Global Shape Processing Research Articles

Interdependency in lateralization of written word and face processing in right-handed individuals

It has been suggested that the right hemisphere lateralization typically observed for face processing may depend on lateralization of written word processing to the left hemisphere; a pattern referred to as the causal complementary principle of lateralization. According to a strong version of this principle, a correlation should be found between the degree of left and right hemisphere lateralization for word and face processing respectively. This has been observed in two studies, but only for left-handed individuals. The present study tested whether a similar lateralization pattern could be found in a relatively large sample of right-handed individuals (N = 210) using behavioral measures (divided visual field paradigms). It was also tested whether the degree of right hemisphere lateralization for face and global shape processing would correlate positively, as predicted by a strong version of the input asymmetry principle of lateralization. This was tested in a subsample (n = 189). Bayesian analyses found no evidence for lateralization interdependency as the observed data were 4–17 times more likely under the null hypothesis. Unfortunately, the reliabilities of the lateralization measures were found to be poor. While this dampens the firmness of the conclusions that can be drawn, it is argued that at present there is no positive evidence for strong interdependency between written word and face processing in right-handed individuals.

The Emergence of Tuning to Global Shape Properties of Radial Frequency Patterns in the Ventral Visual Pathway.

Radial frequency (RF) patterns, created by sinusoidal modulations of a circle's radius, are processed globally when RF is low. These closed shapes therefore offer a useful way to interrogate the human visual system for global processing of curvature. RF patterns elicit greater responses than those to radial gratings in V4 and more anterior face-selective regions of the ventral visual pathway. This is largely consistent with work on nonhuman primates showing curvature processing emerges in V4, but is evident also higher up the ventral visual stream. Rather than contrasting RF patterns with other stimuli, we presented them at varied frequencies in a regimen that allowed tunings to RF to be derived from 8 human participants (3 female). We found tuning to low RF in lateral occipital areas and to some extent in V4. In a control experiment, we added a high-frequency ripple to the stimuli disrupting the local contour. Low-frequency tuning to these stimuli remained in the ventral visual stream, underscoring its role in global processing of shape curvature. We then used representational similarity analysis to show that, in lateral occipital areas, the neural representation was related to stimulus similarity, when it was computed with a model that captured how stimuli are perceived. We therefore show that global processing of shape curvature emerges in the ventral visual stream as early as V4, but is found more strongly in lateral occipital regions, which exhibit responses and representations that relate well to perception.SIGNIFICANCE STATEMENT We show that tuning to low radial frequencies, known to engage global shape processing mechanisms, was localized to lateral occipital regions. When low-level stimulus properties were accounted for such tuning emerged in V4 and LO2 in addition to the object-selective region LO. We also documented representations of global shape properties in lateral occipital regions, and these representations were predicted well by a proxy of the perceptual difference between the stimuli.

Motor Skills and Visual Deficits in Developmental Coordination Disorder: A Narrative Review.

Developmental coordination disorder (DCD) is a developmental disorder in which numerous comorbidities seem to coexist, such as motor and visual impairment and some executive functions; Methods: A narrative review on motor and visual deficits in children with DCD was carried out; Results and Discussion: Fine and gross motor skills are affected in children with DCD. In addition, they seem to be related to visual deficits, such as difficulty in visual perception, sensory processing and visual memory. Limitations have also been found in accommodation. Interventions in children with DCD should be aimed at improving both aspects, since vision affects motor skills and vice versa; Conclusions: In children with DCD, who present a marked deficit in global shape processing, it causes an association between deficiencies in visual perception and motor skills.

Capsule networks as recurrent models of grouping and segmentation.

Classically, visual processing is described as a cascade of local feedforward computations. Feedforward Convolutional Neural Networks (ffCNNs) have shown how powerful such models can be. However, using visual crowding as a well-controlled challenge, we previously showed that no classic model of vision, including ffCNNs, can explain human global shape processing. Here, we show that Capsule Neural Networks (CapsNets), combining ffCNNs with recurrent grouping and segmentation, solve this challenge. We also show that ffCNNs and standard recurrent CNNs do not, suggesting that the grouping and segmentation capabilities of CapsNets are crucial. Furthermore, we provide psychophysical evidence that grouping and segmentation are implemented recurrently in humans, and show that CapsNets reproduce these results well. We discuss why recurrence seems needed to implement grouping and segmentation efficiently. Together, we provide mutually reinforcing psychophysical and computational evidence that a recurrent grouping and segmentation process is essential to understand the visual system and create better models that harness global shape computations.

Local features and global shape information in object classification by deep convolutional neural networks

Deep convolutional neural networks (DCNNs) show impressive similarities to the human visual system. Recent research, however, suggests that DCNNs have limitations in recognizing objects by their shape. We tested the hypothesis that DCNNs are sensitive to an object’s local contour features but have no access to global shape information that predominates human object recognition. We employed transfer learning to assess local and global shape processing in trained networks. In Experiment 1, we used restricted and unrestricted transfer learning to retrain AlexNet, VGG-19, and ResNet-50 to classify circles and squares. We then probed these networks with stimuli with conflicting global shape and local contour information. We presented networks with overall square shapes comprised of curved elements and circles comprised of corner elements. Networks classified the test stimuli by local contour features rather than global shapes. In Experiment 2, we changed the training data to include circles and squares comprised of different elements so that the local contour features of the object were uninformative. This considerably increased the network’s tendency to produce global shape responses, but deeper analyses in Experiment 3 revealed the network still showed no sensitivity to the spatial configuration of local elements. These findings demonstrate that DCNNs’ performance is an inversion of human performance with respect to global and local shape processing. Whereas abstract relations of elements predominate in human perception of shape, DCNNs appear to extract only local contour fragments, with no representation of how they spatially relate to each other to form global shapes.

Who's got the global advantage? Visual field differences in processing of global and local shape

Much evidence suggests that we first perceive the overall layout of a scene or object followed later by the details. This coarse-to-fine temporal dynamic in visual processing is also found in Navon's classical paradigm where information at the global level of compound stimuli is processed faster than information at the local level (global precedence effect), and where information at the global level has larger effects on local level responses than local level information has on global level responses (asymmetric interference effects). Traditionally, global shape primacy in Navon's paradigm has been linked with a right hemisphere preference (left visual field advantage) for global shape processing, and a left hemisphere preference (right visual field advantage) for local shape processing. This link, however, has been based on measures which confound global precedence and interference effects. Indeed, when these measures are de-confounded, we find no evidence for larger global precedence effects in the left compared with the right visual field in a large sample of participants (N = 337). In comparison, global-to-local interference effects are found to be stronger in the left than in the right visual field. We argue that these findings can be accounted for by assuming that the right hemisphere plays a special role in integrating shape information across spatial scales, that is, without assuming the existence of a right hemisphere preference for global shape processing per se.

Convergent evidence for global processing of shape

There is an ongoing debate over whether there is convincing evidence in support of global contour integration in shape discrimination tasks, particularly when using radial frequency (RF) patterns as stimuli (Baldwin, Schmidtmann, Kingdom, & Hess, 2016). The objection lies in the previous use of high-threshold theory (HTT), rather than signal detection theory (SDT) to model the probability summation estimates of observer thresholds to determine whether integration of information is occurring around the contour. Here we used a discrimination at threshold method to establish evidence of global processing of two frequently used RF patterns (RF3 and RF5) that does not require mathematical modeling. To provide a bridge between current and past research we examined the two proposed methods, finding that HTT produced probability summation estimates that were more conservative than SDT (when an appropriate number of channels was used to generate estimates). We found no difference in observer thresholds when an RF pattern was presented as the only test stimulus in a block of trials or when two RF patterns were interleaved, and no evidence for a decrease in psychometric slopes with increasing numbers of stimulus elements. These findings are contrary to what is predicted by SDT for a stimulus whose detection conforms to probability summation. Therefore, our results find no evidence that supports probability summation, further demonstrating the importance of using random phase RF patterns while measuring integration around a contour and providing strong evidence for global shape processing around low frequency RF patterns.

Delayed processing of global shape in developmental prosopagnosia

Delayed processing of global shape in developmental prosopagnosia

Global shape aftereffects in composite radial frequency patterns.

Individual radial frequency (RF) patterns are generated by modulating a circle's radius as a sinusoidal function of polar angle and have been shown to tap into global shape processing mechanisms. Composite RF patterns can reproduce the complex outlines of natural shapes and examining these stimuli may allow us to interrogate global shape mechanisms that are recruited in biologically relevant tasks. We present evidence for a global shape aftereffect in a composite RF pattern stimulus comprising two RF components. Manipulations of the shape, location, size and spatial frequency of the stimuli revealed that this aftereffect could only be explained by the attenuation of intermediate-level global shape mechanisms. The tuning of the aftereffect to test stimulus size also revealed two mechanisms underlying the aftereffect; one that was tuned to size and one that was invariant. Finally, we show that these shape mechanisms may encode some RF information. However, the RF encoding we found was not capable of explaining the full extent of the aftereffect, indicating that encoding of other shape features such as curvature are also important in global shape processing.

Global shape processing: A behavioral and electrophysiological analysis of both contour and texture processing.

The number of corners on the boundary of a closed contour is thought to be particularly critical for shape detection and discrimination. The aim of the current study was to examine the relative contribution of the number of corners and the angle between corners to shape discrimination in complex visual scenes as well as to determine the time course and neural substrates of global shape processing based on the presence or absence of these specific features. In Experiment 1, event-related potentials were recorded while participants discriminated between two radial frequency (RF) patterns with the same maximum local curvature defining corners but varying arrangements of those corners. The results showed that the angle separating corners was more critical than the overall number of corners for discrimination performance. An enhanced negativity (posterior N220) over the occipital lobe was elicited following the presentation of an RF with three modulation cycles (RF3) but not following a circle, suggesting that the posterior N220 is sensitive to variation in curvature on a contour. In Experiment 2, we confirm the primary effect of the presence of corners on the amplitude of the posterior N220 component and extend the stimuli to include shapes defined by texture. Source localization on the N170 and N220 components was conducted in Experiment 2, and a source in cortical area V4 was identified. These findings suggest that corners contain vital information for the discrimination of shapes. Additionally, this study shows that the perceptual characteristics and neuroanatomical substrates can be detected using electrophysiological measures.

Cortical contributions to impaired contour integration in schizophrenia

ObjectivesVisual perceptual organization impairments in schizophrenia (SCZ) are well established, but their neurobiological bases are not. The current study used the previously validated Jittered Orientation Visual Integration (JOVI) task, along with fMRI, to examine the neural basis of contour integration (CI), and its impairment in SCZ. CI is an aspect of perceptual organization in which multiple distinct oriented elements are grouped into a single continuous boundary or shape. MethodsOn the JOVI, five levels of orientational jitter were added to non-contiguous closed contour elements embedded in background noise to progressively increase the difficulty in perceiving contour elements as left- or right-pointing ovals. Multi-site fMRI data were analyzed for 56 healthy control subjects and 47 people with SCZ. ResultsSCZ patients demonstrated poorer CI, and this was associated with increased activation in regions involved in global shape processing and visual attention, namely the lateral occipital complex and superior parietal lobules. There were no brain regions where controls demonstrated more activation than patients. ConclusionsCI impairment in this sample of outpatients with SCZ was related to excessive activation in regions associated with object processing and allocation of visual–spatial attention. There was no evidence for basic impairments in contour element linking in the fMRI data. The latter may be limited to poor outcome patients, where more extensive structural and functional changes in the occipital lobe have been observed.

Differential electrophysiological responses to biological motion in children and adults with and without autism spectrum disorders

Although atypical processing of biological motion (BM) in individuals with autism spectrum disorder (ASD) has been reported, the temporal profile of the neural response to BM is not well explored. In the current study, event-related potentials (ERPs) were measured in 12 individuals with ASD, aged 8–22 years, and 12 age- and gender-matched normal controls, to investigate the electrophysiological response to BM and a control visual stimulus. By introducing a novel experimental paradigm that can dissociate the electrophysiological responses to motion processing and the global shape processing of BM, we found that: (1) the timing of the response was preserved in ASD groups, whereas (2) the ERP response to BM was significantly enhanced compared with scrambled point-light motion (SM) in normal controls; the responses to both BM and SM were not significantly different in subjects with ASD. Because we did not find a significant group effect on the peak and mean amplitude induced by BM, it is presumed that this atypical response in individuals with ASD was due to over-sensitivity to the local motion signals. This experimental paradigm showed atypical local motion processing of BM in individuals with ASD.

Global shape processing involves feature-selective and feature-agnostic coding mechanisms.

Recent research and modeling proposes that a closed shape is accurately described by both the curvature and angular location of its parts relative to the shape center, implying that the shape is coded along with its overall orientation. We tested this proposition. Radial frequency (RF) patterns were employed as stimuli as they can represent a range of familiar closed shapes and are processed globally. We measured a RF amplitude aftereffect (RFAAE) as a function of the shape orientation difference between adapt and test patterns of the same RF. For RF3 and RF4, RFAAEs were largest when adapt and test patterns were the same orientation, and then linearly decreased as the adaptor was rotated away from the test. RFAAEs did not, however, reach zero, instead plateauing significantly above zero. On the other hand, when adapt and test were of opposite luminance polarity, RFAAEs, although lower than same luminance-polarity RFAAEs, were invariant to differences between adapt and test orientations. Our findings provide evidence for two global shape mechanisms: one that is selective for shape orientation and luminance polarity, and one that is agnostic to these characteristics.

Global Shape Processing Deficits Are Amplified by Temporal Masking in Migraine

Individuals with migraine show subtle defects in a range of visual tasks compared to nonmigraineurs. Increased neuronal noise can account for some of these deficits. To examine the generality of increased noise in migraine, masking effects were compared in migraineurs and headache-free controls using a shape discrimination task, thought to involve processing in extrastriate cortical areas. Nine migraineurs with aura, nine migraineurs without aura, and nine headache-free controls participated. observers had to detect deviations in circular shapes with or without a larger contour mask. The nonoverlapping mask was presented at five temporal intervals (stimulus onset asynchronies, SOA): 0 (simultaneous), 66, 100, 133, and 250 ms. Migraineurs with aura performed worse in all tests than migraineurs without aura and controls. Both migraine groups performed poorer than controls at discriminating shapes without masks. Typical masking functions were obtained from all groups, but they were steeper for migraineurs than controls with thresholds raised most dramatically (2.1 and 4.4 times for migraineurs without and with aura relative to controls, respectively) at SOAs where masks had their most detrimental effect (66-100 ms). Modeling the effect of masking showed that raised internal noise alone is insufficient to explain these deficits. Rather, an abnormal nonlinear transducer function (e.g., as part of gain-control) together with increased multiplicative noise is required to capture the data. The findings are consistent with an extrastriate deficit in migraine that cannot be explained completely by defective inhibition.

Global shape processing involves a hierarchy of integration stages

Radial Frequency (RF) patterns can be used to study the processing of familiar shapes, e.g. triangles and squares. Opinion is divided over whether the mechanisms that detect these shapes integrate local orientation and position information directly, or whether local orientations and positions are first combined to represent extended features, such as curves, and that it is local curvatures that the shape mechanism integrates. The latter view incorporates an intermediate processing stage, the former does not. To differentiate between these hypotheses we studied the processing of micro-patch sampled RF patterns as a function of the luminance polarity of successive elements on the contour path. Our first study measures shape after effects involving suprathreshold amplitude RF shapes and shows that alternating the luminance polarity of successive micro-patch elements disrupts adaptation of the global shape. Our second study shows that polarity alternations also disrupt sensitivity to threshold-amplitude RF patterns. These results suggest that neighbouring points of the contour shape are integrated into extended features by a polarity selective mechanism, prior to global shape processing, consistent with the view that for both threshold amplitude and suprathreshold amplitude patterns, global processing of RF shapes involves an intermediate stage of processing.

Global shape processing: Which parts form the whole?

Research suggests that detection of low-frequency radial frequency (RF) patterns involves global shape processing and that points of maximum curvature (corners) contribute more than points of minimum curvature (sides). However, this has only been tested with stimuli presented at the threshold of discriminability from a circle. We used RF pattern adaptation to (a) examine whether a supra-threshold RF pattern is processed as a global shape and (b) determine what the critical features are for representing its shape. We measured the perceived amplitude shift of an RF test pattern after prolonged exposure either to a higher amplitude pattern or to various combinations of its parts (concave maxima, convex maxima, inflections). We found greater shifts in perceived amplitude after adaptation to a "whole" pattern than after adaptation to its component parts, which alternated to produce equal net contrast. Furthermore, when adapting to specific parts of the shape in isolation, we found that each part generated a similar magnitude aftereffect. Although the whole is clearly greater than the sum of the parts, we find that concave maxima, convex maxima, and inflections contribute equally to global shape processing, a fact that is only apparent when using a supra-threshold appearance-based task.

Global or local shape discrimination of radial frequency patterns?

Purpose. Color vision performs below luminance vision on global shape discrimination measured with radial frequency patterns (Mullen & Beaudot, Vision Research, 2002). This effect might reflect a genuine deficit of color on global shape processing, or it might originate from the fact that color vision is poorer at encoding local orientation. In a series of experiments we test and compare the role of local orientation cues in the shape discrimination of radial frequency (RF) patterns for chromatic and achromatic vision. Methods. Stimuli are contours described by the 4th derivative of a gaussian and form radial patterns (global task) or parts of the pattern (local task). We designed the following tasks in order to separate the role of local-orientation and global cues in the shape discrimination task: 1) we compare performance for the discrimination of local parts (corners or sides) of a RF pattern versus circular arcs of the same radius; 2) we compare the performance for the detection of line modulation versus radial frequency modulation for otherwise identical contours; 3) we compare the effect on shape discrimination of varying the spatial frequencies of the contour of the radial frequency pattern. Data were fitted with a model that uses changes in local curvature to discriminate shape. Results. We find very little performance difference between the local-orientation based tasks and the global shape discrimination task, regardless of whether stimuli were chromatic or achromatic. Conclusions. We conclude that performance on the shape discrimination of radial frequency patterns is limited by local orientation cues in our experiments. While RF patterns presumably exploit a global shape processing mechanism, this affords little or no performance advantages in the task.

Emergence of global shape processing continues through adolescence.

The developmental trajectory of perceptual organization in humans is unclear. This study investigated perceptual grouping abilities across a wide age range (8-30 years) using a classic compound letter global/local (GL) task and a more fine-grained microgenetic prime paradigm (MPP) with both few- and many-element hierarchical displays. In the GL task, contrary to adults, both children and adolescents exhibited a classic local bias. In the MPP, all 3 age groups evinced a bias to individuate the few-element displays; however, the ability to encode the global shape of the many-element displays at the short prime durations increased with age. These results indicate that the full process of garnering shape information from perceptual grouping, which is essential for the ability to do fast and efficient object recognition and identification, develops late into adolescence.

Missing the big picture: impaired development of global shape processing in autism

Individuals with autism exhibit hypersensitivity to local elements of the input, which may interfere with the ability to group visual elements perceptually. We investigated the development of perceptual grouping abilities in high-functioning individuals with autism (HFA) across a wide age range (8-30 years) using a classic compound letter global/local (GL) task and a more fine-grained microgenetic prime paradigm (MPP), including both few- and many-element hierarchical displays. In the GL task, contrary to the typically developing (TD) controls, HFA participants did not develop an increasing sensitivity to the global information with age. In the MPP, like the TD controls, individuals with autism at all three age groups evinced a bias to individuate the few-element displays. However, contrary to the TD controls, the HFA group failed to show age-related improvements in the ability to encode the global shape of the many-element displays. In fact, across the age range, the HFA group was consistently faster than the TD controls at perceiving the local elements in these displays. These results indicate that in autism the full process of garnering shape information from perceptual grouping, which is essential for the ability to do fast and efficient object recognition and identification, never matures, and this is especially evident in adolescence when this ability begins to improve in TD individuals. The atypical development of these perceptual organizational abilities may disrupt processing of visually presented objects, which may, in turn, fundamentally impede the development of major aspects of the social and emotional behaviors in individuals with autism.

Maintaining Subspecialty Certification in Neuroradiology

<h3>Abstract</h3> Classically, visual processing is described as a cascade of local feedforward computations. Feedforward Convolutional Neural Networks (ffCNNs) have shown how powerful such models can be. Previously, using visual crowding as a well-controlled challenge, we showed that no classic model of vision, including ffCNNs, can explain human global shape processing (1). Here, we show that Capsule Neural Networks (CapsNets; 2), combining ffCNNs with a grouping and segmentation mechanism, solve this challenge. We also show that ffCNNs and standard recurrent networks do not, suggesting that the grouping and segmentation capabilities of CapsNets are crucial. Furthermore, we provide psychophysical evidence that grouping and segmentation is implemented recurrently in humans, and show that CapsNets reproduce these results well. We discuss why recurrence seems needed to implement grouping and segmentation efficiently. Together, we provide mutually reinforcing psychophysical and computational evidence that a recurrent grouping and segmentation process is essential to understand the visual system and create better models that harness global shape computations. <h3>Author Summary</h3> Feedforward Convolutional Neural Networks (ffCNNs) have revolutionized computer vision and are deeply transforming neuroscience. However, ffCNNs only roughly mimic human vision. There is a rapidly expanding literature investigating differences between humans and ffCNNs. Several findings suggest that, unlike humans, ffCNNs rely mostly on local visual features. Furthermore, ffCNNs lack recurrent connections, which abound in the brain. Here, we use visual crowding, a well-known psychophysical phenomenon, to investigate recurrent computations in global shape processing. Previously, we showed that no model based on the classic feedforward framework of vision, including ffCNNs, can explain global effects in crowding. Here, we show that Capsule Networks (CapsNets), combining ffCNNs with recurrent grouping and segmentation, solve this challenge. Lateral and top-down recurrent connections do not, suggesting that grouping and segmentation are crucial for human-like global computations. Based on these results, we hypothesize that one computational function of recurrence is to efficiently implement grouping and segmentation. We provide psychophysical evidence that, indeed, recurrent processes implement grouping and segmentation in humans. CapsNets reproduce these results too. Together, we provide mutually reinforcing computational and psychophysical evidence that a recurrent grouping and segmentation process is essential to understand the visual system and create better models that harness global shape computations.