Search PubMed⌕ Search

Biomedical subjects

Daniel Kersten

Publications and source records attributed to Daniel Kersten.

15 recordsLinked to original sources

Vision as Bayesian inference: analysis by synthesis?

We argue that the study of human vision should be aimed at determining how humans perform natural tasks with natural images. Attempts to understand the phenomenology of vision from artificial stimuli, although worthwhile as a starting point, can lead to faulty generalizations about visual systems, because of the enormous complexity of natural images. Dealing with this complexity is daunting, but Bayesian inference on structured probability distributions offers the ability to design theories of vision that can deal with the complexity of natural images, and that use 'analysis by synthesis' strategies with intriguing similarities to the brain. We examine these strategies using recent examples from computer vision, and outline some important implications for cognitive science.

Algorithms↗

The representation of perceived angular size in human primary visual cortex.

Two objects that project the same visual angle on the retina can appear to occupy very different proportions of the visual field if they are perceived to be at different distances. What happens to the retinotopic map in primary visual cortex (V1) during the perception of these size illusions? Here we show, using functional magnetic resonance imaging (fMRI), that the retinotopic representation of an object changes in accordance with its perceived angular size. A distant object that appears to occupy a larger portion of the visual field activates a larger area in V1 than an object of equal angular size that is perceived to be closer and smaller. These results demonstrate that the retinal size of an object and the depth information in a scene are combined early in the human visual system.

Brain Mapping↗

Spatially specific FMRI repetition effects in human visual cortex.

The functional MRI (fMRI) response to a pair of identical, successively presented stimuli can result in a smaller signal than the presentation of two nonidentical stimuli. This "repetition effect" has become a frequently used tool to make inferences about neural selectivity in specific cortical areas. However, little is known about the mechanism(s) underlying the effect. In particular, despite many successful applications of the technique in higher visual areas, repetition effects in lower visual areas [e.g., primary visual cortex (V1)] have been more difficult to characterize. One property that is well understood in early visual areas is the mapping of visual field locations to specific areas of the cortex (i.e., retinotopy). We used the retinotopic organization of V1 to activate progressively different populations of neurons in a rapid fMRI experimental design. We observed a repetition effect (reduced signal) when localized stimulus elements were repeated in identical locations. We show that this effect is spatially tuned and largely independent of both interstimulus interval (100-800 ms) and the focus of attention. Using the same timing parameters for which we observed a large effect of spatial position, we also examined the response to orientation changes and observed no effect of an orientation change on the response to repeated stimuli in V1 but significant effects in other retinotopic areas. Given these results, we discuss the possible causes of these repetition effects as well as the implications for interpreting other experiments that use this potentially powerful imaging technique.

Female↗

Orientation-tuned FMRI adaptation in human visual cortex.

Adaptation is a general property of almost all neural systems and has been a longstanding tool of psychophysics because of its power to isolate and temporarily reduce the contribution of specific neural populations. Recently, adaptation designs have been extensively applied in functional MRI (fMRI) studies to infer neural selectivity in specific cortical areas. However, there has been considerable variability in the duration of adaptation used in these experiments. In particular, although long-term adaptation has been solidly established in psychophysical and neurophysiological studies, it has been incorporated into few fMRI studies. Furthermore, there has been little validation of fMRI adaptation using stimulus dimensions with well-known adaptive properties (e.g., orientation) and in better understood regions of cortex (e.g., primary visual cortex, V1). We used an event-related fMRI experiment to study long-term orientation adaptation in the human visual cortex. After long-term adaptation to an oriented pattern, the fMRI response in V1, V2, V3/VP, V3A, and V4 to a test stimulus was proportional to the angular difference between the adapting and test stimuli. However, only V3A and V4 showed this response pattern with short-term adaptation. In a separate experiment, we measured behavioral contrast detection thresholds after adaptation and found that the fMRI signal in V1 closely matched the psychophysically derived contrast detection thresholds. Similar to the fMRI results, adaptation induced threshold changes strongly depended on the duration of adaptation. In addition to supporting the existence of adaptable orientation-tuned neurons in human visual cortex, our results show the importance of considering timing parameters in fMRI adaptation experiments.

Adaptation, Physiological↗

Is prior knowledge of object geometry used in visually guided reaching?

We investigated whether humans use prior knowledge of the geometry of faces in visually guided reaching. When viewing the inside of a mask of a face, the mask is often perceived as being a normal (convex) face, instead of the veridical, hollow (concave) shape. In this "hollow-face illusion," prior knowledge of the shape of faces dominates perception, even when in conflict with information from binocular disparity. Computer images of normal and hollow faces were presented, such that depth information from binocular disparity was consistent or in conflict with prior knowledge of the geometry. Participants reached to touch either the nose or cheek of the faces or gave verbal estimates of the corresponding distances. We found that reaching to touch was dominated by prior knowledge of face geometry. However, hollow faces were estimated to be flatter than normal faces. This suggests that the visual system combines binocular disparity and prior assumptions, rather than completely discounting one or the other. When comparing the magnitude of the hollow-face illusion in reaching and verbal tasks, we found that the flattening effect of the illusion was similar for verbal and reaching tasks.

Cognition↗

BOLD fMRI and psychophysical measurements of contrast response to broadband images.

We have measured the relationship between image contrast, perceived contrast, and BOLD fMRI activity in human early visual areas, for natural, whitened, pink noise, and white noise images. As root-mean-square contrast increases, BOLD response to natural images is stronger and saturates more rapidly than response to the whitened images. Perceived contrast and BOLD fMRI responses are higher for pink noise than for white noise patterns, by the same ratio as between natural and whitened images. Spatial phase structure has no measurable effect on perceived contrast or BOLD fMRI response. The fMRI and perceived contrast response results can be described by models of spatial frequency response in V1, that match the contrast sensitivity function at low contrasts, and have more uniform spatial frequency response at high contrasts.

Adult↗

Object perception as Bayesian inference.

We perceive the shapes and material properties of objects quickly and reliably despite the complexity and objective ambiguities of natural images. Typical images are highly complex because they consist of many objects embedded in background clutter. Moreover, the image features of an object are extremely variable and ambiguous owing to the effects of projection, occlusion, background clutter, and illumination. The very success of everyday vision implies neural mechanisms, yet to be understood, that discount irrelevant information and organize ambiguous or noisy local image features into objects and surfaces. Recent work in Bayesian theories of visual perception has shown how complexity may be managed and ambiguity resolved through the task-dependent, probabilistic integration of prior object knowledge with image features.

Bayes Theorem↗

Visuomotor sensitivity to visual information about surface orientation.

We measured human visuomotor sensitivity to visual information about three-dimensional surface orientation by analyzing movements made to place an object on a slanted surface. We applied linear discriminant analysis to the kinematics of subjects' movements to surfaces with differing slants (angle away form the fronto-parallel) to derive visuomotor d's for discriminating surfaces differing in slant by 5 degrees. Subjects' visuomotor sensitivity to information about surface orientation was very high, with discrimination "thresholds" ranging from 2 to 3 degrees. In a first experiment, we found that subjects performed only slightly better using binocular cues alone than monocular texture cues and that they showed only weak evidence for combining the cues when both were available, suggesting that monocular cues can be just as effective in guiding motor behavior in depth as binocular cues. In a second experiment, we measured subjects' perceptual discrimination and visuomotor thresholds in equivalent stimulus conditions to decompose visuomotor sensitivity into perceptual and motor components. Subjects' visuomotor thresholds were found to be slightly greater than their perceptual thresholds for a range of memory delays, from 1 to 3 s. The data were consistent with a model in which perceptual noise increases with increasing delay between stimulus presentation and movement initiation, but motor noise remains constant. This result suggests that visuomotor and perceptual systems rely on the same visual estimates of surface slant for memory delays ranging from 1 to 3 s.

Acceleration↗

Bayesian models of object perception.

The human visual system is the most complex pattern recognition device known. In ways that are yet to be fully understood, the visual cortex arrives at a simple and unambiguous interpretation of data from the retinal image that is useful for the decisions and actions of everyday life. Recent advances in Bayesian models of computer vision and in the measurement and modeling of natural image statistics are providing the tools to test and constrain theories of human object perception. In turn, these theories are having an impact on the interpretation of cortical function.

Animals↗

Is color an intrinsic property of object representation?

The role of color in object representation was examined by using a variation of the Stroop paradigm in which observers named the displayed colors of objects or words. In experiment 1, colors of color-diagnostic objects were manipulated to be either typical or atypical of the object (eg a yellow banana versus a purple banana). A Stroop-like effect was obtained, with faster color-naming times for the typical as compared to the atypical condition. In experiment 2, naming colors on words specifying these same color-diagnostic objects reversed this pattern, with the typical condition producing longer response times than the atypical condition. In experiment 3, a blocked condition design that used the same words and colors as experiment 2 produced the standard Stroop-like facilitation for the typical condition. These results indicate that color is an intrinsic property of an object's representation at multiple levels. In experiment 4, we examined the specific level(s) at which color-shape associations arise by following the tasks used in experiments 1 and 2 with a lexical-decision task in which some items were conceptually related to items shown during color naming (eg banana/monkey). Priming for these associates was observed following color naming of words, but not pictures, providing further evidence that the color-shape associations responsible for the differing effects obtained in experiments 1 and 2 are due to the automatic activation of color-shape associations at different levels of representation.

Adult↗

Bootstrapped learning of novel objects.

Recognition of familiar objects in cluttered backgrounds is a challenging computational problem. Camouflage provides a particularly striking case, where an object is difficult to detect, recognize, and segment even when in "plain view." Current computational approaches combine low-level features with high-level models to recognize objects. But what if the object is unfamiliar? A novel camouflaged object poses a paradox: A visual system would seem to require a model of an object's shape in order to detect, recognize, and segment it when camouflaged. But, how is the visual system to build such a model of the object without easily segmentable samples? One possibility is that learning to identify and segment is opportunistic in the sense that learning of novel objects takes place only when distinctive clues permit object segmentation from background, such as when target color or motion enables segmentation on single presentations. We tested this idea and discovered that, on the contrary, human observers can learn to identify and segment a novel target shape, even when for any given training image the target object is camouflaged. Further, perfect recognition can be achieved without accurate segmentation. We call the ability to build a shape model from high-ambiguity presentations bootstrapped learning.

Animals↗

Three-dimensional symmetric shapes are discriminated more efficiently than asymmetric ones.

Objects with bilateral symmetry, such as faces, animal shapes, and many man-made objects, play an important role in everyday vision. Because they occur frequently, it is reasonable to conjecture that the brain may be specialized for symmetric objects. We investigated whether the human visual system processes three-dimensional (3D) symmetric objects more efficiently than asymmetric ones. Human subjects, having learned a symmetric wire object, discriminated which of two distorted copies of the learned object was more similar to the learned one. The distortion was achieved by adding 3D Gaussian positional perturbations at the vertices of the wire object. In the asymmetric condition, the perturbation was independent from one vertex to the next. In the symmetric condition, independent perturbations were added to only half of the object; perturbations on the other half retained the symmetry of the object. We found that subjects' thresholds were higher in the symmetric condition. However, since the perturbation in the symmetric condition was correlated, a stimulus image provided less information in the symmetric condition. Taking this in to consideration, an ideal-observer analysis revealed that subjects were actually more efficient at discriminating symmetric objects. This reversal in interpretation underscores the importance of ideal-observer analysis. A completely opposite, and wrong, conclusion would have been drawn from analyzing only human discrimination thresholds. Given the same amount of information, the visual system is actually better able to discriminate symmetric objects than asymmetric ones.

Differential Threshold↗

Shape perception reduces activity in human primary visual cortex.

Visual perception involves the grouping of individual elements into coherent patterns that reduce the descriptive complexity of a visual scene. The physiological basis of this perceptual simplification remains poorly understood. We used functional MRI to measure activity in a higher object processing area, the lateral occipital complex, and in primary visual cortex in response to visual elements that were either grouped into objects or randomly arranged. We observed significant activity increases in the lateral occipital complex and concurrent reductions of activity in primary visual cortex when elements formed coherent shapes, suggesting that activity in early visual areas is reduced as a result of grouping processes performed in higher areas. These findings are consistent with predictive coding models of vision that postulate that inferences of high-level areas are subtracted from incoming sensory information in lower areas through cortical feedback.

Brain Mapping↗

Perceptual grouping and the interactions between visual cortical areas.

Visual perception involves the grouping of individual elements into coherent patterns, such as object representations, that reduce the descriptive complexity of a visual scene. The computational and physiological bases of this perceptual remain poorly understood. We discuss recent fMRI evidence from our laboratory where we measured activity in a higher object processing area (LOC), and in primary visual cortex (V1) in response to visual elements that were either grouped into objects or randomly arranged. We observed significant activity increases in the LOC and concurrent reductions of activity in V1 when elements formed coherent shapes, suggesting that activity in early visual areas is reduced as a result of grouping processes performed in higher areas. In light of these results we review related empirical findings of context-dependent changes in activity, recent neurophysiology research related to cortical feedback, and computational models that incorporate feedback operations. We suggest that feedback from high-level visual areas reduces activity in lower areas in order to simplify the description of a visual image-consistent with both predictive coding models of perception and probabilistic notions of 'explaining away.'

Animals↗