Receptive Field Properties Research Articles

Feature-specific divisive normalization improves natural image encoding for depth perception.

Vision science and visual neuroscience seek to understand how stimulus and sensor properties limit the precision with which behaviorally-relevant latent variables are encoded and decoded. In the primate visual system, binocular disparity-the canonical cue for stereo-depth perception-is initially encoded by a set of binocular receptive fields with a range of spatial frequency preferences. Here, with a stereo-image database having ground-truth disparity information at each pixel, we examine how response normalization and receptive field properties determine the fidelity with which binocular disparity is encoded in natural scenes. We quantify encoding fidelity by computing the Fisher information carried by the normalized receptive field responses. Several findings emerge from an analysis of the response statistics. First, broadband (or feature-unspecific) normalization yields Laplace-distributed receptive field responses, and narrowband (or feature-specific) normalization yields Gaussian-distributed receptive field responses. Second, the Fisher information in narrowband-normalized responses is larger than in broadband-normalized responses by a scale factor that grows with population size. Third, the most useful spatial frequency decreases with stimulus size and the range of spatial frequencies that is useful for encoding a given disparity decreases with disparity magnitude, consistent with neurophysiological findings. Fourth, the predicted patterns of psychophysical performance, and absolute detection threshold, match human performance with natural and artificial stimuli. The current computational efforts establish a new functional role for response normalization, and bring us closer to understanding the principles that should govern the design of neural systems that support perception in natural scenes.

The influence of attentional load on population receptive field properties

The influence of attentional load on population receptive field properties

Training-induced changes in population receptive field properties in visual cortex: Impact of eccentric vision training on population receptive field properties and the crowding effect.

This study aimed to investigate the impact of eccentric-vision training on population receptive field (pRF) estimates to provide insights into brain plasticity processes driven by practice. Fifteen participants underwent functional magnetic resonance imaging (fMRI) measurements before and after behavioral training on a visual crowding task, where the relative orientation of the opening (gap position: up/down, left/right) in a Landolt C optotype had to be discriminated in the presence of flanking ring stimuli. Drifting checkerboard bar stimuli were used for pRF size estimation in multiple regions of interest (ROIs): dorsal-V1 (dV1), dorsal-V2 (dV2), ventral-V1 (vV1), and ventral-V2 (vV2), including the visual cortex region corresponding to the trained retinal location. pRF estimates in V1 and V2 were obtained along eccentricities from 0.5° to 9°. Statistical analyses revealed a significant decrease of the crowding anisotropy index (p = 0.009) after training, indicating improvement on crowding task performance following training. Notably, pRF sizes at and near the trained location decreased significantly (p = 0.005). Dorsal and ventral V2 exhibited significant pRF size reductions, especially at eccentricities where the training stimuli were presented (p < 0.001). In contrast, no significant changes in pRF estimates were found in either vV1 (p = 0.181) or dV1 (p = 0.055) voxels. These findings suggest that practice on a crowding task can lead to a reduction of pRF sizes in trained visual cortex, particularly in V2, highlighting the plasticity and adaptability of the adult visual system induced by prolonged training.

Ultra Fast Deep Lane Detection With Hybrid Anchor Driven Ordinal Classification.

Modern methods mainly regard lane detection as a problem of pixel-wise segmentation, which is struggling to address the problems of efficiency and challenging scenarios like severe occlusions and extreme lighting conditions. Inspired by human perception, the recognition of lanes under severe occlusions and extreme lighting conditions is mainly based on contextual and global information. Motivated by this observation, we propose a novel, simple, yet effective formulation aiming at ultra fast speed and the problem of challenging scenarios. Specifically, we treat the process of lane detection as an anchor-driven ordinal classification problem using global features. First, we represent lanes with sparse coordinates on a series of hybrid (row and column) anchors. With the help of the anchor-driven representation, we then reformulate the lane detection task as an ordinal classification problem to get the coordinates of lanes. Our method could significantly reduce the computational cost with the anchor-driven representation. Using the large receptive field property of the ordinal classification formulation, we could also handle challenging scenarios. Extensive experiments on four lane detection datasets show that our method could achieve state-of-the-art performance in terms of both speed and accuracy. A lightweight version could even achieve 300+ frames per second(FPS). Our code is at https://github.com/cfzd/Ultra-Fast-Lane-Detection-v2.

Visual Corticotectal Neurons in Awake Rabbits: Receptive Fields and Driving Monosynaptic Thalamocortical Inputs.

The superior colliculus receives powerful synaptic inputs from corticotectal neurons in the visual cortex. The function of these corticotectal neurons remains largely unknown due to a limited understanding of their response properties and connectivity. Here, we use antidromic methods to identify corticotectal neurons in awake male and female rabbits, and measure their axonal conduction times, thalamic inputs and receptive field properties. All corticotectal neurons responded to sinusoidal drifting gratings with a nonlinear (nonsinusoidal) increase in mean firing rate but showed pronounced differences in their ON-OFF receptive field structures that we classified into three groups, Cx, S2, and S1. Cx receptive fields had highly overlapping ON and OFF subfields as classical complex cells, S2 had largely separated ON and OFF subfields as classical simple cells, and S1 had a single ON or OFF subfield. Thus, all corticotectal neurons are homogeneous in their nonlinear spatial summation but very heterogeneous in their spatial integration of ON and OFF inputs. The Cx type had the fastest conducting axons, the highest spontaneous activity, and the strongest and fastest visual responses. The S2 type had the highest orientation selectivity, and the S1 type had the slowest conducting axons. Moreover, our cross-correlation analyses found that a subpopulation of corticotectal neurons with very fast conducting axons and high spontaneous firing rates (largely "Cx" type) receives monosynaptic input from retinotopically aligned thalamic neurons. This previously unrecognized fast-conducting thalamic-mediated corticotectal pathway may provide specialized information to superior colliculus and prime recipient neurons for subsequent corticotectal or retinal synaptic input.

Tryptamine psychedelics shape population receptive fields in early visual cortex in the presence of perceptual distortions

Aims/Purpose: N, N‐dimethyltryptamine (DMT) is a psychedelic tryptamine associated with conspicuous visual hallucinatory phenomena and changes in perceived visual space. The neural basis of these effects remain unidentified. We asked whether changes in population receptive field properties in the primary visual cortex (V1) occur in parallel with changes in visual phenomenology.Methods: We tested this hypothesis using magnetic resonance imaging using a placebo‐controlled design of inhaled DMT. We used a population receptive field (pRF) mapping method, which assesses neural population visual response properties and retinotopic maps in visual cortex. One‐Way Repeated Measures ANOVA was performed to determine if populations of visual neurons do reorganize along eccentricity in response to inhaled DMT. We also tested the phenomenology of visual hallucinations using the Hallucinogen Rating scale (HRS) with a focus on the visual cluster of symptoms.Results: We show that the pRF sizes in V1 are significantly increased in peripherical field in the active condition (inhaled DMT) versus placebo (F (3, 54) = 22.6790, p‐value &lt;0.0001). Post hoc Paired Sample T‐Tests were used to determine which pairs of bins of eccentricity were associated to significant differences. We found statistically significant differences for the peripherical bins of eccentricity (3.75 deg, 6.25 deg, 8.75 deg). These effects were accompanied by higher visual scores in Hallucinogen Rating scale (HRS) F(1,9) = 46 434, p &lt; 0.05).Conclusions: This evidence for fast effects of DMT resulting in increased population receptive fields is consistent with the presence of visual manifestations and may explain perceptual distortions induced by psychedelics such as field blurring and the enlargement of nearby visual space, particularly at the visual periphery.

Heterogeneous presynaptic receptive fields contribute to directional tuning in starburst amacrine cells.

The processing of visual information by retinal starburst amacrine cells (SACs) involves transforming excitatory input from bipolar cells (BCs) into directional calcium output. While previous studies have suggested that an asymmetry in the kinetic properties of BCs along the soma-dendritic axes of the postsynaptic cell could enhance directional tuning at the level of individual branches, it remains unclear whether biologically relevant presynaptic kinetics contribute to direction selectivity (DS) when visual stimulation engages the entire dendritic tree. To address this question, we built multicompartmental models of the bipolar-SAC circuit and trained them to boost directional tuning. We report that despite significant dendritic crosstalk and dissimilar directional preferences along the dendrites that occur during whole-cell stimulation, the rules that guide BC kinetics leading to optimal DS are similar to the single-dendrite condition. To correlate model predictions to empirical findings, we utilized two-photon glutamate imaging to study the dynamics of bipolar release onto ON- and OFF-starburst dendrites in the murine retina. We reveal diverse presynaptic dynamics in response to motion in both BC populations; algorithms trained on the experimental data suggested that the differences in the temporal release kinetics are likely to correspond to heterogeneous receptive field properties among the different BC types, including the spatial extent of the center and surround components. In addition, we demonstrate that circuit architecture composed of presynaptic units with experimentally recorded dynamics could enhance directional drive but not to levels that replicate empirical findings, suggesting other DS mechanisms are required to explain SAC function. Our study provides new insights into the complex mechanisms underlying DS in retinal processing and highlights the potential contribution of presynaptic kinetics to the computation of visual information by SACs.

Heterogeneous presynaptic receptive fields contribute to directional tuning in starburst amacrine cells

The processing of visual information by retinal starburst amacrine cells (SACs) involves transforming excitatory input from bipolar cells (BCs) into directional calcium output. While previous studies have suggested that an asymmetry in the kinetic properties of BCs along the soma-dendritic axes of the postsynaptic cell could enhance directional tuning at the level of individual branches, it remains unclear whether biologically relevant presynaptic kinetics contribute to direction selectivity (DS) when visual stimulation engages the entire dendritic tree. To address this question, we built multicompartmental models of the bipolar–SAC circuit and trained them to boost directional tuning. We report that despite significant dendritic crosstalk and dissimilar directional preferences along the dendrites that occur during whole-cell stimulation, the rules that guide BC kinetics leading to optimal DS are similar to the single-dendrite condition. To correlate model predictions to empirical findings, we utilized two-photon glutamate imaging to study the dynamics of bipolar release onto ON- and OFF-starburst dendrites in the murine retina. We reveal diverse presynaptic dynamics in response to motion in both BC populations; algorithms trained on the experimental data suggested that the differences in the temporal release kinetics are likely to correspond to heterogeneous receptive field properties among the different BC types, including the spatial extent of the center and surround components. In addition, we demonstrate that circuit architecture composed of presynaptic units with experimentally recorded dynamics could enhance directional drive but not to levels that replicate empirical findings, suggesting other DS mechanisms are required to explain SAC function. Our study provides new insights into the complex mechanisms underlying DS in retinal processing and highlights the potential contribution of presynaptic kinetics to the computation of visual information by SACs.

Poster Session I: Gain, not changes in spatial receptive field properties, improves task performance in a neural network attention model.

Attention allows us to focus sensory processing on behaviorally relevant aspects of the visual world. One potential mechanism of attention is a change in the gain of sensory responses. However, changing gain at early stages could have multiple downstream consequences for visual processing. Which, if any, of these effects can account for the benefits of attention for detection and discrimination? Using a model of primate visual cortex we document how a Gaussian-shaped gain modulation results in changes to spatial tuning properties. Forcing the model to use only these changes failed to produce any benefit in task performance. Instead, we found that gain alone was both necessary and sufficient to explain category detection and discrimination during attention. Our results show how gain can give rise to changes in receptive fields which are not necessary for enhancing task performance.

Local field potentials, spiking activity, and receptive fields in human visual cortex.

The concept of receptive field (RF) is central to sensory neuroscience. Neuronal RF properties have been substantially studied in animals, while those in humans remain nearly unexplored. Here, we measured neuronal RFs with intracranial local field potentials (LFPs) and spiking activity in human visual cortex (V1/V2/V3). We recorded LFPs via macro-contacts and discovered that RF sizes estimated from low-frequency activity (LFA, 0.5-30 Hz) were larger than those estimated from low-gamma activity (LGA, 30-60 Hz) and high-gamma activity (HGA, 60-150 Hz). We then took a rare opportunity to record LFPs and spiking activity via microwires in V1 simultaneously. We found that RF sizes and temporal profiles measured from LGA and HGA closely matched those from spiking activity. In sum, this study reveals that spiking activity of neurons in human visual cortex could be well approximated by LGA and HGA in RF estimation and temporal profile measurement, implying the pivotal functions of LGA and HGA in early visual information processing.

Layers of inhibitory networks shape receptive field properties of AII amacrine cells

In the retina, rod and cone pathways mediate visual signals over a billion-fold range in luminance. AII ("A-two") amacrine cells (ACs) receive signals from both pathways via different bipolar cells, enabling AIIs to operate at night and during the day. Previous work has examined luminance-dependent changes in AII gap junction connectivity, but less is known about how surrounding circuitry shapes AII receptive fields across light levels. Here, we report that moderate contrast stimuli elicit surround inhibition in AIIs under all but the dimmest visual conditions, due to actions of horizontal cells and at least two ACs that inhibit presynaptic bipolar cells. Under photopic (daylight) conditions, surround inhibition transforms AII response kinetics, which are inherited by downstream ganglion cells. Ablating neuronal nitric oxide synthase type-1 (nNOS-1) ACs removes AII surround inhibition under mesopic (dusk/dawn), but not photopic, conditions. Our findings demonstrate how multiple layers of neural circuitry interact to encode signals across a wide physiological range.

Bio-inspired XYW parallel pathway edge detection network

Edge detection is of critical importance for middle-level and high-level tasks in computer vision. Existing edge detection methods usually use VGG16 as the encoding network and achieve exceptional performance through transfer learning, which has the characteristics of high parameters and high computational cost. Researchers have implemented edge detection by designing decoding networks. Unlike existing methods, this paper is inspired by the effective mechanisms of edge detection in the parallel pathway of biological vision and proposes a new lightweight encoding–decoding structure for edge detection networks, which we refer to as XYW-Net. In the encoding network, we drew inspiration from the receptive field properties of X-type cells, Y-type cells, and W-type cells involved in the parallel pathway, and developed more meticulous convolutional models. Based on the structure of the parallel pathway, the three cell modules are combined into a novel encoding network and excellent feature extraction performance is obtained. Within the decoding network, this paper draws inspiration from the feature integration ability of the inferotemporal cortex (IT) and introduces a novel feature integration module to constitute the decoding network. Experiments show that the proposed network in this paper obtains the Optimal Dataset Scale (ODS) = 0.812 on the BSDS500 dataset with only 0.79 M parameters. Additionally, the model has exhibited outstanding ODS performance on various publicly available edge detection datasets, including NYUD, BIPEDv1, and Multicue. Moreover, future research could concentrate on implementing more efficient antagonistic mechanisms and devising networks grounded in higher-level visual cortical mechanisms. This could further augment the performance and efficiency of edge detection networks. We encourage researchers to delve into these directions within the field. The codes are available at https://github.com/PXinTao/XYW-Net.

Contributed Session III: In vivo calcium imaging of macaque foveolar retinal ganglion cells reveals spatiochromatic receptive field properties.

Here, we optically record responses to spatial and chromatic stimuli using a calcium indicator in the living macaque eye to characterize the receptive field (RF) properties of retinal ganglion cells (RGCs) serving the foveal center. GCaMP6s was expressed in three female macaques. Adaptive optics ophthalmoscopy was used to image fluorescence (488nm ex, 520/35nm em) from RGCs whose RF centers were driven by cones in the central 36 arcmin of the fovea and additional RGCs driven by cones in the central 6 arcmin of the foveola. Using cone isolating and luminance flicker (1.3deg, 0.15Hz, LED 420nm, 530nm, 660nm), we derived cone weights in over 250 RGCs. Using drifting gratings (1.9deg, 6Hz, 4-50c/deg, 561nm), we derived the spatial frequency responses of 15 L vs. M cone opponent RGCs at the foveolar center. Employing computational modeling (ISETbio), we inferred the full spatial difference of gaussians center and surround structure for those 15 cells. Of the 34 foveolar RGCs, 44% exhibited L vs. M cone opponency, 15% were L+M ON, 6% were -L-M OFF, and 35% showed only L or only M responses. The spatial frequency response functions of 12/15 L vs. M opponent cells peaked at high spatial frequencies (25-40c/deg) and had a strong bandpass characteristic. Our model indicates that the responses of all 15 L vs. M opponent cells are consistent with single cone input to their RF centers and that our data are consistent with extrafoveal data when the blurring of the optics is accounted for.

Differences in non-linearities determine retinal cell types.

Classifying neurons in different types is still an open challenge. In the retina, recent works have taken advantage of the ability to record a large number of cells to classify ganglion cells into different types based on functional information. While the first attempts in this direction used the receptive field properties of each cell to classify them, more recent approaches have proposed to cluster ganglion cells directly based on their response to standard stimuli. These two approaches have not been compared directly. Here we recorded the responses of a large number of ganglion cells and compared two methods for classifying them into types, one based on the receptive field properties, and the other one using their responses to standard stimuli. We show that the stimulus-based approach allows separating more types than the receptive field-based method, leading to a better classification. This better granularity is due to the fact that the stimulus-based method takes into account not only the linear part of ganglion cell function, but also non-linearities. A careful characterization of non-linear processing is thus key to allow functional classification of sensory neurons.

Population receptive field properties change dynamically within milliseconds

Population receptive field properties change dynamically within milliseconds

Gain, not concomitant changes in spatial receptive field properties, improves task performance in a neural network attention model.

Attention allows us to focus sensory processing on behaviorally relevant aspects of the visual world. One potential mechanism of attention is a change in the gain of sensory responses. However, changing gain at early stages could have multiple downstream consequences for visual processing. Which, if any, of these effects can account for the benefits of attention for detection and discrimination? Using a model of primate visual cortex we document how a Gaussian-shaped gain modulation results in changes to spatial tuning properties. Forcing the model to use only these changes failed to produce any benefit in task performance. Instead, we found that gain alone was both necessary and sufficient to explain category detection and discrimination during attention. Our results show how gain can give rise to changes in receptive fields which are not necessary for enhancing task performance.

Retinal motion statistics during natural locomotion.

Walking through an environment generates retinal motion, which humans rely on to perform a variety of visual tasks. Retinal motion patterns are determined by an interconnected set of factors, including gaze location, gaze stabilization, the structure of the environment, and the walker's goals. The characteristics of these motion signals have important consequences for neural organization and behavior. However, to date, there are no empirical in situ measurements of how combined eye and body movements interact with real 3D environments to shape the statistics of retinal motion signals. Here, we collect measurements of the eyes, the body, and the 3D environment during locomotion. We describe properties of the resulting retinal motion patterns. We explain how these patterns are shaped by gaze location in the world, as well as by behavior, and how they may provide a template for the way motion sensitivity and receptive field properties vary across the visual field.

Impact of acute visual experience on development of LGN receptive fields in the ferret.

Selectivity for direction of motion is a key feature of primary visual cortical neurons. Visual experience is required for direction selectivity in carnivore and primate visual cortex, but the circuit mechanisms of its formation remain incompletely understood. Here we examined how developing lateral geniculate nucleus (LGN) neurons may contribute to cortical direction selectivity. Using in vivo electrophysiology techniques, we examined LGN receptive field properties of visually naïve female ferrets before and after exposure to 6 hours of motion stimuli in order to assess the effect of acute visual experience on LGN cell development. We found that acute experience with motion stimuli did not significantly affect the weak orientation or direction selectivity of LGN neurons. In addition, we found that neither latency nor sustainedness or transience of LGN neurons significantly changed with acute experience. These results suggest that the direction selectivity that emerges in cortex after acute experience is computed in cortex and cannot be explained by changes in LGN cells.SIGNIFICANCE STATEMENT:The development of typical neural circuitry requires experience-independent and experience-dependent factors. In the visual cortex of carnivores and primates, selectivity for motion arises as a result of experience, but we do not understand whether the major brain area that sits between the retina and the visual cortex - the lateral geniculate nucleus of the thalamus - also participates. Here we found that lateral geniculate neurons do not exhibit changes as a result of several hours of visual experience with moving stimuli, at a time when visual cortical neurons undergo a rapid change. We conclude that lateral geniculate neurons do not participate in this plasticity, and that changes in cortex are likely responsible for the development of direction selectivity in carnivores and primates.

Multi-finger receptive field properties in primary somatosensory cortex: A revised account of the spatiotemporal integration functions of area 3b

The leading view in the somatosensory system indicates that area 3b serves as a cortical relay site that primarily encodes (cutaneous) tactile features limited to individual digits. Our recent work argues against this model by showing that area 3b cells can integrate both cutaneous and proprioceptive information from the hand. Here, we further test the validity of this model by studying multi-digit (MD) integration properties in area 3b. In contrast to the prevailing view, we show that most cells in area 3b have a receptive field (RF) that extends to multiple digits, with the size of the RF (i.e., the number of responsive digits) increasing across time. Further, we show that MD cells' orientation angle preference is highly correlated across digits. Taken together, these data show that area 3b plays a larger role in generating neural representations of tactile objects, as opposed to just being a "feature detector" relay site.

Encoding fidelity of binocular receptive fields with internal noise in the presence of external variability from natural scenes

Characterizing the properties of receptive fields that are involved in encoding behaviorally-relevant latent variables (binocular disparity, motion) is a focus of visual neuroscience research. The factors that determine which receptive field properties are most useful for encoding a particular latent variable in natural scenes are not well-understood. Here, using a stereo-image database with groundtruth disparity information at each pixel and a nonlinear-linear-nonlinear subunit modeling framework, we examine how external variability and internal noise jointly determine the binocular receptive field properties that are most useful for encoding binocular disparity in natural scenes. We compute the encoding fidelity, as indexed by Fisher information, of binocular receptive field pairs across preferred spatial frequency (SF) and differences in preferred phase disparity (binocular phase shift). Two major findings emerge. First, stimulus disparity and external variability (image and depth variability) determine the usefulness of different encoding subspaces, as defined by the preferred SFs of the binocular receptive field pairs. The most useful preferred SF decreases with stimulus disparity but increases with local depth variability. Second, internal noise determines the usefulness within the encoding subspaces of different basis elements, as defined by the binocular phase shifts of the receptive fields. In particular, with internal noise, encoding fidelity varies with the particular receptive field pairs within the subspace. Non-orthogonal receptive fields are most useful, a result that deviates from the quadrature-pairs posited by the disparity energy model. Most subunit modeling frameworks do not inject noise into the individual subunit responses. This modeling choice renders all subunit receptive field pairs spanning a subspace computationally equivalent. Incorporating subunit response noise makes more specific predictions regarding the particular receptive fields that are most useful for encoding a particular latent variable. The current modeling efforts bring us closer to understanding the principles governing the design of real neural systems in natural scenes.