The Mechanism For Processing Motion And Orientation Signals In Early Visual Cortices

Human sensation of the external world mainly comes from the visual system. Based on the specific property, visual information in the natural environment includes motion signal, form signal, color signal, depth signal, etc. The processing of these elementary signals by the visual system leads to the perception of real and apparent motion, object and face recognition, figure and ground segregation, as well as all other high-level visual functions. After decades of research, we have understood the basic architecture and function of the visual system. In the early cortical stage of the visual system (such as V1and V2), neurons with similar orientation, direction, and spatial frequency preferences tend to cluster together to form columns. However, it is still unclear how these neurons contribute to the cue-invariant motion and orientation perception, and how these neurons overcome the aperture problem due to their small receptive field sizes. Moreover, motion information includes three different components that are direction, speed, and motion axis (spatial orientation information related with motion trajectory). Whether these three components are segregated or combined in the visual pathway is unknown. Lastly, in the natural environment, except from contours defined by luminance changes (first-order contours), there exist contours defined by visual cues other than luminance (second-order contours), such as contrast, motion, and texture. How neurons process the orientation of second-order contours needs to be elucidated.To answer the above questions, we used intrinsic signal optical imaging, cytochrome oxidase (CO) staining, moving random-noise and two second-order contour stimuli, and spatio-temporal energy model simulation to study early visual motion and orientation signal processing in V1, V2, and V4of macaque ventral visual pathway at the column level. To further investigate whether different types of neuronal populations use different mechanisms in motion processing, we used intrinsic signal optical imaging and in vivo single-unit recording to examine the response properties of different types of neurons to the motion of random-dot stimuli in cat visual areas17and18. We also systematically simulated all the responses recorded using the spatio-temporal energy model. We found:1) In V1, V2, and V4of macaque ventral visual pathway, the classical orientation columns that in charge of processing orientation information also served to process the motion-axis signals generated by moving objects. The moving speed of the noise stimuli could affect the preferred motion axes of these orientation columns. Specifically, at low speed, the preferred motion axes of these orientation columns were perpendicular to their preferred orientations, while at high speed, the preferred motion axes became parallel to the orientations they encoded.2) Direction-selective population responses were only observed in V2; these direction encoding domains were precisely in register with the CO stained thick stripes. Whereas in area V2, color encoding domains located in the thin stripes and domains that encoded orientation/motion axis corresponded to the pale and thick stripes.3) Spatio-temporal energy model was able to simulate the speed dependent population responses to the motion-axis signals of noise stimuli in the orientation columns of macaque V1and V2.4) In cat areas17and18, the preferred motion axes of orientation columns to random-dot stimuli also critically depended on the moving speed. The transition speeds, at which orientation columns turned to encode the motion axes that were parallel to their preferred grating orientations, were about31°/s and15°/s in areas17and18, respectively.5) In cat areas17and18, direction-selective neurons only encoded their preferred directions at low speed, while at high speed, they switched to encode the motion axes perpendicular to their preferred directions. In addition, there was almost no difference in response property to random-dot stimuli moving at both low and high speeds between simple and complex cells.6) At both single-cell and population levels, spatio-temporal energy model simulated not only the responses of orientation-selective cells, but also the responses of direction-selective cells to random dots moving at different speeds.7) Orientation columns in macaque V1and V2could invariantly represent the orientation of contrast-modulated and phase-reversed contours. However, population responses to these second-order contours were significantly weaker than those to grating stimuli (first order). Spatio-temporal energy model could also simulate the orientation-cue invariant population responses to second-order contour stimuli in macaque V1and V2.Our study indicates that orientation-selective neurons in the early visual cortices can help to process the motion of objects. Furthermore, both orientation-and direction-selective neurons contribute to the motion streak or speed line phenomenon observed in human psychophysics studies. We also confirmed that direction-selective cells cluster in the thick stripes of V2, which supports the hypothesis that visual signals are parallel and separately processed in the primate visual system. In addition, our results suggest that the cortical representation of the orientation of some second-order contours starts in V1, which will facilitate future studies on the mechanisms underlying second-order signal processing. Finally, all our results from spatio-temporal energy model simulations strongly indicate that the linear filtering properties of neurons in the early visual cortices underlie all the responses recorded and thus suggest that higher level and more complex models should regard V1neurons as spatio-temporal energy detectors. All our optical imaging results support that the orientation maps in the early visual cortices of both cat and monkey can be described as spatio-temporal energy maps.