FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W4281296799> ?p ?o ?g. }

• W4281296799 endingPage "R476" @default.
• W4281296799 startingPage "R474" @default.
• W4281296799 abstract "We can distinguish between the direction and speed of a moving object effortlessly, but this is actually a very challenging computational task. A new study demonstrates that this process begins at the first stages of visual processing in the retina. We can distinguish between the direction and speed of a moving object effortlessly, but this is actually a very challenging computational task. A new study demonstrates that this process begins at the first stages of visual processing in the retina. Anyone who has ever driven a car can attest to the importance of accurately judging the trajectory of the other cars on the road. This need to precisely estimate the direction that other cars are moving remains important whether one is racing along the Autobahn or stuck behind the Sunday joyrider traveling well below the speed limit. Confounding the speed and direction of other cars on the road could have disastrous consequences. Indeed, humans can judge the direction of moving objects with similar accuracy across a wide (∼100-fold) range of stimulus speeds1De Bruyn B. Orban G.A. Human velocity and direction discrimination measured with random dot patterns.Vision Res. 1988; 28: 1323-1335Crossref PubMed Scopus (229) Google Scholar. This ability to represent direction information separately from other pieces of information is a type of invariant neural representation, and the brain contains several types of these invariant representations that are critical to behavior and survival2Albright T.D. Form-cue invariant motion processing in primate visual cortex.Science. 1992; 255: 1141-1143Crossref PubMed Scopus (257) Google Scholar,3Rust N.C. Dicarlo J.J. Selectivity and tolerance (“invariance”) both increase as visual information propagates from cortical area V4 to IT.J. Neurosci. 2010; 30: 12978-12995Crossref PubMed Scopus (219) Google Scholar. A new study by Summers and Feller reported in this issue of Current Biology demonstrates that invariant representations of motion direction begin at the earliest stages of visual processing in the retina4Summers M.T. Feller M.B. Distinct inhibitory pathways control velocity and directional tuning in the mouse retina.Curr. Biol. 2022; 32: 2130-2143Abstract Full Text Full Text PDF PubMed Scopus (3) Google Scholar. A great deal of research has been devoted to understanding how invariant representations emerge as sensory information flows through biological and artificial neural networks3Rust N.C. Dicarlo J.J. Selectivity and tolerance (“invariance”) both increase as visual information propagates from cortical area V4 to IT.J. Neurosci. 2010; 30: 12978-12995Crossref PubMed Scopus (219) Google Scholar,5Hubel D.H. Wiesel T.N. Receptive fields, binocular interaction and functional architecture in the cat’s visual cortex.J. Physiol. 1962; 160: 106-154Crossref PubMed Scopus (8713) Google Scholar, 6Bengio Y. Learning deep architectures for AI.Foundations Trends® Machine Learning. 2009; 2: 1-127Crossref Scopus (6112) Google Scholar, 7Serre T. Kreiman G. Kouh M. Cadieu C. Knoblich U. Poggio T. A quantitative theory of immediate visual recognition.Prog. Brain Res. 2007; 165: 33-56Crossref PubMed Scopus (169) Google Scholar, 8Riesenhuber M. Poggio T. Hierarchical models of object recognition in cortex.Nat. Neurosci. 1999; 2: 1019-1025Crossref PubMed Scopus (2403) Google Scholar, 9Rust N.C. Stocker A.A. Ambiguity and invariance: two fundamental challenges for visual processing.Curr. Opin. Neurobiol. 2010; 20: 382-388Crossref PubMed Scopus (42) Google Scholar. According to the standard model, invariant and other complex representations gradually emerge as information is processed by successive layers of hierarchically organized neural circuits. Thus, they are thought to first arise in the cortex after passing through many levels of neural processing5Hubel D.H. Wiesel T.N. Receptive fields, binocular interaction and functional architecture in the cat’s visual cortex.J. Physiol. 1962; 160: 106-154Crossref PubMed Scopus (8713) Google Scholar,8Riesenhuber M. Poggio T. Hierarchical models of object recognition in cortex.Nat. Neurosci. 1999; 2: 1019-1025Crossref PubMed Scopus (2403) Google Scholar (Figure 1A). Moreover, deep artificial neural networks with such hierarchical organization and many layers have successfully learned invariant features from sensory input, which has been interpreted as evidence that deep networks are needed for complex feature extraction6Bengio Y. Learning deep architectures for AI.Foundations Trends® Machine Learning. 2009; 2: 1-127Crossref Scopus (6112) Google Scholar,7Serre T. Kreiman G. Kouh M. Cadieu C. Knoblich U. Poggio T. A quantitative theory of immediate visual recognition.Prog. Brain Res. 2007; 165: 33-56Crossref PubMed Scopus (169) Google Scholar. These representations are not expected to emerge from shallow neural networks consisting of only a few layers. The vertebrate retina constitutes a clear exception to this orderly model of hierarchical neural processing. Utilizing only two synaptic layers, retinal circuits extract complex features from the raw light inputs, including object orientation, object/background motion, and motion direction10Gollisch T. Meister M. Eye smarter than scientists believed: neural computations in circuits of the retina.Neuron. 2010; 65: 150-164Abstract Full Text Full Text PDF PubMed Scopus (444) Google Scholar. Further, a class of these direction-encoding neurons, direction-selective ganglion cells, shows evidence for invariant neural encoding — reliably discriminating motion direction across a broad range of stimulus contrasts, sizes, and speeds11Grzywacz N.M. Amthor F.R. Robust directional computation in on-off directionally selective ganglion cells of rabbit retina.Vis. Neurosci. 2007; 24: 647-661Crossref PubMed Scopus (24) Google Scholar. This indicates that complex feature extraction including invariant stimulus selectivity can emerge from shallow neural networks. However, current theoretical models of motion detection cannot account for this invariant motion selectivity12Hassenstein B. Reichardt W. Systemtheoretische Analyse der Zeit-, Reihenfolgen-und Vorzeichenauswertung bei der Bewegungsperzeption des Rüsselkäfers Chlorosphanus.Z. Naturforsch. B. 1956; 11: 513-524Crossref Scopus (610) Google Scholar, 13Barlow H.B. Levick W.R. The mechanism of directionally selective units in rabbit’s retina.J. Physiol. 1965; 178: 477-504Crossref PubMed Scopus (1007) Google Scholar, 14Adelson E.H. Bergen J.R. Spatiotemporal energy models for the perception of motion.J. Opt. Soc. Am. A. 1985; 2: 284-299Crossref PubMed Scopus (2613) Google Scholar. The three widely accepted models for direction selectivity predict a bandpass relationship between speed and direction selectivity. The details vary between these models, but they all share a common feature — they detect motion by computing the correlation between flanking receptive-field regions with fast and slow (that is, delayed) kinetics12Hassenstein B. Reichardt W. Systemtheoretische Analyse der Zeit-, Reihenfolgen-und Vorzeichenauswertung bei der Bewegungsperzeption des Rüsselkäfers Chlorosphanus.Z. Naturforsch. B. 1956; 11: 513-524Crossref Scopus (610) Google Scholar, 13Barlow H.B. Levick W.R. The mechanism of directionally selective units in rabbit’s retina.J. Physiol. 1965; 178: 477-504Crossref PubMed Scopus (1007) Google Scholar, 14Adelson E.H. Bergen J.R. Spatiotemporal energy models for the perception of motion.J. Opt. Soc. Am. A. 1985; 2: 284-299Crossref PubMed Scopus (2613) Google Scholar. For example, the Barlow-Levick paradigm, which models retinal direction selectivity, features a fast excitatory component on one side of the receptive field and delayed inhibition on the other side (Figure 1B). This circuit motif produces a receptive field that is slanted in space–time, which biases the cell’s responses to objects moving in the same direction as the slant (Figure 1C). This bias occurs because of the differences in kinetics in the inhibitory and excitatory circuitry. As an object moves first through the side of the receptive field with delayed inhibition (termed the null direction), it engages the inhibitory and excitatory components at about the same time, causing the inhibition to veto the excitatory component and produce a weak response. When the object moves in the opposite direction (termed the preferred direction), the delayed inhibition occurs later than the excitation and the cell responds vigorously. This receptive-field motif predicts that a direction-selective cell will distinguish motion direction best when the object is moving at a speed that corresponds to the slope of the space-time receptive field, and direction discrimination will deteriorate at faster and slower speeds (Figure 1C). Thus, according to this model, direction selectivity will show a bandpass speed tuning (Figure 1D). Indeed, one type of ganglion cell shows such a bandpass speed tuning with a bias for slowly moving objects. This cell responds to the onset of a light flash over its receptive field and is called the On direction-selective ganglion cell. However, another cell class that responds at both the onset and offset of a light flash — On-Off direction-selective ganglion cells — shows relatively flat speed tuning across a wide (∼60-fold) range of speeds. This speed invariance runs contrary to the prediction of the Barlow-Levick model11Grzywacz N.M. Amthor F.R. Robust directional computation in on-off directionally selective ganglion cells of rabbit retina.Vis. Neurosci. 2007; 24: 647-661Crossref PubMed Scopus (24) Google Scholar. The new study by Summers and Feller demonstrates that these differences arise in the circuits that provide synaptic inhibition to these cell types4Summers M.T. Feller M.B. Distinct inhibitory pathways control velocity and directional tuning in the mouse retina.Curr. Biol. 2022; 32: 2130-2143Abstract Full Text Full Text PDF PubMed Scopus (3) Google Scholar. On direction-selective cells receive input from an amacrine cell that provides inhibition via synaptic glycine release. This inhibitory input does not vary in strength based on the direction that an object is moving, but it does vary with motion speed. Inhibition is strong at faster motion speeds and is weak at slow speeds; thus, these inhibitory neurons suppress the direction-selective cell’s spike responses at fast speeds, which explains the lack of direction selectivity at those speeds. In contrast, On-Off direction-selective cells show inhibitory inputs that are largely independent of motion speed. Summers and Feller demonstrate that this speed-invariant inhibition arises in the starburst amacrine cells, the principal substrate for direction selectivity in the vertebrate retina4Summers M.T. Feller M.B. Distinct inhibitory pathways control velocity and directional tuning in the mouse retina.Curr. Biol. 2022; 32: 2130-2143Abstract Full Text Full Text PDF PubMed Scopus (3) Google Scholar. These findings show that the retina extracts a highly refined representation of visual motion within three synapses of collecting the raw light signals in the photoreceptors. Thus, deep neural networks are not required to produce complex and invariant representations of some sensory information. These results have implications for understanding the properties of biological neural networks and for developing smarter artificial networks. The Barlow-Levick model was developed to explain the properties of On-Off direction-selective cells13Barlow H.B. Levick W.R. The mechanism of directionally selective units in rabbit’s retina.J. Physiol. 1965; 178: 477-504Crossref PubMed Scopus (1007) Google Scholar. However, this model does not account for the invariant properties of these cells, indicating that more nuanced computations are occurring in the starburst cell circuit. Discrepancies between receptive field predictions and cellular response properties have also been observed in cortical direction-selective cells15Conway B.R. Livingstone M.S. Space-time maps and two-bar interactions of different classes of direction-selective cells in macaque V-1.J. Neurophysiol. 2003; 89: 2726-2742Crossref PubMed Scopus (51) Google Scholar. Many of these cells show canonical direction selectivity to moving objects, but their measured receptive fields lack the space-time slant predicted by the Adelson-Bergen model, the standard model of direction-selectivity in primates14Adelson E.H. Bergen J.R. Spatiotemporal energy models for the perception of motion.J. Opt. Soc. Am. A. 1985; 2: 284-299Crossref PubMed Scopus (2613) Google Scholar,15Conway B.R. Livingstone M.S. Space-time maps and two-bar interactions of different classes of direction-selective cells in macaque V-1.J. Neurophysiol. 2003; 89: 2726-2742Crossref PubMed Scopus (51) Google Scholar. These discrepancies between model predictions and neural behavior indicate that we still have much to learn about how these circuits process visual motion, including where and how motion information is extracted from visual inputs4Summers M.T. Feller M.B. Distinct inhibitory pathways control velocity and directional tuning in the mouse retina.Curr. Biol. 2022; 32: 2130-2143Abstract Full Text Full Text PDF PubMed Scopus (3) Google Scholar,16Hillier D. Fiscella M. Drinnenberg A. Trenholm S. Rompani S.B. Raics Z. Katona G. Juettner J. Hierlemann A. Rozsa B. et al.Causal evidence for retina-dependent and -independent visual motion computations in mouse cortex.Nat. Neurosci. 2017; 20: 960-968Crossref PubMed Scopus (63) Google Scholar,17Liu B. Hong A. Rieke F. Manookin M.B. Predictive encoding of motion begins in the primate retina.Nat. Neurosci. 2021; 24: 1280-1291Crossref PubMed Scopus (11) Google Scholar. The groundbreaking experiments and model building of Barlow, Levick, and others were critical steps in understanding how neural circuits detect visual motion13Barlow H.B. Levick W.R. The mechanism of directionally selective units in rabbit’s retina.J. Physiol. 1965; 178: 477-504Crossref PubMed Scopus (1007) Google Scholar. In the nearly six decades since the Barlow-Levick model was first proposed, Dr. Feller and other key figures in the field have provided a deeper understanding of how direction selectivity arises in the retina18Morrie R.D. Feller M.B. An asymmetric increase in inhibitory synapse number underlies the development of a direction selective circuit in the retina.J. Neurosci. 2015; 35: 9281-9286Crossref PubMed Scopus (23) Google Scholar, 19Vlasits A.L. Morrie R.D. Tran-Van-Minh A. Bleckert A. Gainer C.F. DiGregorio D.A. Feller M.B. A role for synaptic input distribution in a dendritic computation of motion direction in the retina.Neuron. 2016; 89: 1317-1330Abstract Full Text Full Text PDF PubMed Scopus (58) Google Scholar, 20Mauss A.S. Vlasits A. Borst A. Feller M. Visual circuits for direction selectivity.Annu. Rev. Neurosci. 2017; 40: 211-230Crossref PubMed Scopus (95) Google Scholar. Future work that integrates this more sophisticated understanding into computational models will provide insights into how starburst cells and other shallow neural networks can produce complex and invariant representations of sensory information. The author declares no competing interests. Distinct inhibitory pathways control velocity and directional tuning in the mouse retinaSummers et al.Current BiologyApril 7, 2022In BriefFeature detectors in the retina construct visual representations with limited neural resources. Summers and Feller show that ON and ON-OFF DSGCs, which have distinct velocity preferences, share an invariant mechanism of directional tuning via starburst amacrine cell inhibition, while an independent glycinergic amacrine cell shapes speed tuning. Full-Text PDF Open Access" @default.
• W4281296799 created "2022-05-24" @default.
• W4281296799 creator A5051436273 @default.
• W4281296799 date "2022-05-01" @default.
• W4281296799 modified "2023-09-30" @default.
• W4281296799 title "Neuroscience: Reliable and refined motion computations in the retina" @default.
• W4281296799 cites W1593057674 @default.
• W4281296799 cites W1990037101 @default.
• W4281296799 cites W2018837082 @default.
• W4281296799 cites W2062074945 @default.
• W4281296799 cites W2076519224 @default.
• W4281296799 cites W2085921385 @default.
• W4281296799 cites W2099616257 @default.
• W4281296799 cites W2108992228 @default.
• W4281296799 cites W2111392332 @default.
• W4281296799 cites W2112719643 @default.
• W4281296799 cites W2116360511 @default.
• W4281296799 cites W2149194912 @default.
• W4281296799 cites W2165838718 @default.
• W4281296799 cites W2169259454 @default.
• W4281296799 cites W2299153360 @default.
• W4281296799 cites W2516546718 @default.
• W4281296799 cites W2616960090 @default.
• W4281296799 cites W3191281378 @default.
• W4281296799 cites W4223438179 @default.
• W4281296799 cites W4231109964 @default.
• W4281296799 doi "https://doi.org/10.1016/j.cub.2022.04.037" @default.
• W4281296799 hasPubMedId "https://pubmed.ncbi.nlm.nih.gov/35609547" @default.
• W4281296799 hasPublicationYear "2022" @default.
• W4281296799 type Work @default.
• W4281296799 citedByCount "1" @default.
• W4281296799 countsByYear W42812967992023 @default.
• W4281296799 crossrefType "journal-article" @default.
• W4281296799 hasAuthorship W4281296799A5051436273 @default.
• W4281296799 hasBestOaLocation W42812967991 @default.
• W4281296799 hasConcept C104114177 @default.
• W4281296799 hasConcept C169760540 @default.
• W4281296799 hasConcept C2777093970 @default.
• W4281296799 hasConcept C31972630 @default.
• W4281296799 hasConcept C41008148 @default.
• W4281296799 hasConcept C86803240 @default.
• W4281296799 hasConceptScore W4281296799C104114177 @default.
• W4281296799 hasConceptScore W4281296799C169760540 @default.
• W4281296799 hasConceptScore W4281296799C2777093970 @default.
• W4281296799 hasConceptScore W4281296799C31972630 @default.
• W4281296799 hasConceptScore W4281296799C41008148 @default.
• W4281296799 hasConceptScore W4281296799C86803240 @default.
• W4281296799 hasFunder F4320337350 @default.
• W4281296799 hasIssue "10" @default.
• W4281296799 hasLocation W42812967991 @default.
• W4281296799 hasLocation W42812967992 @default.
• W4281296799 hasOpenAccess W4281296799 @default.
• W4281296799 hasPrimaryLocation W42812967991 @default.
• W4281296799 hasRelatedWork W1480401973 @default.
• W4281296799 hasRelatedWork W1898957235 @default.
• W4281296799 hasRelatedWork W2044601454 @default.
• W4281296799 hasRelatedWork W2063324237 @default.
• W4281296799 hasRelatedWork W2089216837 @default.
• W4281296799 hasRelatedWork W2319305617 @default.
• W4281296799 hasRelatedWork W2478250964 @default.
• W4281296799 hasRelatedWork W2602150136 @default.
• W4281296799 hasRelatedWork W4245071553 @default.
• W4281296799 hasRelatedWork W4377023576 @default.
• W4281296799 hasVolume "32" @default.
• W4281296799 isParatext "false" @default.
• W4281296799 isRetracted "false" @default.
• W4281296799 workType "article" @default.