NEURO-ANATOMY OF PRIMATE VISUAL CORTEX
-
-
THE VISUAL CORTEX
--
A cortex [Latin: skin or bark] is basically
a 2-dimensional sheet that is highly convoluted.
Many parts of the brain have a cortical structure.
--
The striate cortex (also called the visual cortex or V1)
is the main part for visual processing.
--
Its dimensions are about 100 to 1000 mm-square,
and about 1 mm thick.
--
The visual cortex has many layers called the fine structure
or functional architecture [Hubel-Wiesel'74].
--
At least 10 layers are identified.
--
Fibers from the LGN (lateral geniculate nucleus) enter layer IV.
--
[An equivalent number of fibers project back to the LGN. Their
functions are unknown, but there are speculations that this
is used for imagination and reuse of stored visual data.]
-
-
EXISTENCE OF A 1-1 MAP BETWEEN THE RETINA
AND LAYER VI OF THE STRIATE CORTEX
--
There is a point in the cortex that is maximally responsive
to stimulus at a point in the visual field
[Hubel-Wiesel'74].
-
-
LOCAL PROPERTIES OF THIS MAP:
--
The magnification factor is the distance (in mm) along
the cortical surface that corresponds to change in
stimulus position of 1 degree in the visual field.
--
This factor is independent of local direction, and vary
roughly inversely with eccentricity (=distance from fovea).
[Daniel-Whitteridge'61].
--
Circles of constant eccentricity in retina are mapped to
parallel straight lines in the cortex.
[Daniel-Whitteridge'61, Talbot-Marshall'41].
--
Up to a third of the visual cortex is devoted to
processing the data from the fovea.
-
-
LOGMAP MODEL
--
Schwartz [1977] proposed the complex logarithm
(x,y) --> (r, \theta)
where
\log (z+\alpha) = r . e^{i\theta}
and z= x+iy
as the functional form of this map.
--
This map has properties that agree with the above experimental data.
--
The real constant \alpha is species-dependent.
(e.g., \alpha=0.3 for some monkeys)
--
It is understood that this map applies to the right
half of the retina, and the left half is obtained by symmetry.
--
DRAW A PICTURE OF THIS MAP FOR THE HEMIDISK.
-
-
HYPERCOLUMN STRUCTURE
--
The left and right eye signals terminate in the input layer
of the striate cortex, in a zebra-striped pattern.
[SHOW PICTURE]
--
The individual stripes are called ocular dominance columns
which are about 0.5mm wide.
--
These columns from the left and right eye are interlaced,
and account for stereo vision.
--
Another kind of structure, called orientation columns
that are about 50 microns by 500 microns) are slabs of tissue
cells that respond to edges of constant orientation in
the visual field.
--
Cells in a small patch of the cortex (about 1 mm square) contains
complete left-right eye representation (from the
ocular dominance columns) and orientation column representation
for a full 180 degrees. Hubel and Wiesel call such a patch containing
a complete visual processing unit a hypercolumn.
-
-
HOW TO OBTAIN BRAIN DATA:
--
Basically two ways: micro-electrodes or staining.
--
2-Deoxyglucose (2DG) method: Expose image to visual field of
primate for 30 minutes (the primate must be immobilized).
This produces images of associated neural
activity on the brain surface. After killing the primate (gulp!),
its brain (layer VI) is stained, and slices made. Then comes the
reconstruction from these slices (including flattening the
folds of the cortical surfaces). Computer techniques are
needed to help the reconstruction.
-
-
REFERENCES:
1. [Schwartz-Merker'86]
2. [Rojer'90] -- PhD Thesis, NYU
3. [Franzblau'91] -- Masters thesis, NYU
4. [DeValois-DeValois'88] -- Book called "Spatial Vision"