Speed of processing in the human visual system

Mammalian visual systems are characterized by their ability to recognize stimuli invariant to various transformations. Here, we investigate the hypothesis that this ability is achieved by the temporal encoding of visual stimuli. By using a model of a cortical network, we show that this encoding is invariant to several transformations and robust with respect to stimulus variability. Furthermore, we show that the proposed model provides a rapid encoding, in accordance with recent physiological results. Taking into account properties of primary visual cortex, the application of the encoding scheme to an enhanced network demonstrates favorable scaling and high performance in a task humans excel at. M ammals demonstrate highly evolved visual object recognition skills, tolerating considerable changes in images caused by, for instance, different viewing angles and deformations. Elucidating the mechanisms of such invariant pattern recognition is an active field of research in neuroscience (1-5). However, very little is known about the underlying algorithms and mechanisms. A number of models have been proposed which aim to reproduce capabilities of the biological visual system, such as invariance to shifts in position, rotation, and scaling (6-8). Most of these models are based on the ''Neocognitron'' (9), a hierarchical multilayer network of spatial feature detectors. As a result of a gradual increase of receptive field sizes, translation-invariant representations emerge in the form of activity patterns at the highest level. These models do not consider time as a coding dimension for neural representations. Recently, however, the importance of the temporal dynamics of neuronal activity in representing visual stimuli has gained increased attention (10, 11). Hence, it seems timely to consider the role of temporal coding in the context of tasks like invariant object recognition. In recent years, several modeling studies have addressed properties of temporal codes (12)(13)(14). For instance, Buonomano and Merzenich (15) proposed a model for positioninvariant pattern recognition which uses temporal coding. In this model, feed-forward inhibition modulates the spike-timing such that stimuli are represented by the response latencies of the neurons in the network. This architecture naturally leads to translation invariant representations. This model assigns a critical role to inhibitory interactions in the feed-forward path of the visual system (retina-LGN-V1), whereas anatomical studies suggest that these connections are predominantly excitatory (16). Furthermore, the majority of inputs to cortical neurons are excitatory and of cortical origin (17). Indeed, a recent theoretical study has shown that lateral excitatory coupling has pronounced effects on the global network dynamics (18). In particular, the combination of intracortical connectivity and dendritic processing allowed context-dependent representations of different stimuli to be expressed in the temporal dynamics of the network. Here, we build on these previous proposal...

show abstract

“…1e). This property of our encoding scheme is compatible with the impressive speed of processing found in the mammalian visual system (22).…”

Section: Resultssupporting

confidence: 70%

Invariant representations of visual patterns in a temporal population code

Wyss

König

Verschure

2002

Proc. Natl. Acad. Sci. U.S.A.

View full text Add to dashboard Cite

show abstract

“…In other words, a low-level visual property, such as LSF, might provide the foundation for a high-level cognitive faculty, such as categorization. Indeed, event-related potentials studies (56,57) have repeatedly shown that humans can distinguish between the category of objects in scenes extremely rapidly, which we argue can be mediated by coarse LSF representations, and that this distinction is first apparent in prefrontal cortex time courses Ϸ150 ms from stimulus onset. That such prefrontal representations subsequently guide the activation of IT representations is supported by fMRI studies in human patients (9,58) and electrophysiology recordings in monkeys (10).…”

Section: Discussionmentioning

confidence: 69%

Top-down facilitation of visual recognition

Bär

Kassam

Ghuman

et al. 2006

Proc. Natl. Acad. Sci. U.S.A.

1,379

183

1,242

View full text Add to dashboard Cite

10.1073͞pnas.0504604102), the authors note that Eq. 1 was incorrectly given as P ϭ f ssͩ ͩ1 ϩ m ͪ ϩ Dͩ 2 ϩ m ͪͪ both in the text and in Fig. 2. The correct equation is as follows:The corrected figure and its legend appear below. The error does not affect the conclusions of the article. Fig. 2.Comprehensive characterization of the promoter library. Several orthogonal metrics were used to characterize the promoter library and ensure the consistent behavior of all its members for various genes and culturing conditions. We show here three metrics that were chosen for quantifying transcriptional of the promoters: (i) the dynamics of GFP production based on fluorescence, (ii) measurement of the relative mRNA transcript levels in the cultures, and (iii) testing of the MIC for chloramphenicol in an additional library of constructs where the promoter drove the expression of chloramphenicol acetyltransferase. The overall strong correlation between the various metrics suggests a broad-range utility of the promoter library for a variety of genes and conditions. (A) Mouse C127 cells were transduced with retrovirus expressing BPV-1 E7 with a FLAG͞HA epitope tag at either the C terminus (E7-C) or N terminus (E7-N), or with no tag (E7). Cells were lysed, and proteins were immunoprecipitated by using either an anti-FLAG antibody (Left) or an anti-BPV-1 E7 antibody (Right). Proteins were resolved by SDS͞PAGE on a 15% polyacrylamide gel and probed by immunoblotting using the anti-E7 antibody. (B) Cells were assayed for anchorage-independent growth with transduced BPV-1 oncogenes: C127 control cells, cells expressing BPV-1 E7 alone, BPV-1 E6 alone, E6 and E7, E6 and C-terminal FLAG͞HA-tagged E7 (E7-C), and E6 and N-terminal FLAG͞HA-tagged E7 (E7-N). Cells were suspended in 0.3% Noble agar, DMEM, and 10% FBS and grown for 14 days. Representative fields are shown at ϫ10 magnification. For further details, see Cortical analysis related to visual object recognition is traditionally thought to propagate serially along a bottom-up hierarchy of ventral areas. Recent proposals gradually promote the role of top-down processing in recognition, but how such facilitation is triggered remains a puzzle. We tested a specific model, proposing that low spatial frequencies facilitate visual object recognition by initiating top-down processes projected from orbitofrontal to visual cortex. The present study combined magnetoencephalography, which has superior temporal resolution, functional magnetic resonance imaging, and a behavioral task that yields successful recognition with stimulus repetitions. Object recognition elicited differential activity that developed in the left orbitofrontal cortex 50 ms earlier than it did in recognition-related areas in the temporal cortex. This early orbitofrontal activity was directly modulated by the presence of low spatial frequencies in the image. Taken together, the dynamics we revealed provide strong support for the proposal of how top-down facilitation of object recognition is initiated, and our observations a...

show abstract

“…The preliminary processing that was unaltered by attention in our task may reflect a feed-forward phase of perceptual processing that precedes activity that is modulated by a more distributed neural system (2,29,30). Later responses show stronger modulation by effects that suggest top-down control, such as attention, priming, and familiarity (13,15,17,19,20).…”

Section: Discussionmentioning

confidence: 87%

Dissociation of face-selective cortical responses by attention

Furey

Tanskanen²,

Beauchamp³

et al. 2006

Proc. Natl. Acad. Sci. U.S.A.

119

104

View full text Add to dashboard Cite

We studied attentional modulation of cortical processing of faces and houses with functional MRI and magnetoencephalography (MEG). MEG detected an early, transient face-selective response. Directing attention to houses in ''double-exposure'' pictures of superimposed faces and houses strongly suppressed the characteristic, face-selective functional MRI response in the fusiform gyrus. By contrast, attention had no effect on the M170, the early, face-selective response detected with MEG. Late (>190 ms) category-related MEG responses elicited by faces and houses, however, were strongly modulated by attention. These results indicate that hemodynamic and electrophysiological measures of face-selective cortical processing complement each other. The hemodynamic signals reflect primarily late responses that can be modulated by feedback connections. By contrast, the early, face-specific M170 that was not modulated by attention likely reflects a rapid, feedforward phase of face-selective processing.functional MRI ͉ human ͉ magnetoencephalography ͉ visual V iewing faces evokes responses in ventral temporal cortex that have a distinctive spatial topography that can be observed with functional magnetic resonance imaging (1-8) and a distinctive temporal course that can be observed with magnetoencephalography (MEG) (1, 9-13) and electroencephalography (EEG) (10, 14-21). The spatial pattern of hemodynamic cortical response that is measured with fMRI has a maximum in the fusiform face area in the lateral fusiform gyrus, (3,4,7,8,22,23). The temporal electrophysiological response that is measured with MEG (or EEG) contains an early field (or potential) that peaks Ϸ170 ms after the appearance of a face, the M170 (or N170) response (1, 9-21). Both source modeling of MEG signals (9, 10, 12) and EEG recordings made directly from the cortical surface (14) suggest that the M170͞N170 response is generated by activity in ventral and lateral extrastriate regions, including the fusiform gyrus.MEG͞EEG and fMRI complement each other insofar as MEG and EEG signals generated by neural activity can resolve temporal events on a millisecond time scale but with limited spatial resolution, especially for multiple, distributed sources (24), whereas fMRI has relatively high spatial resolution but coarse temporal resolution because it measures slower hemodynamic changes elicited by neural activity (25). The fMRI hemodynamic measures reflect metabolic demand, whereas MEG͞EEG measures reflect electrophysiological activity and are influenced strongly by the synchrony of neuronal activity. Consequently, these measures may be dissociated when the activity is synchronous but brief, resulting in a small metabolic demand, or when it reflects a resetting of the phase of the spontaneous activity with no change in power. The faceselectivity and similar cortical locations of the fusiform face area and M170͞N170 responses suggest that they reflect the same neural activity. Here we show, however, that these responses can be dissociated by the effect of attention. Th...

show abstract

Speed of processing in the human visual system

Cited by 3,080 publications

References 0 publications

Invariant representations of visual patterns in a temporal population code

Invariant representations of visual patterns in a temporal population code

Top-down facilitation of visual recognition

Dissociation of face-selective cortical responses by attention

Contact Info

Product

Resources

About