• No se han encontrado resultados

CAPÍTULO IV: ASPECTOS ADMINISTRATIVOS

4.2. ANÁLISIS DE SENSIBILIDAD

4.2.4. Análisis del ROI

Key Notes

The sorting of fibers in the optic chiasm means that the left LGN maps the right side of the visual field. The primate LGN has six layers, two magnocellular layers get input from M ganglion cells and contain movement-sensitive cells, four parvocellular layers are innervated by P ganglion cells, and cells in these layers are wavelength selective. Each layer gets a precise retinotopic input from just one eye. The properties of LGN cells are similar to those of the ganglion cells which supply them and have circular receptive fields with surround inhibition. The LGN projects to the primary visual cortex and receives extensive connections from it which may be involved in visual attention.

The striate cortex in the occipital lobe is the primary visual cortex (V1) and gets a retinotopic projection from the LGN, with the fovea occupying a disproportionately big area. The M and P LGN cells project to different sublayers in layer 4C of the cortex, giving rise to distinct streams of information flow through the cortex. Most cells in V1 have elongated receptive fields and respond to linear features rather than spots. Simple cells are position sensitive. Complex cells are less sensitive to position than simple cells, many having a preference for linear stimuli moving at right angles to the long axis of their receptive fields.

All cells lying within radial columns extending through V1 respond to linear features having approximately the same orientation. All orientations are represented for each point on the retina. These

orientation columns are ordered so that a smooth gradient for orientation exists; columns with the same orientation are aligned in stripes across the cortex.

Many cells in V1 get input from both eyes, but most show ocular dominance in that they are preferentially driven by one eye. These cells occur in discrete ocular dominance columns that are aligned in stripes across the cortex in which ipsilateral and contralateral eye dominance alternates. Binocular cells get input from corresponding positions on the two retinas and measure retinal disparity, from which the visual system computes the depth of an object in three-dimensional space.

The volume of cortex in which every orientation is mapped for corresponding positions on both retinas is called a hypercolumn. It consists of a complete set of orientation columns and ocular dominance columns for a single pixel of the visual field.

Related topics Organization of the central Eye and visual pathways (G2)

nervous system (A5) Parallel processing in the visual

Attributes of vision (G1) system (G7)

Lateral geniculate nucleus Primary visual cortex Orientation columns Binocular cells Hypercolumns

Pathways for visual perception start with the retinogeniculate fibers, axons of ganglion cells that end in the lateral geniculate nucleus (LGN). Because of the manner in which fibers are sorted in the optic chiasm, the left optic tract and left LGN carry axons from the left side of both retinas. Thus the left LGN represents the right side of the visual field (see Fig. 2, Topic G2).

The primate LGN has six layers (Fig. 1). The two most ventral are the magno- cellular(large-cell) layers which receive input from M ganglion cells. Dorsal to these are the four parvocellular (small-cell) layers that are innervated by P ganglion cells. Interleaved between these major layers are koniocellular layers containing very small cells. These are thought to receive input from small slowly conducting retinal ganglion cells with large dendritic trees and large receptive fields which signal average illumination.

LGN cells have circular RFs with surround antagonism. They show little or no response to diffuse light covering the whole receptive field. Each layer in the LGN gets input from only one eye and no cells show binocular responses (responses to both eyes). It is not until the visual cortex is reached that input from both eyes is integrated. The responses of the LGN cells match those of the ganglion cells which supply them, so on and off channels remain independent and P cells display precisely the same color opponency properties as retinal ganglion cells. There is a very precise topographic (retinotopic) mapping from the retina onto each of the layers of the LGN, with the representation of the fovea taking up about half of the nucleus. The maps in each layer are in precise register with each other so that any given vertical axis through the LGN passes through cells with RFs representing the same place in visual space.

The LGN contains two populations of neurons. Those that project to the primary visual cortex are geniculostriate neurons. In addition, there is a substantial population of smaller interneurons the exact function of which is not known. Furthermore only about 20% of the synaptic connections on geniculo- striate neurons are from retinal ganglion cells. Other synapses are made by back projections from the visual cortex and by the reticular formation. They may play a role in visual attention, modifying geniculostriate neuron responses so that only a selected proportion of retinal input is transmitted through to the visual cortex.

The fibers of the optic tract terminate in the striate cortex (Brodmann’s area 17) on the medial surface of the tip of the occipital lobe. This region is the primary visual cortex (V1). Precise retinotopic mapping is maintained up to V1 with the fovea having a disproportionate representation.

Primary visual cortex

Lateral geniculate nucleus

G6 – Early visual processing 175

Magnocellular layers Contralateral eye Parvocellular layers Ipsilateral eye 6 6 5 5 4 4 3 3 2 2 1 1

There are at least three parallel streams of information into the primary visual cortex. The movement-sensitive M LGN cells input into layer 4Cα, the P LGN cells go to layer 4Cβ, whereas the koniocellular layers of the LGN project to layers 2 and 3. These streams remain quasi-independent throughout the visual system. The connections of the primary visual cortex are illustrated for the primate in Fig. 2.

The great majority of cells in V1 have elongated receptive fields (RFs) with both inhibitory and excitatory regions, and respond to bars, slits, edges and corners rather than spots of light. Most fall into two categories based on their RF properties, simple or complex cells. Both are orientation selective, in that they respond to linear features in only a narrow range of orientations.

1. Simple cells are pyramidal cells found mostly in layers 4 and 6. They are highly sensitive to the position of a stimulus on the retina. They have small oval RFs with center–surround antagonism (Fig. 3). A simple cell gets its input from a linear array of LGN cells having the same RF properties, so the RF of the simple cell emerges as a consequence of the RFs of the LGN inputs. 2 Complex cells are most abundant in layers 2, 3 and 5. They have larger RFs

than simple cells and, lacking distinct inhibitory or excitatory regions, a stim- ulus of the appropriate orientation anywhere in the RF evokes a response. Hence, complex cells are much less fussy about position than simple cells. Many complex cells show a preference for movement at right angles to the long axis of the stimulus. Some complex cells receive their inputs from simple cells but others get their input directly from the LGN.

1 2 3 4A 4B 4Cα 4Cβ 5 6 Pial surface Spiny stellate excitatory interneuron Pyramidal cell Aspiny inhibitory interneuron LGN

K M P To ipsilateral andcontralateral extrastriate cortex To pulvinar, superior colliculus and pons To LGN

Fig. 2. Canonical circuitry in the primary visual cortex illustrated for parvocellular (P) LGN input. Magnocellular (M) circuitry (not illustrated) has its input to 4Cα. Spiny stellate cells here send axons to pyramidal cells in 4B and these send collaterals to pyramidal cells in layers 5 and 6 directly rather than via more superficial layers. Koniocellular input is directly to pyramidal cells in blobs of 2 + 3. Feedback collaterals are dotted. Cortical layers are designated by Arabic numbers.

In common with other sensory cortex, the primary visual cortex is divided into radial columns 30–100 µm across. In each of these all cells respond preferentially to linear features with a given orientation so they are called orientation columns. The cortex is organized so that adjacent columns have an orientation preference that differs by only about 15°; in other words, orientation is repre- sented in a systematic way across the cortex. Columns which have the same orientation are arranged in stripes across the cortex. The obvious inference, that orientation selectivity is how the visual system represents straight-line segments which can be built up to give the form of an image, need not be true. Computer modeling shows that orientation selectivity is a property of neural networks that learn the curvature of curved surfaces from their shading. Hence orientation selectivity might, counter-intuitively, be concerned with representations of curves rather than linear features in the visual world.

Binocular cells V1 is the first region in which input from both eyes is combined. Many cells,

particularly in layers 4B, 2 and 3 show binocular responses in that they can be driven by either eye. This is a necessary condition for stereopsis. Most binocular cellsshow a preference for one eye, a phenomenon referred to as ocular domi- nance. Cells which have the same ocular dominance (e.g. those that are driven preferentially by the ipsilateral eye) occupy ocular dominance columns that are situated in long stripes about 500 µm across. Columns representing ipsilateral and contralateral input alternate regularly over the cortex which when visual- ized at the level of layer 4C look like the pattern of stripes on a zebra.

The receptive fields of binocularly driven cells resemble those of simple or complex cells, lie in corresponding positions in the two retinas, have identical orientation properties and have similar arrangements of excitatory and inhibitory regions. Similar input from both eyes into arrays of binocular cells is needed for perception of a fused image. To the extent that inputs into these cells are unequal they measure retinal disparity and so the depth of an object in three-dimensional visual space. Cells that respond to visual disparity have been discovered in the primate visual cortex (including V1). These are responsible for stereopsis.

Hypercolumns A higher-order modularity exists in the primary visual cortex. Called a hyper-

column (Fig. 4), it represents a given corresponding position for both retinas, and maps every orientation for that position. It consists of a full-thickness slab Orientation

columns

G6 – Early visual processing 177

(b) LGN cell inputs Layer 4 stellate cells Simple cell Simple cell RF RFs of LGN cells

Fig. 3. (a) Receptive fields of three simple cells; (b) a diagram depicting how lateral geniculate nucleus (LGN) cells contribute to the simple cell receptive field (RF), four on-center LGN cell RFs generate an on-centre simple cell RF.

of cortex with an area of about 1 mm2

containing a complete set of orientation columns for both ipsilateral and contralateral ocular dominance. The retinotopic map in V1 occurs because adjacent pixels of the retina map to adjacent columns in an orderly fashion. Complex cells Complex cells Simple cells Interstripe interval (V2) Thin stripe (V2) Cortical blob (color) Upper surface of a single hypercolumn (area ~ 1 mm2) 2 mm Ocular dominance column Orientation column P channel

(form, color) M channel

(movement, visual attention) LGN 6 5 4 3 2 1 Parvocellular (P) layers Magnocellular (M) layers 500µm30–100µm I I C C I II III IVA IVB IVC-α IVC-β V VI

Fig. 4. Modular structure of the primary visual cortex. Cortical layers are designated by Roman numerals. I, ipsilateral; C, contralateral (blobs are described in Topic G7).

Section G – Vision

G7

PARALLEL PROCESSING IN

Documento similar