Lecture Notes - 6/29/99 and 7/1/99
Alex Huk
Visual Pathways

Visual Pathways: Lecture Notes

In these lectures we will follow the flow of information about visual world, picking up at the photoreceptors and finishing in the cortex. We will learn many important facts about the structure and function of the parts of the brain dedicated to vision. At several points along the way ("Interludes"), we will pause to see how these details fit into more general, "big picture" themes in neuroscience and psychology.

Remember that the retina is a part of the brain, "reaching out" to see the world. It is the only part of the brain that can be viewed directly (see "Light and Eyes" lecture).
A quick overview of the route: the outputs of the retina then pass through the optic nerve, cross and split at the optic chiasm, through the optic tract to the LGN. From there, they pass to V1, or primary visual cortex, and then on to "higher-level" brain areas. Get oriented by looking at [the visual pathways from below], and [the visual pathways from the side].

1. Photoreceptor Signals

When a photoreceptor "captures" a photon, the protein part of the photopigment ("opsin") can change shape ("become activated", "be isomerized"). This is often notated as R -> R*.

This change in the shape of the photopigment triggers a biochemical cascade (we'll skip the details). This results in the hyperpolarization of the receptor (i.e., its charge decreases relative to the outside).

In the dark (at rest), the photoreceptors are quite active, constantly releasing neurotransmitter. After absorption of a photon, the resulting hyperpolarization decreases the amount of neurotransmitter released. This means that light actually turns receptors off.

These hyperpolarizations are graded responses. Gradual increases in light intensity have gradual effects on neurotransmitter release.

The rods and cones are connected to horizontal and bipolar cells. These cells then connect to the retinal ganglion cells. Refer to detailed and more schematized figures.

2. Retinal Ganglion Cells

The retinal ganglion cells represent the output of the retina. They exhibit several important properties that are characteristic of many visual neurons. Begin by reviewing the basic anatomy of a neuron: cell body ("soma"), dendrites (which receive inputs from other cells), and the axon (which sends outputs to other cells).

Action Potentials. As opposed to graded responses, ganglion cells fire action potentials. These electrical "spikes" are the basic way that information is transmitted in the brain, and so we'll spend some time understanding how they work.
- Retinal ganglion cells have a negative resting potential (~ -70 mV). At this resting potential, there is a tension between the concentrations and charges of sodium and potassium ions inside out outside the neuron. Sodium ions would like to enter the neuron, while potassium ions would like to leave it.
- When the ganglion cell receives a supra-threshold level of input from the bipolar cells, voltage-gated sodium channels suddenly open, and Na+ ions rush in.
- This causes a sudden reversal of the charge (from negative to positive), and this propogates down the body (axon) of the neuron. [diagram of an action potential].
- Then, the sodium channels close, and voltage-gated potassium channels open, which restores the cell to its (negative) resting potential.
- Key points about action potentials:
  - Spikes are all-or-none, discrete, stereotyped events.
  - A certain threshold-level input must be achieved in order to produce a spike. Below that, no spikes.
  - Refractory period: after a neuron has fired an action potential, it cannot fire another until some time has passed.
  - The firing rate of the neuron (e.g., spikes per second) is what represents information about stimulus intensity within each neuron.
Receptive Fields. Retinal ganglion cells fire action potentials in response to certain types of retinal stimulation. The part of the retina that needs to be stimulated in order to elicit a spike is the retinal ganglion cell's receptive field.
- The receptive field of a neuron can be defined, more generally, as:
  - The part of the visual world that the neuron is responsive to (that it "sees").
  - What the visual stimulus needs to be in order to elicit spikes.
- Most retinal ganglion cells have concentric (or center-surround) receptive fields. [Retinal ganglion cell receptive field].
- These receptive fields are divided into 2 parts (center/surround), one of which is excitatory ("ON"), the other inhibitory ("OFF"). For an ON/OFF center/surround cell, a spot of light shown on the inside (center) of the receptive field will elicit spikes, while light falling on the outside ring (surround) will suppress firing below the baseline rate. Opposite results for an OFF/ON cell.
- In class, we will watch a film of the original researchers who first characterized receptive fields (Hubel and Weisel) defining the receptive field of retinal ganglion cells.
- A perceptual consequence of these receptive field shapes can be seen in a Hermann grid. When you stare at a set of dark squares separated by a grid of white lines, you will see darkness at the intersections of the white lines.
  - Referring to the 2 following figures, note that at the intersections, there is relatively more light (white) falling on the inhibitory surround of ON/OFF receptive fields (compare to the amount of white falling on the inhibitory surround of receptive fields that are not centered over intersections. This relatively-larger amount of inhibition suppresses the firing of cells with receptive fields that overlie the intersections, decreasing their firing rates, and yielding the perception of darkness. [response at an intersection] [response off an intersection]
  - However, note that the darkness disappears at an intersection that you stare at directly. Why do you think this happens?
Lateral Inhibition. Visual neurons do not simply "pipe" the output of the retina through the visual pathways. Instad, the activity of a given neuron is affected by the activity of nearby neurons. Lateral inhibition in the retinal ganglion cells is a prime example.
- When a retinal ganglion cell fires action potentials, it also inhibits the firing of nearby (lateral) ganglion cells.
- Lateral inhibition performs edge (contrast) enhancement.
- A perceptual example of this edge enhancement: Mach bands. The borders between light and dark parts of the image are exaggerated and appear as extra-light and extra-dark bars. Note that your perceptual experience can be predicted by the responses of the retinal ganglion cells.

Interlude. The Representation of the Visual Field in the Brain.

As the pattern of light reflected off objects in the world enters the eye, it is flipped upside-down(recall from last lecture).

This upside-down pattern of light is the retinal image.
The spatial structure of the retinal image is preserved as neurons from the retina connect to the LGN, and is still preserved further along in the cortex.
Although you have two eyes, and slightly different images in each eye, the brain does not keep the information from the two retinas separate. Instead, it splits the world into a left and right visual field. Check out the figure to learn what the horizontal and vertical meridia are.
After leaving the retina, the outputs of each eye are split. The nasal (toward the nose) half of each eye's visual field crosses from one side to the other at the optic chiasm. The temporal half (towards the temple) remains on the same side as its eye-of-origin. This splitting and crossing re-organizes the retinal outputs so that the left hemisphere processes information from the right visual field, and the right hemisphere processes information from the left visual field. Check out these two figures which diagram this tricky wiring pattern: [visual maps 1] [visual maps 2]

3. Parallel Pathways and the LGN

Retinal ganglion cells actually come in 2 sorts: M (magnocellular, or parasol) and P (parvocellular, or midget).

P cells also exhibit color-opponent responses: their firing is also dependent on the wavelength of light in their receptive field. M cells do not exhibit color-opponency.
M cells make transient responses: they fire action potentials when a stimulus is introduced, but quickly fade if the stimulus does not change. P cells, meanwhile, give sustained responses to stimuli in their receptive field.

The division of M and P pathways becomes anatomically evident in the lateral geniculate nucleus (LGN), a part of the thalamus that acts as a relay station between the retina and the cortex.

While the receptive fields of LGN neurons look a lot like retinal ganglion cells, the LGN does re-organize these circuits into distinct layers.

There are 6 layers in the LGN. Bottom 2 layers are M cells, upper 4 are P cells. Layers alternate eye of origin. [LGN anatomy]
Parallel pathways generally exhibit these 4 main characteristics:
1. Physiologically/functionally distinct. For example, the M cells conduct neural signals faster, while P cells represent more constant stimulus presence. A simple hypothesis is that M cells contribute to fast/transient processing (visual motion perception, eye movements) while P cells contribute more to recognition (object reccognition, face recognition, etc).
2. Anatomically distinct. The dendritic trees of P cells are always smaller than the M cells (remember that they're also called "midget" cells). Note that dendritic trees of both types of cell get larger as you move from fovea to periphery.
3. Complete coverage (or nearly complete). Both M and P cells cover the entire retina.
4. Recombine. The M and P cells are separated in the LGN (different layers) but recombine in visual cortex (although some separation still exists).

4. Gross Anatomy of the Cortex

Before we discuss visual cortex in detail, let's stop and get oriented in the brain as a whole.

As you know, there are left and right hemispheres of the brain. They are connected by a tract of fibers known as the corpus callosum.
In each hemisphere, there are 4 "lobes": frontal, temporal, parietal, occipital.
Visual information is processed primarily in the occipital lobes, but parallel pathways extend into the temporal and parietal lobes as information-processing becomes increasingly specialized.

Interlude. Spatial Organization in the Brain

The spatial organization of the brain often provides hints about what the brain does to transform sensory input to useful information for the guidance of action and thought. Spatial organization can be seen at many different levels:

Functional specialization: different types of information are processed in different parts of the brain (with varying degrees of separation).
Columnar architecture: within a brain area, neurons with similar (or complimentary) sensitivities lie close together, often in "columns" or "pinwheels".
Topography/Retinotopy: a "map" of the visual world (or, a map of the retina) is preserved in many visual brain areas. E.g., adjacent points of the visual world/retinal image are mapped onto (or processed by) adjacent neurons. Just as there is a "retinal image", there is a "neural image" in each visual area. People who study the visual system often use the existence of multiple retinotopic maps to localize different brain areas. A technique for localizing visual areas in humans using retinotopy measurements was developed here at Stanford.

5. Primary Visual Cortex: V1

V1 has a topographic/retinotopic map of the visual world (see above). This means that there is a "neural image" that retains the spatial layout of the pattern of light that falls on the retina. This map has several interesting characteristics:

Remember that there are 2 V1s in each person (left and right hemispheres). Each V1 has a representation of the opposite half of the visual field (e.g., left V1 has a map of the right visual field, and vice versa). Note that each V1 does not simply receive input from the opposite eye. The outputs of each retina are split (left half/right half) and then run through the LGN to the appropriate V1.This diagram of the visual fields is helpful.
Just as othe image of the world is inverted when projected onto the retina, the retinotopic V1 map is upside down. As discussed earlier, the right hemisphere's V1 has a topographic map of the left visual field, and vice versa.
Cortical magnification: more cortical space is dedicated to the fovea than the periphery (remember the higher density of photoreceptors in the fovea, hence clearer vision).

There are 3 main types of cells in primary visual cortex.

Simple: receptive fields often have a long, narrow bar of light (ON) and flanking (OFF) parts. Other types are the opposite (responding to dark bars) or simply respond to a light/dark edge. [simple cell receptive field types]

Complex: bars of light must be oriented correctly, but can appear anywhere in the receptive field. Moving the bar through the field produces a sustained response. Complex cells often show direction-selectivity: they fire more when the bar moves in one direction, and are suppressed by motion in the opposite direction. [complex cell responses]

End-stopped (formerly Hypercomplex): Many simple and complex cells exhibit length summation: if an appropriate bar is placed in the visual field, they fire action potentials; if the bar is made longer, they fire more, up to the extent of the full receptive field. However, end-stopped cells increase their responses with increases in bar length up to a limit that is smaller than the receptive field.

In class, we'll see more of Hubel and Weisel, this time defining the receptive field of various V1 cells.
Architecture of V1
- Ocular dominance columns: as you move parallel to the surface of V1, there are alternating columns of cells that are driven predominantly by inputs to a single eye. These alternations between left and right eye are the ocular dominance columns.[ocular dominance columns]
- Orientation columns: as you move perpendicular to the surface, the preferred orientation of the cells changes gradually from horizontal to vertical and back again.
- View a schematic of the ocular dominance and orientation columns together.

Interlude. Defining and Separating Different Brain Areas

Brain areas can be differentiated according to 4 main criteria:

Function: physiology. Neurons in different parts of the brain are responsive to different aspects of the stimulus (= do different things).
Architecture: microanatomy can differ widely across brain areas. For example, V1 is also referred to as "striate cortex" because it has a series of stripes that run parallel to the surface; these stripes end abruptly at the end of V1.
Connections: different areas feed forward and also receive backward-reaching connections from distinct areas.
Topography: e.g., retinotopy. Each distinct visual area has its own retinotopic map.

Remember 'FACT' as a mnemonic.

6. Secondary Visual Areas

There are approximately 30 visual areas after V1. The functional specialization hypothesis drives much of the research about these areas. Some areas seem specialized for processing a certain aspect of visual information. E.g., MT - motion, V4 - color (?).

Cortical areas dedicated to vision are densely interconnected, and can seem quite confusing at first glance.
However, a more general organization is evident in a pair of parallel pathways.
- What pathway. Temporal lobe; recognition of objects.
- Where pathway. Parietal lobe; motion, spatial orientation, localization.

Interlude. Methods for Measuring Brain Activity

Methods of historical interest. Blood flow is selectively increased in parts of the brain that are active. Mosso (late 1800s) measured "brain pulsations" in infants, and found larger pulsations during periods of mental activity.
Electrophysiology. Can be done at the single-cell level in research animals. Invasive technique using an electrode implanted into cortex to measure neuronal spikes.
Optical imaging. Measure of blood flow. By opening the animal's skull, pointing a video camera at the exposed cortex, and collecting images that reflect the amount of oxygenated and deoxygenated blood researchers can measure the activity of a section of brain.
PET = "positron emission tomography". Subject is injected with a mild radioactive substance and lies in a PET scanner. Because blood is directed to areas of increased activity, more radioactive substance ends up in the active areas, and the PET scanner detects this.
fMRI = "functional magnetic resonance imaging". Just as a regular MRI machine can be used to image soft tissue in your knee, it can be fine-tuned to detect relative changes in the concentration of oxygenated and deoxygenated blood in the brain. Therefore, we can detect areas of increased neural activity. This technique is non-invasive (no surgery, no radiation), so it is the current technique of choice for measuring brain activity in humans.
EEG = "electroencephalograph". Electrodes placed on the scalp detect small changes in electrical activity. However, it is difficult to localize the source of these electrical signals.

Alex Huk

Last modified: Tue Jun 29 10:58:53 PDT