-
Notifications
You must be signed in to change notification settings - Fork 0
Analysis plan context
Our goal is to understand how neuronal activity evolves during learning. We are especially interested into the first few stages of sensory processing where the visual stimulus is transformed from a very large number of temporal "pixels" into meaningful decision that are relevant to the animal. To this end, the cortex needs to drastically reduce the dimensionality of the problem. In this page, I hope to provide some context into our behavior experiments.
TBD
The eye is paved with hundreds of thousands of photoreceptors cells. Mice are nocturnal animals so their retinas is dominated by rods. The output of the retina is made by Ganglions cells. Those are the only cells in the retina that project an axon out of the eye. Mice typically have about 50,000 axons per eye. See Source or Source
Our stimuli are about 40 degrees diameters. The mouse retina typically covers about 110 degrees vertical, 140 degrees horizontal Source. Let's assume that all axons cover an equal solid angle (which is typically not true but not crazy See about area centralis ). Let's also assume the total solid angle is 110 degrees or 2pi(1-cos(55*pi/180) (See), i.e. 2.7 sr. With 50,000 axons, we end up with 18500 axons per s.r.
A 40 degree stimulus is 2pi(1-cos(20*pi/180)) = 0.37 s.r. large so cover approximately the receptive field of about 6800 ganglion cells.
All ganglions cells do not project directly to the cortex via the thalamus, some have other brain targets (See the general circuitry of the mouse visual system). However given that the center of the mouse retina is a little denser (about 3 times) than the periphery, it would seem that all together, 5000 axons is a good order of magnitude as the number of output signal activated within a 40 deg large stimulus. Given the large functional repertoire of the retina, some of these cells will see increase in firing, some will see decreases, others will focus on some particular spatio-temporal component of the visual stimulus, some will not respond. See this review for an extended description of these classes.
Our eyes are not fixed but can move. Each eye has 3 degrees of rotations of their own : 2 spherical rotations and one torsion. Each eye is controlled by 6 independent muscles that form 3 pairs. Grossly, one pair controls lateral movements, one pair vertical movements, and one other pair control the torsion.
In humans and monkeys, it has been shown that the muscle system is organized in such a way to simplify the motor control of the eye so that there is always a single possible torsion for every gaze of the eye. This ensures that the image on retina does not change with the succession of rotation that the eye follow.
For those interested, you can easily demonstrate that rotation are not commutative. To guarantee one and only one torsion per eye gaze, the motor system of the eye fulfill the Listing Law using a system of miniature pulleys.
The 50,000 axons that leave the eye to the thalamus preserve their original retinal coordinate system. They are topologically organized so that once the information reach the cortex, you can precisely map the retinotopy of the visual cortex. Given that both the eye and the head can move, the cortex needs to integrate both the head positon AND the eye position with the visual input to reference any object in a more general coordinate system.
The overal anatomy of the visual system is supposed to help with that work.