We could not demonstrate any effects of dietary taurine level on retinal anatomy, nor the functional properties of luminous sensitivity and temporal resolution (measured as flicker fusion frequency). We did, however, find an effect on spectral sensitivity. The peak of spectral sensitivity of individuals fed a 5% taurine diet was rightward shifted (i.e., towards longer wavelengths) relative to that of fish fed a 0% or 1.5% taurine diet. This difference in in spectral sensitivity was due to a relatively lower level of middle wavelength pigment (maximum absorbance .500 nm) in fish fed a 5% taurine diet. Changes in spectral sensitivity resulting from diets containing different taurine levels are unlikely to be detrimental to fish destined for market, but could be in fishes that are being reared for stock enhancement programs.

Connexin-36 distribution and layer-specific topography in the cat retina

Connexin-36 (Cx36) is the major constituent of mammalian retinal gap junctions positioned in key signal pathways. Here, we examined the laminar and large-scale topographical distribution of Cx36 punctate immunolabels in the retina of the cat, a classical model of the mammalian visual system. Calretinin-immunoreactive (CaR-IR) cell populations served to outline the nuclear and plexiform layers and to stain specific neuronal populations. CaR-IR cells included horizontal cells in the outer retina, numerous amacrine cells, and scattered cells in the ganglion cell layer. Cx36-IR plaques were found among horizontal cell dendrites albeit without systematic colocalization of the two labels. Diffuse Cx36 immunoreactivity was found in the cytoplasm of AII amacrine cells, but no colocalization of Cx36 plaques was observed with either the perikarya or the long varicose dendrites of the CaR-IR non-AII amacrine cells. Cx36 puncta were seen throughout the entire inner plexiform layer showing their highest density in the ON sublamina. The densities of AII amacrine cell bodies and Cx36 plaques in the ON sublamina were strongly correlated across a wide range of eccentricities suggesting their anatomical association. However, the high number of plaques per AII cell suggests that a considerable fraction of Cx36 gap junctions in the ON sublamina is formed by other cell types than AII amacrine cells drawing attention to extensive but less studied electrically coupled networks.

Hierarchical visual attention model for saliency detection inspired by avian visual pathways

Visual attention is a mechanism that enables the visual system to detect potentially important objects in complex environment. Most computational visual attention models are designed with inspirations from mammalian visual systems. However, electrophysiological and behavioral evidences indicate that avian species are animals with high visual capability that can process complex information accurately in real time. Therefore, the visual system of the avian species, especially the nuclei related to the visual attention mechanism, are investigated in this paper. Afterwards, a hierarchical visual attention model is proposed for saliency detection. The optic tectum neuron responses are computed and the self-information is used to compute primary saliency maps in the first hierarchy. The &#x201C winner-take-all &#x201D network in the tecto-isthmal projection is simulated and final saliency maps are estimated with the regularized random walks ranking in the second hierarchy. Comparison results verify that the proposed model, which can define the focus of attention accurately, outperforms several state-of-the-art models. This study provides insights into the relationship between the visual attention mechanism and the avian visual pathways. The computational visual attention model may reveal the underlying neural mechanism of the nuclei for biological visual attention.

Connecting spectral radiometry of anthropogenic light sources to the visual ecology of organisms

AbstractHumans have drastically altered nocturnal environments with electric lighting. Animals depend on natural night light conditions and are now being inundated with artificial lighting. There are numerous artificial light sources that differ in spectral composition that should affect the perception of these light sources and due to differences in animal visual systems, the differences in color perception of these anthropogenic light sources should vary significantly. The ecological and evolutionary ramifications of these perceptual differences of light sources remain understudied. Here, we quantify the radiance of nine different street lights comprised of four different light sources: Metal Halide, Mercury Vapor, Light Emitting Diodes, and High‐Pressure Sodium and model how five animal visual systems will be stimulated by these light sources. We calculated the number of photons that photoreceptors in different visual systems would detect. We selected five visual systems: avian UV/VIS, avian V/VIS, human, wolf and hawk moth. We included non‐visual photoreceptors of vertebrates known for controlling circadian rhythms and other physiological traits. The nine light types stimulated visual systems and the photoreceptors within the visual systems differently. Furthermore, we modelled the chromatic contrast (Just Noticeable Differences [JNDs]) and color space overlap for each light type comparison for each visual system to see if organisms would perceive the lights as different colors. The JNDs of most light type comparisons were very high, indicating most visual systems would detect all light types as different colors, however mammalian visual systems would perceive many lights as the same color. We discuss the importance of understanding not only the brightness of artificial light types, but also the spectral composition of light types, as most organisms have different visual systems from humans. Thus, for researchers to understand how artificial light sources affect the visual environment of specific organisms and thus mitigate the effects, spectral information is crucial.

Rod Photoresponse Kinetics Limit Temporal Contrast Sensitivity in Mesopic Vision.

The mammalian visual system operates over an extended range of ambient light levels by switching between rod and cone photoreceptors. Rod-driven vision is sluggish, highly sensitive, and operates in dim or scotopic lights, whereas cone-driven vision is brisk, less sensitive, and operates in bright or photopic lights. At intermediate or mesopic lights, vision transitions seamlessly from rod-driven to cone-driven, despite the profound differences in rod and cone response dynamics. The neural mechanisms underlying such a smooth handoff are not understood. Using an operant behavior assay, electrophysiological recordings, and mathematical modeling we examined the neural underpinnings of the mesopic visual transition in mice of either sex. We found that rods, but not cones, drive visual sensitivity to temporal light variations over much of the mesopic range. Surprisingly, speeding up rod photoresponse recovery kinetics in transgenic mice improved visual sensitivity to slow temporal variations, in the range where perceptual sensitivity is governed by Weber's law of sensation. In contrast, physiological processes acting downstream from phototransduction limit sensitivity to high frequencies and temporal resolution. We traced the paradoxical control of visual temporal sensitivity to rod photoresponses themselves. A scenario emerges where perceptual sensitivity is limited by: (1) the kinetics of neural processes acting downstream from phototransduction in scotopic lights, (2) rod response kinetics in mesopic lights, and (3) cone response kinetics as light levels rise into the photopic range.SIGNIFICANCE STATEMENT Our ability to detect flickering lights is constrained by the dynamics of the slowest step in the visual pathway. Cone photoresponse kinetics limit visual temporal sensitivity in bright (photopic) lights, whereas mechanisms in the inner retina limit sensitivity in dim (scotopic) lights. The neural mechanisms underlying the transition between scotopic and photopic vision in mesopic lights, when both rods are cones are active, are unknown. This study provides a missing link in this mechanism by establishing that rod photoresponse kinetics limit temporal sensitivity during the mesopic transition. Surprisingly, this range is where Weber's Law of Sensation governs temporal contrast sensitivity in mouse. Our results will help guide future studies of complex and dynamic interactions between rod-cone signals in the mesopic retina.

Extrastriate connectivity of the mouse dorsal lateral geniculate thalamic nucleus.

The mammalian visual system is one of the most well-studied brain systems. Visual information from the retina is relayed to the dorsal lateral geniculate nucleus of the thalamus (LGd). The LGd then projects topographically to primary visual cortex (VISp) to mediate visual perception. In this view, the VISp is a critical network hub where visual information must traverse LGd-VISp circuits to reach higher order "extrastriate" visual cortices, which surround the VISp on its medial and lateral borders. However, decades of conflicting reports in a variety of mammals support or refute the existence of extrastriate LGd connections that can bypass the VISp. Here, we provide evidence of bidirectional extrastriate connectivity with the mouse LGd. Using small, discrete coinjections of anterograde and retrograde tracers within the thalamus and cortex, our cross-validated approach identified bidirectional connectivity between LGd and extrastriate visual cortices. We find robust reciprocal connectivity of the medial extrastriate regions with LGd neurons distributed along the "ventral strip" border with the intergeniculate leaflet. In contrast, LGd input to lateral extrastriate regions is sparse, but lateral extrastriate regions return stronger descending projections to localized LGd areas. We show further evidence that axons from lateral extrastriate regions can overlap onto medial extrastriate-projecting LGd neurons in the ventral strip, providing a putative subcortical LGd pathway for communication between medial and lateral extrastriate regions. Overall, our findings support the existence of extrastriate LGd circuits and provide novel understanding of LGd organization in rodent visual system.