Processing in the Primary Visual Cortex

Overview: The primary visual cortex constructs local image features

The visual system provides a description of the location and identification of objects that have survival value to the species. This description must be made in a dynamic world in which our gaze is constantly shifting and in which objects move. To be useful, the description must select items of importance depending on context and memory; in essence the visual system is designed to make useful predictions about visual reality, not to actually represent reality. Implicit in the architecture of the visual system reside mechanisms tuned by evolution and development to optimize useful vision. Beginning in the retina, the visual system selects what is needed to accomplish this goal. The retina contributes to this selection process by eliminating information about absolute light intensity, emphasizing local image contours, and compressing the visual signal information into a manageable size during a manageable time window to be transmitted to the lateral geniculate nucleus (LGN, see Chapter 29 ). The LGN contributes by regulating the flow of visual signals. Primary visual cortex (V1), as we will learn, contributes by coding aspects of local image features including their size, orientation, local direction of movement, and binocular disparity. All of these local descriptions of stimulus quality are critical for the more global and complex identification of objects (“what”) and spatial relations (“where”).

Classically, V1 is described as an area where information provided by the separate LGN channels is combined in different ways before it is sent to extrastriate visual areas for further processing. In order to do its job, V1 must solve the geometry puzzle of representing all stimulus qualities necessary for the subsequent steps of analyses within tissue that maps the outside world, i.e. responds to different parts of the visual field. V1 accomplishes this goal by a division of labor between different layers and different iterated modules within each layer. In this classic view, V1 is pictured as the beginning stage for several sets of serial processes that take place in feedforward hierarchies of visual areas; areas specialized to perform a variety of higher-order tasks. Currently, however, there is growing recognition that V1 does more. V1 does not represent stimulus qualities statically. Via dense and intricate horizontal and feedback pathways, V1 dynamically interacts with higher visual and non-visual areas, as well as the thalamus, and alters its activity and message depending on attention, memory, context, and reward. Therefore, for efficiency, V1 cells bias their messages to emphasize moment to moment relevance. In this chapter we initially provide the reader with a road map to the organization of V1, we show how the anatomical architecture of the V1 sets limits on how visual signals can be processed, and we emphasize the laminar architecture, connections, and cell types. Next we review the traditional view of how receptive field properties are put together in V1 and review the hypothesis that the parallel inputs from the LGN relate in functionally specific ways to output channels of V1 (see also Chapter 31 ). The second to last section then briefly reviews more controversial ideas concerning the functions of V1, including whether V1 is conscious and participates in memory or is modified by reward and attention. The final section of this chapter summarizes key points.

Overview of cortical organization: a general road map

Human V1 ranges in size from 1400 to 3400 mm ² , is about 2 mm thick, and is located within the occipital lobe extending from the posterior pole along the medial wall of the hemisphere ( Fig. 30.1 ). Like the rest of the cerebral cortex, V1 contains 6 principal layers. V1 is often called striate cortex in recognition of its original identification by an Italian medical student (Francesco Gennari) more than 200 years ago. Gennari observed that if he sliced into the posterior pole of an unfixed human brain and observed the cortical layers of the slice a striking white stripe was visible, known today as the stria of Gennari . The stria of Gennari actually marks one heavily myelinated layer in the middle of the gray matter of V1. V1 also is referred to by several other names including, most commonly, area 17 of Brodmann or primary visual cortex. For convenience we will use the term V1 in this chapter.

Damage to V1 results in a hole (scotoma) or blind spot in one visual hemifield. The location of damage to V1 can be predicted based upon the topographic map of the opposite visual hemifield that is known to exist in this region in all mammalian species. In humans, detailed knowledge about the manner in which the visual field is mapped onto V1 has been obtained from a variety of sources including clinical assessments of damage, results of electrical stimulation, and transcranial magnetic stimulation (TMS). The latter is a non-invasive method used to excite neurons by using rapidly changing magnetic fields to produce weak electrical currents. Both direct invasive electrical stimulation and non-invasive TMS can give rise to the experience of localized phosphenes (flashes of light) experienced by the subject. Other methods of developing functional maps of V1 have used positron emission tomography (PET) and, most commonly, functional magnetic resonance imaging (fMRI). In the latter case the subject is presented with topographically restricted stimuli in different locations in space and the relative activation of V1 measured as a change in blood flow. Knowledge about the retinotopic organization of V1 in humans is relevant not only to the clinical evaluation of damage, but also to research efforts designed to develop visual prosthetic devices involving visual cortical electrical stimulation for individuals with incurable retinal diseases or loss of eyes ( Box 30.1 ). All of the methods used to understand the map of the visual field in V1 are in general agreement showing, as illustrated in Figure 30.1 , that the fovea is represented in the occipital pole and the far periphery is represented in the anterior margin of the calcarine fissure with the upper and lower visual fields being mapped onto its lower (lingual gyrus) and upper (cuneus gyrus) banks, respectively. As in the LGN, the visual field map in human cortex is distorted such that the representation of central vision occupies much more tissue than does peripheral vision. Whether the foveal representation in V1 is expanded over what would be predicted simply by assigning each retinal ganglion cell or LGN cell the same amount of cortical tissue has been the subject of considerable debate. One explanation for this controversy is that the proportion of cortex devoted to central vision has been shown to be highly individually variable, at least in macaque monkeys. The latter finding indicates that the relative amount of tissue devoted to the fovea could be magnified at the cortical level relative to the retina in some individuals but not others.

Layers, connections, and cells of V1: The inputs, outputs, and general wiring

As in other cortical areas, V1 has six main layers that can be identified in a cell stain as shown in Figure 30.2 . These layers contain two main types of cells, pyramidal and stellate, in an 80/20 proportion. The layers and sublayers of V1 have been named in different ways depending on investigator interpretation. The most common laminar scheme is the one adopted by Brodmann. Although Brodmann likely made a major mistake in his interpretation of V1 layers in all primates his scheme is the most accepted so we use it here ( Fig. 30.2 ). As in other areas of neocortex, input from the thalamus ends primarily in the middle layers (IVC of V1), with the superficial layers (above IVC) sending information to other cortical areas and the layers below IVC sending signals to subcortical targets ( Fig. 30.2 ).

LGN inputs

Studies done in anesthetized monkeys have demonstrated that activation of V1 neurons depends completely on input from the LGN, because when the LGN is inactivated, visually evoked potentials in V1 are blocked. LGN axons carrying signals from the left and right eyes and from K, M, and P layers remain segregated at the first synapse in V1 in non-human primates and presumably also humans ( Fig. 30.3 ). The left and right eye segregation of axons from the M and P LGN layers ending in layer IVC of V1 is referred to as ocular dominance columns. These “columns” have been directly demonstrated to exist in humans as shown in the complete reconstruction of the pattern of left and right eye input to V1 in Figure 30.4 . These zebra-stripe-like patterns of monocular segregation from LGN can be revealed in tangential sections of layer IVC of V1 in humans postmortem following long-standing loss of one eye. As shown in Figure 30.4 , the patterns are revealed by a histochemical stain for the mitochondrial enzyme, cytochrome oxidase (CO) which becomes down-regulated in the zones of layer IVC innervated by LGN axons from the missing eye. As mentioned and shown schematically in Figure 30.3 , K, M, and P LGN axons in macaque monkeys also terminate in separate V1 layers or sublayers. M and P axons synapse in sublayers of IVC, and K LGN axons synapse in layers IVA, III, and I. Also, K LGN axons are likely made up of several classes since those that end in different V1 layers appear to come from different populations of LGN cells in macaques. It is noteworthy, however, that the LGN does not project to layer IVA in apes and humans, reminding us that care must be taken in translating data directly from other primates to humans.

The fact that input from the left and right eyes remains segregated at the first synapse in V1 raises an interesting question concerning the retinotopic map. Recall that in the LGN each layer contains a continuous map of the opposite visual hemifield. This means that in cortex there must be two topographic maps, one for each eye, within the same layer, at least at the first synapse from the LGN. This is exactly what was found in layer IV of the macaque monkey using detailed electrophysiological recordings. Tangential recordings made within layer IVC revealed that the segments of the visual field mapped in one ocular dominance column for one eye were also represented in the adjacent column that received input from the other eye. As we will see below, information from the two eyes and from K, M, and P cells is combined in different ways within other layers of V1, hence, most cells in V1 receive a combination of many extrinsic and intrinsic inputs.