Beyond Phototransduction: What Happens after the Retina
In the first part of this series of three articles we reviewed in some detail how photons are ‘converted’ or transduced into a neurochemical signal, the language of the brain, by photoreceptor neurons in the sensory retina. In this part we briefly discuss how visual information is processed by the rest of the retina and other parts of the brain. The last article in this series we will discuss some of the information theoretic implications of visual information. In other words, how is the information in the visual system actually encoded from a mathematical perspective. Intellectually, the topics we will introduce in Part III are fascinating and deep subjects.
The inner retina
After the photoreceptors the biology gets more complicated. As we move through higher levels of the brain, the number of open questions that remain about how the brain represents and processes visual information are much greater than the process of phototransduction itself. There is still a lot we do not know. In the neural retina, there is an increasing complexity (heterogeneity) in the number and types of neurons as we move to layers beyond the photoreceptors, both in terms of their structure (morphology) and function (physiology). At its most basic, photoreceptors pass their signals to bipolar cells, which in turn signal retinal ganglion cells. As I mentioned in the first article, the axons of the ganglion cells then project to other parts of the brain. There are about eleven different types of bipolar cells, and many more types of ganglion cells, twenty- two identified types in the mouse retina alone for example. Ganglion cells are very diverse, and the receptive fields of different classes of ganglion cells, the piece of retinal real estate they respond to, are tiled so that although the receptive field of one neuron ends where the another field begins for ganglion cells of the same class, different classes of ganglion cells overlap the entire surface of the retina, and therefore visual space, repeatedly. In total then, ganglion cells send about twenty or so parallel pathways of information to other parts of the brain, each encoding different aspects of the visual scene.
In addition, there are neuron classes and circuits orthogonal to the primary direction of information flow from photoreceptors to bipolar cells to ganglion cells, which are also be very diverse and heterogeneous. Amacrine cell neurons send processes that synapse at the bipolar cell-ganglion cell layer. Because one of these neurons can synapse across multiple bipolar-ganglion synapses, they create neural circuits that connect different bipolar cells and ganglion cells that may otherwise not have been connected. In effect creating a sort of physiological ‘short circuit’ designed to filter and modify visual information through feed back loops as it passes from the bipolar cells to the ganglion cells. This seems to be important in the pre-processing of the raw visual information by the retina before it goes off to other parts of the brain. Important properties such as the detection of edges, changes in contrast, and other properties that provide key information about the visual scene are selected and filtered out. Similarly, horizontal cells at the photoreceptor-bipolar cell synapse and interplexiform cells that send processes to both the photoreceptor-bipolar cell and bipolar cell- ganglion cell synaptic layers also participate in this pre-processing of visual information by the retina. Exactly what all these neuron classes do and how they do it continues to be extensively studied, and although we understand quite a bit we still do not fully understand their physiology. The neurobiology of the inner retinal circuit continues to surprise us.
Projections from the retina to visual cortex
Beyond the retina, the axons of the ganglion cells bundle together to form the optic nerves, which project to a part of the brain called the lateral geniculate nucleus (LGN) located deep in a structure of the brain called the thalamus. The thalamus is a sort of integration and relay center, where information comes together through the spinal cord, other parts of the brain such as the hippocampus, and all the senses with exception of smell. Connections from the thalamus project to various parts of the cerebral cortex for higher level processing, although importantly, there are numerous feed-back loops back to the thalamus so that there is a continuous crosstalk between the thalamus and cortex (corticothalamic loops). The LGN acts as the primary relay of vis-ual information to primary visual cortex (V1) in the occipital lobes at the back of the cortex. It has a layered anatomical structure where ganglion cell projections from the retina overlap with local LGN circuit connections. This structure serves important functions associated with the correlation and de-correlation of spatial and temporal visual information that allow us to build representations of objects in three-dimensional space. Interestingly, how the ganglion cell axons project to the LGN’s (we have two, one on each side of the brain) is not trivial.
For the left eye, the temporal (or lateral) visual field, which is the half of the visual field from the middle of your eye to the outside towards the left ear, is detected by the nasal half of the retina, the half of the retina on the inside of the eye closest to the nose. The nasal visual field for your left eye, that is the half of the visual field from the middle of your eye to the inside closest to the nose, is detected by the temporal half of the retina of the left eye. Now switch everything for the right eye. The nasal visual field is detected by the temporal half retina and the temporal visual field by the nasal half of the retina. So now we have to follow each half of the retinal projections from both eyes to the LGN’s. The nasal half of the retina in the left eye projects to the right LGN, as does the temporal half of the retina from the right eye. This means that the inner (nasal) half of the visual field from your left eye and outer (temporal) half of the visual field from your right eye both project to the right LGN. And as you can probably guessed by now, the temporal half of the retina from the left eye and nasal half of the retina from the right eye both project to the left LGN, so that the left LGN receives information about the nasal left visual field and temporal right visual field. This crossing of the fibers occurs anatomically at a location in front of the LGN called the optic chiasm.
The reason for this seemingly overly complicated crossing is because the visual cortex takes advantage of superimposed in- formation from the same hemispheric visual fields to construct binocular vision. Clinically this can produce some unique vision defects. If you are not aware of how the ganglion cell fibers divide and cross it can lead to a rather strange experience for patients, but is quite telling for a neuro-ophthalmologist. Hemianopia is a distortion or loss of the half of the visual field (blindness) in one eye or both. There is usually a descriptive term that precedes it that describes exactly what that loss is. For example, temporal hemianopia means a loss of the lateral temporal half of the visual field in one eye or both. Bi-nasal hemianopia means loss of the nasal half of the visual field in both eyes. Depending on the cause though these defects often correct themselves with some time.
We then finally arrive at the visual cortex itself. The visual cortex is anatomically and functionally sub-divided into five different regions, referred to as V1 through V5. We know and understand progressively less the higher up we go. All of this continues to be very active and important areas of research. We know quite a bit about the anatomy, connections, circuits, and cell types of V1. V1 is the primary visual cortex and plays key roles in extracting and processing much of the key visual information from the retina that leads to the perception of the visual world, such as the detection of edges and contrast. V1 maintains a retinotopic map of the retina, meaning that the spatial organization and positions of ganglion cells in the retina are preserved by the spatial organization of neurons in V1. The LGN also maintains retinotopic maps of the retina, which it then preserves as it projects to V1.
The other regions of the visual cortex presumably use this information to create our internal representation and perception of the visual world. They are also responsible for the use of visual information as it relates to other cognitive functions, such as the triggering of memories from an image for example. And we know that V5 plays a key role in the perception of motion. At present however we still only have a partial picture about how the visual system works at the level of the cortex. There is a lot we know and understand both anatomically and physiologically, but our understanding remains incomplete. In certain cases we only have fragments of the correct picture, and in some cases there are controversial and even conflicting viewpoints that remain to be resolved. What we lack most is an accepted and complete picture of the computational aspects of how vision is produced. What I mean by this is that even for much of the data and biological details we do have, we are no where near a complete mathematical and computational framework that puts this data into an appropriate descriptive and context that tell us how vision is produced, how we perceive the visual world, or how visual information is represented and processed to produce the cognitive properties that result from vision. The brain is a dynamic system, and to understand it we have to treat it as such.
At a high level though, we know that although the retina is responsible for taking in and in part processing the raw visual information, the perception of the visual world is achieved in the cortex, even though we do not fully understand how. Or put another way, we actually see with our brains and not with our eyes. This is why optical illusions work or why one person can see (or not see) something completely different than someone else in a piece of abstract art or a cloud. Our interpretation of the physical world around us through all our senses, including vision, is a very personal and private experience. I can try to communicate to you what I see, or what the color red looks like to me, but you cannot ever truly know, nor can I ever truly tell you, and vice versa.