Extra-classical Receptive Field Research Articles

Recurrent neural circuits for perceptual grouping

Neurons in the visual cortex are sensitive to context: Responses to stimuli presented within their classical receptive fields (CRFs) are modulated by stimuli in their surrounding extra-classical receptive fields (eCRFs). However, the circuits underlying these contextual effects are not well understood, and little is known about how these circuits drive perception during everyday vision. We tackle these questions by approximating circuit-level eCRF models with a differentiable discrete-time recurrent neural network that is trainable with gradient-descent. After optimizing model synaptic connectivity and dynamics for object contour detection in natural images, the neural-circuit model rivals human observers on the task with far better sample efficiency than state-of-the-art computer vision approaches. Notably, the model also exhibits CRF and eCRF phenomena typically associated with primate vision. The model’s ability to accurately detect object contours also critically depends on these effects, and these contextual effects are not found in ablated versions of the model. Finally, we derive testable predictions about the neural mechanisms responsible for contextual integration and illustrate their importance for accurate and efficient perceptual grouping.

Contextual Modulation of Feedforward Inputs to Primary Visual Cortex.

Throughout the brain, parallel processing streams compose the building blocks of complex neural functions. One of the most salient models for studying the functional specialization of parallel visual streams in the primate brain is the lateral geniculate nucleus (LGN) of the dorsal thalamus, through which the parvocellular and magnocellular channels, On-center and Off-center channels, and ipsilateral and contralateral eye channels are maintained and provide the foundation for cortical processing. We examined three aspects of neural processing in these streams: (1) the relationship between extraclassical surround suppression, a widespread visual computation thought to represent a canonical neural computation, and the parallel channels of the LGN; (2) the magnitude of binocular interaction in the parallel streams; and (3) the magnitude of suppression elicited by perceptual competition (binocular rivalry) in each stream. Our results show that surround suppression is almost exclusive to Off channel cells; further, we found evidence for two different components of monocular surround suppression—an early-stage suppression exhibited by all magnocellular cells, and a late-stage suppression exhibited only by Off cells in both the parvocellular and magnocellular pathways. This finding indicates that stream-specific circuits contribute to surround suppression in the primate LGN and suggests a distinct role for suppression in the Off channel to the cortex. We also examined the responses of LGN neurons in alert macaque monkeys to determine whether neurons that supply the cortex with visual information are influenced by stimulation of both eyes. Our results demonstrate that LGN neurons are not influenced by stimulation of the non-dominant eye. This was the case when dichoptic stimuli were presented to classical receptive fields of neurons, extraclassical receptive fields of neurons, and when stimuli were appropriate to produce the perception of binocular rivalry.

Adaptive Surround Modulation of MT Neurons: A Computational Model.

The classical receptive field (CRF) of a spiking visual neuron is defined as the region in the visual field that can generate spikes when stimulated by a visual stimulus. Many visual neurons also have an extra-classical receptive field (ECRF) that surrounds the CRF. The presence of a stimulus in the ECRF does not generate spikes but rather modulates the response to a stimulus in the neuron's CRF. Neurons in the primate Middle Temporal (MT) area, which is a motion specialist region, can have directionally antagonistic or facilitatory surrounds. The surround's effect switches between directionally antagonistic or facilitatory based on the characteristics of the stimulus, with antagonistic effects when there are directional discontinuities but facilitatory effects when there is directional coherence. Here, we present a computational model of neurons in area MT that replicates this observation and uses computational building blocks that correlate with observed cell types in the visual pathways to explain the mechanism of this modulatory effect. The model shows that the categorization of MT neurons based on the effect of their surround depends on the input stimulus rather than being a property of the neurons. Also, in agreement with neurophysiological findings, the ECRFs of the modeled MT neurons alter their center-surround interactions depending on image contrast.

Distinct spatiotemporal mechanisms underlie extra-classical receptive field modulation in macaque V1 microcircuits.

Complex scene perception depends upon the interaction between signals from the classical receptive field (CRF) and the extra-classical receptive field (eCRF) in primary visual cortex (V1) neurons. Although much is known about V1 eCRF properties, we do not yet know how the underlying mechanisms map onto the cortical microcircuit. We probed the spatio-temporal dynamics of eCRF modulation using a reverse correlation paradigm, and found three principal eCRF mechanisms: tuned-facilitation, untuned-suppression, and tuned-suppression. Each mechanism had a distinct timing and spatial profile. Laminar analysis showed that the timing, orientation-tuning, and strength of eCRF mechanisms had distinct signatures within magnocellular and parvocellular processing streams in the V1 microcircuit. The existence of multiple eCRF mechanisms provides new insights into how V1 responds to spatial context. Modeling revealed that the differences in timing and scale of these mechanisms predicted distinct patterns of net modulation, reconciling many previous disparate physiological and psychophysical findings.

Pattern Motion Processing by MT Neurons

Based on stimulation with plaid patterns, neurons in the Middle Temporal (MT) area of primate visual cortex are divided into two types: pattern and component cells. The prevailing theory suggests that pattern selectivity results from the summation of the outputs of component cells as part of a hierarchical visual pathway. We present a computational model of the visual pathway from primary visual cortex (V1) to MT that suggests an alternate model where the progression from component to pattern selectivity is not required. Using standard orientation-selective V1 cells, end-stopped V1 cells, and V1 cells with extra-classical receptive fields (RFs) as inputs to MT, the model shows that the degree of pattern or component selectivity in MT could arise from the relative strengths of the three V1 input types. Dominance of end-stopped V1 neurons in the model leads to pattern selectivity in MT, while dominance of V1 cells with extra-classical RFs result in component selectivity. This model may assist in designing experiments to further understand motion processing mechanisms in primate MT.

Development of Human Infants’ Receptive Field Mechanisms in Visual Motion Processing

Motion direction of a large high-contrast pattern is more difficult to perceive than that of a small one [1]. This counterintuitive perceptual phenomenon is considered to reflect surround suppression, a receptive field property observed in the visual cortex [2-5]. Here, we demonstrate that this phenomenon can be observed in human infants. Infants at 7-8 months of age showed higher sensitivity for a small motion stimulus than for a large one. However, infants under 6 months showed the opposite result; motion sensitivity was higher for a large stimulus. These results suggest that suppressive surround regions beyond classical receptive fields develop in the second half of the first year. Moreover, we examined the size of receptive fields in infants using this phenomenon and found that the center region of receptive fields shrinks from 3 months to 8 months. Our findings suggest that receptive fields related to motion processing are broad and do not have extra-classical receptive fields in early infancy, and that they become narrower and acquire suppressive surround regions in the first year of life.

Complementary surrounds explain diverse contextual phenomena across visual modalities.

Context is known to affect how a stimulus is perceived. A variety of illusions have been attributed to contextual processing-from orientation tilt effects to chromatic induction phenomena, but their neural underpinnings remain poorly understood. Here, we present a recurrent network model of classical and extraclassical receptive fields that is constrained by the anatomy and physiology of the visual cortex. A key feature of the model is the postulated existence of near- versus far- extraclassical regions with complementary facilitatory and suppressive contributions to the classical receptive field. The model accounts for a variety of contextual illusions, reveals commonalities between seemingly disparate phenomena, and helps organize them into a novel taxonomy. It explains how center-surround interactions may shift from attraction to repulsion in tilt effects, and from contrast to assimilation in induction phenomena. The model further explains enhanced perceptual shifts generated by a class of patterned background stimuli that activate the two opponent extraclassical regions cooperatively. Overall, the ability of the model to account for the variety and complexity of contextual illusions provides computational evidence for a novel canonical circuit that is shared across visual modalities. (PsycINFO Database Record (c) 2018 APA, all rights reserved).

Identification of a Retinal Circuit for Recurrent Suppression Using Indirect Electrical Imaging

Understanding the function of modulatory interneuron networks is a major challenge, because such networks typically operate over long spatial scales and involve many neurons of different types. Here, we use an indirect electrical imaging method to reveal the function of a spatially extended, recurrent retinal circuit composed of two cell types. This recurrent circuit produces peripheral response suppression of early visual signals in the primate magnocellular visual pathway. We identify a type of polyaxonal amacrine cell physiologically via its distinctive electrical signature, revealed by electrical coupling with ON parasol retinal ganglion cells recorded using a large-scale multi-electrode array. Coupling causes the amacrine cells to fire spikes that propagate radially over long distances, producing GABA-ergic inhibition of other ON parasol cells recorded near the amacrine cell axonal projections. We propose and test a model for the function of this amacrine cell type, in which the extra-classical receptive field of ON parasol cells is formed by reciprocal inhibition from other ON parasol cells in the periphery, via the electrically coupled amacrine cell network.

Dynamic functional connectivity among neuronal population during modulation of extra-classical receptive field in primary visual cortex.

The neuronal activity evoked by stimuli confined in a receptive field can be modulated by surround stimuli of the extra-classical receptive field (eCRF). The surrounding modulation, hypothesized to be the basis of visual feature integration and figure-ground segregation, has drawn much attention in the field of neuroscience and engineering. However, most studies focused on surround modulation of individual neuronal response. In this study, we analyzed surround modulation of the population response recorded from rat primary visual cortex, and further investigated dynamic functional connectivity modulated by the surrounding stimuli. The functional connectivity was estimated using Granger causality (GC) and then determined by thresholding the p-matrix with different significance α values. Four scalar indexes were calculated to describe the functional connectivity of neuronal population: averaged connection strength (mGC), connection density (D), clustering coefficient (C) and path length (L). The statistical results from 5 rats showed that these network characteristics were dynamically changed during modulation of surrounding stimuli, which suggested that the neuronal population may connect in a dynamic way during modulation of eCRF. We further guessed that the neurons may happened to be organized in a more efficient way underlying surrounding modulation conditions, which helps to process larger images efficiently with the same number of neurons. This study provided new insights for a better understanding of the underlying neural mechanisms responsible for surround modulation.

Computing with a canonical neural circuits model with pool normalization and modulating feedback.

Evidence suggests that the brain uses an operational set of canonical computations like normalization, input filtering, and response gain enhancement via reentrant feedback. Here, we propose a three-stage columnar architecture of cascaded model neurons to describe a core circuit combining signal pathways of feedforward and feedback processing and the inhibitory pooling of neurons to normalize the activity. We present an analytical investigation of such a circuit by first reducing its detail through the lumping of initial feedforward response filtering and reentrant modulating signal amplification. The resulting excitatory-inhibitory pair of neurons is analyzed in a 2D phase-space. The inhibitory pool activation is treated as a separate mechanism exhibiting different effects. We analyze subtractive as well as divisive (shunting) interaction to implement center-surround mechanisms that include normalization effects in the characteristics of real neurons. Different variants of a core model architecture are derived and analyzed--in particular, individual excitatory neurons (without pool inhibition), the interaction with an inhibitory subtractive or divisive (i.e., shunting) pool, and the dynamics of recurrent self-excitation combined with divisive inhibition. The stability and existence properties of these model instances are characterized, which serve as guidelines to adjust these properties through proper model parameterization. The significance of the derived results is demonstrated by theoretical predictions of response behaviors in the case of multiple interacting hypercolumns in a single and in multiple feature dimensions. In numerical simulations, we confirm these predictions and provide some explanations for different neural computational properties. Among those, we consider orientation contrast-dependent response behavior, different forms of attentional modulation, contrast element grouping, and the dynamic adaptation of the silent surround in extraclassical receptive field configurations, using only slight variations of the same core reference model.

Binocular neurons in parastriate cortex: interocular 'matching' of receptive field properties, eye dominance and strength of silent suppression.

Spike-responses of single binocular neurons were recorded from a distinct part of primary visual cortex, the parastriate cortex (cytoarchitectonic area 18) of anaesthetized and immobilized domestic cats. Functional identification of neurons was based on the ratios of phase-variant (F1) component to the mean firing rate (F0) of their spike-responses to optimized (orientation, direction, spatial and temporal frequencies and size) sine-wave-luminance-modulated drifting grating patches presented separately via each eye. In over 95% of neurons, the interocular differences in the phase-sensitivities (differences in F1/F0 spike-response ratios) were small (≤0.3) and in over 80% of neurons, the interocular differences in preferred orientations were ≤10°. The interocular correlations of the direction selectivity indices and optimal spatial frequencies, like those of the phase sensitivies and optimal orientations, were also strong (coefficients of correlation r ≥0.7005). By contrast, the interocular correlations of the optimal temporal frequencies, the diameters of summation areas of the excitatory responses and suppression indices were weak (coefficients of correlation r ≤0.4585). In cells with high eye dominance indices (HEDI cells), the mean magnitudes of suppressions evoked by stimulation of silent, extra-classical receptive fields via the non-dominant eyes, were significantly greater than those when the stimuli were presented via the dominant eyes. We argue that the well documented ‘eye-origin specific’ segregation of the lateral geniculate inputs underpinning distinct eye dominance columns in primary visual cortices of mammals with frontally positioned eyes (distinct eye dominance columns), combined with significant interocular differences in the strength of silent suppressive fields, putatively contribute to binocular stereoscopic vision.

Spatiotemporal characteristics of surround suppression in primary visual cortex and lateral geniculate nucleus of the cat.

In the primary visual cortex (V1), a neuronal response to stimulation of the classical receptive field (CRF) is predominantly suppressed by a stimulus presented outside the CRF (extraclassical receptive field, ECRF), a phenomenon referred to as ECRF suppression. To elucidate the neuronal mechanisms and origin of ECRF suppression in V1 of anesthetized cats, we examined the temporal properties of the spatial extent and orientation specificity of ECRF suppression in V1 and the lateral geniculate nucleus (LGN), using stationary-flashed sinusoidal grating. In V1, we found three components of ECRF suppression: (1) local and fast, (2) global and fast, and (3) global and late. The local and fast component, which resulted from within 2° of the boundary of the CRF, started no more than 10 ms after the onset of the CRF response and exhibited low specificity for the orientation of the ECRF stimulus. These spatiotemporal properties corresponded to those of geniculate ECRF suppression, suggesting that the local and fast component of V1 is inherited from the LGN. In contrast, the two global components showed rather large spatial extents ∼5° from the CRF boundary and high specificity for orientation, suggesting that their possible origin is the cortex, not the LGN. Correspondingly, the local component was observed in all neurons of the thalamocortical recipient layer, while the global component was biased toward other layers. Therefore, we conclude that both subcortical and cortical mechanisms with different spatiotemporal properties are involved in ECRF suppression.

Motion based prediction and development of response to an "on the way" stimulus

Prior information on regularities of the world is essential for optimal performance of sensory processing. In the particular case of the detection of visual motion, the prior knowledge on the temporal coherency of motion facilitates estimation of predictable trajectories. Based on such prior information, we have proposed a generic bayesian modeling framework to implement anisotropic diffusion of estimated motion information. By applying particle filtering (Sequential Monte Carlo) we were able to study the emergence of a global neural response through a probabilistic pool of local sensory information. To explore if such diffusive mechanisms could explain neurophysiological recordings from the responses of neural populations to a predictable moving stimulus that we have recorded from V1 neurons in macaque monkeys, we simulated such response in the model. Our main experimental protocol consisted of a bar moving to a fixed direction but with different trajectory lengths before reaching the receptive fields of the recorded neurons. Experimental results show that a bar moving towards a long trajectory generates an anticipatory response that build-up gradually with time before the bar actually enters into the receptive field of the cell. In the modeling part we aim to asses the role of prediction in the dynamical development of neural activity at, or before, the arrival of the sensory stimulus. We simulated the response of the model in target receptive fields to a smoothly moving stimulus whose trajectory crosses the receptive field. This was done under various conditions for two classic stimuli: a dot and a slanted bar. We found that, as expected, a smoothly moving stimulus triggers a transient response as it crosses the receptive field and that he peak time of this response matches the arrival of the stimulus to the it’s center. We found that the development of a response to an ‘on the way stimulus’ is enhanced by increasing the reliability of sensory information along the trajectory of motion. Also we observed an advance of the peak time with respect to the degree of predictability of the trajectory. This is in agreement with the neural activity that we have recorded from V1 neurons in response to a smoothly moving slanted bar with different trajectory lengths. This results may be explained with the definition of classical and extra classical receptive fields while former is activated only by having a stimulus inside it and later covers bigger area and collects information from a wider neighborhood [1]. Motion based prediction by highlighting the anisotropic diffusion of estimated motion information can develop another perspective on lateral spread of activity in primary visual cortex and better understanding of modulatory effect of cells on each other.

Spontaneous emergence of simple and complex receptive fields in a spiking model of V1

Brain Corporation is engaged in a multi-year project to build a spiking model of vision, paying special attention to the anatomy and physiology of the mammalian visual system. While it is relatively easy to hand-tune V1 to get simple and complex cells, it is not clear how to arrange connectivity in other cortical areas to get appropriate receptive fields, or what the appropriate receptive fields even should be. Instead of pre-wiring cortical connectivity according to a computational theory of how vision should work, we start with a generic tabula rasa spiking network having multiple cortical layers and neuronal types (single-compartment RS, FS, LTS cells). The goal is to find the anatomical and physiological parameters so that the appropriate connectivity emerges through STDP and visual experience. Since we know exactly what kind of receptive fields and visual responses are in V1, we build a smaller model of retina-LGN-V1 pathway and tune the STDP parameters so that the expected responses emerge. Admittedly, there are many free parameters that could be tuned to get V1-like responses; our choice is motivated by in-vitro recordings and other published data, whenever possible. Once we trust what we see in V1, we are ready to copy and paste the cortical model to implement V2, V3, V4, and IT areas with the hope that useful connectivity, receptive fields, and visual responses emerge. Our large-scale simulations of the spiking model of the visual system show spontaneous emergence of simple and complex cells, orientation domains, end-stopping receptive fields, extra-classical receptive fields with tuned surround suppression, color opponency that depends on the eccentricity of the receptive field, contrast invariance, and many other features that are routinely recorded in V1. Since the visual model exhibits micro- and full saccades, we observe perceptual behavior, such as the emergence of bottom-up (pop-out) attention. The model underscores the importance of spike-timing dynamics, inhibition, saccadic mechanism, and it imposes important restrictions on the possible types of STDP to model early visual processing. This presentation is meant to be a flagship introduction to the whole effort by Brain Corporation to build spiking model of vision. There are multiple other submissions elaborating details of our modeling effort. The figure summarizes the gross anatomy of the spiking model of the visual system, including area MT, Superior Colliculus (SC), Frontal Pursuit Area (FPA), Flocculus (Floc), and Brainstem (B.Stem). The retina covers 10 by 10 degrees of visual field with the density of L, M, and S cones and receptive field sizes of Midget, Parasol, and SBC retino-ganglion cells corresponding to 4 degrees of eccentricity of a primate retina (collaboration with EJ Chichilnisky lab at Salk Institute). The model consists of more than 1M single-compartment neurons of RS, FS, and LTS types (Izhikevich, 2007 Dynamical Systems in Neuroscience) and 0.25B synapses having AMPA, NMDA, GABA_A and GABA_B conductances, axonal conduction delays, and STDP. In this presentation, we summarize these and other anatomical and physiological assumptions, describe major results and point to other submissions by Brain Corporation scientists for more details. Figure 1

Spatial sensitivity, responsivity, and surround suppression of LGN cell responses in the macaque

Responses of cells in the lateral geniculate nucleus (LGN) of the macaque monkey have been measured for different sizes of chromatic and achromatic stimuli, with relative luminance spanning a range of 3-6 log units. Homogeneous illuminated test fields, centered on the receptive field, were used. Responses to these stimuli deviated from what is expected for the grating stimuli used to study the contrast-sensitive mechanisms in the visual pathway. For test fields smaller than the center of the receptive field, both the excitatory and the inhibitory cone-opponent components were present in the response, and the sensitivity to both components increased with the same factor when the test field increased in size (area summation). For test field areas extending into the suppressive surround of the extraclassical receptive field, the excitatory and the inhibitory cone opponents were both suppressed, again by the same factor. This suppression of the cell's responsiveness, as a function of test spot area, was described by a logarithmic function, and the spatial sensitivity of attenuation could therefore be described by a power function of radius. The logarithmic suppression was clear for parvocellular and koniocellular cells but was more prominent for magnocellular cells. The surround field suppression was also found for the prepotential inputs to LGN cells, indicating a retinal origin. The difference of Gaussian (DOG) model has been used successfully to describe the cells' contrast behavior for grating stimuli. However, this model fails to describe the constant excitatory/inhibitory response balance needed to obtain color (hue) stability for light stimuli of different sizes but with the same Commission Internationale de l'Eclairage (CIE) chromaticity and luminance factor. Neither the constant responsiveness found in the center of the receptive field nor the suppressive response in the surround can be described by the DOG model.

Functional Characterization of the Extraclassical Receptive Field in Macaque V1: Contrast, Orientation, and Temporal Dynamics

Neurons in primary visual cortex, V1, very often have extraclassical receptive fields (eCRFs). The eCRF is defined as the region of visual space where stimuli cannot elicit a spiking response but can modulate the response of a stimulus in the classical receptive field (CRF). We investigated the dependence of the eCRF on stimulus contrast and orientation in macaque V1 cells for which the laminar location was determined. The eCRF was more sensitive to contrast than the CRF across the whole population of V1 cells with the greatest contrast differential in layer 2/3. We confirmed that many V1 cells experience stronger suppression for collinear than orthogonal stimuli in the eCRF. Laminar analysis revealed that the predominant bias for collinear suppression was found in layers 2/3 and 4b. The laminar pattern of contrast and orientation dependence suggests that eCRF suppression may derive from different neural circuits in different layers, and may be comprised of two distinct components: orientation-tuned and untuned suppression. On average tuned suppression was delayed by ∼25 ms compared with the onset of untuned suppression. Therefore, response modulation by the eCRF develops dynamically and rapidly in time.

Stability of simple/complex classification with contrast and extraclassical receptive field modulation in macaque V1

A key property of neurons in primary visual cortex (V1) is the distinction between simple and complex cells. Recent reports in cat visual cortex indicate the categorization of simple and complex can change depending on stimulus conditions. We investigated the stability of the simple/complex classification with changes in drive produced by either contrast or modulation by the extraclassical receptive field (eCRF). These two conditions were reported to increase the proportion of simple cells in cat cortex. The ratio of the modulation depth of the response (F1) to the elevation of response (F0) to a drifting grating (F1/F0 ratio) was used as the measure of simple/complex. The majority of V1 complex cells remained classified as complex with decreasing contrast. Near contrast threshold, an equal proportion of simple and complex cells changed their classification. The F1/F0 ratio was stable between optimal and large stimulus areas even for those neurons that showed strong eCRF suppression. There was no discernible overall effect of surrounding spatial context on the F1/F0 ratio. Simple/complex cell classification is relatively stable across a range of stimulus drives, produced by either contrast or eCRF suppression.

Disinhibition Among the Extra-Classical Receptive Field of Retinal Ganglion Cells Contributes to Color Constancy

Disinhibition Among the Extra-Classical Receptive Field of Retinal Ganglion Cells Contributes to Color Constancy

Imaging extraclassical receptive fields in early visual cortex

Imaging extraclassical receptive fields in early visual cortex

Long-Range, Pattern-Dependent Contextual Effects in Early Human Visual Cortex

The standard view of neurons in early visual cortex is that they behave like localized feature detectors. Here we demonstrate that processing in early visual areas goes beyond feature detection by showing that neural responses are greater when a feature deviates from its context compared to when it does not deviate from its context. Using psychophysics, fMRI, and electroencephalography methodologies, we measured neural responses to an oriented Gabor ("target") embedded in various visual patterns as defined by the relative orientation of flanking stimuli. We first show using psychophysical contrast adaptation and fMRI that a target that differs from its context results in more neural activity compared to a target that is contained within an alternating sequence, suggesting that neurons in early visual cortex are sensitive to large-scale orientation patterns. Next, we use event-related potentials to show that orientation deviations affect the earliest sensory components of the target response. Finally, we use forced-choice classification of "noise" stimuli to show that we are more likely to "see" orientations that deviate from the context. Our results suggest that early visual cortex is sensitive to global patterns in images in a way that is markedly different from the predictions of standard models of cortical visual processing.