Human cortical responses to binocular disparity and stereoscopic depth

Functional magnetic resonance imaging (fMRI) can be used to visualise activity in the brains of human subjects. When a region of the brain becomes active the neurons increase their activity. An increased flow of oxygen is required in that local region to fuel this activity. Since oxygenated (fresh) and de-oxygenated blood have different magnetic properties, it is possible to use the ratio of these two substances to measure changes in neural activity.

You may have seen ?Magic Eye? pictures in which an object ?jumps? out in depth from the page of a book. This is because there is a slight difference in the images going to the two eyes: it is known as stereoscopic depth perception. Using stimuli similar to these Magic Eye pictures, we can employ fMRI to investigate which areas of the human brain are responsible for producing this depth percept from the small differences in the images.

Deficits in depth perception are common, as any misalignment of the two eyes during early childhood due to ?squint? or ?lazy eye? can prevent the development of neurons that receive input from both eyes. Although the eyes can be realigned surgically, most children do not regain their binocular vision. Once we have established the areas of the brain involved in depth perception, it will be possible to compare the effectiveness of different treatments for these deficits.

In natural viewing, humans acquire information about the depth of objects from the slightly different views of the scene obtained by the left and right eyes. In the laboratory, the same information can be provided simply by adding a slight horizontal offset (binocular disparity) to regions of the images presented to the two eyes. Our earlier neurophysiological studies demonstrate that multiple computational steps are required in the visual cortex to produce this depth percept.

This project will use functional magnetic resonance imaging (fMRI) to investigate the stages in the human visual cortex that begin with the detection and measurement of binocular disparity and lead to the perception of depth. This brain scanning technique is particularly valuable because cortical activity can be measured simultaneously across all visual areas while human subjects perform a perceptual task.

Based on neurophysiological results in animal studies, we propose four experiments.
The combination of these experiments will allow us to describe the network of areas involved in the computation of binocular depth perception from binocular disparity.
(1) Comparison between the response to correlated and anticorrelated random dot stereograms (RDS). While V1 neurons respond to both types of RDS, anticorrelated RDS do not deliver a consistent depth percept. Visual areas that reflect perceptual judgments should respond only to correlated RDS. (2) Investigation of absolute and relative disparity. Early stages of vision measure binocular disparity with respect to where the eyes are aligned (absolute disparity), whilst later stages that dominate perception compare the disparity between two or more visible features (relative disparity). (3) Separation of local and global disparity information using sinewave grating stimuli. When a disparity is added to a patch of grating, V1 neurons respond to the local disparity of each individual bar of the grating, but subjects report the global change in depth of the patch. (4) Cancellation of spatial disparities in depth perception. Binocular depth perception in dynamic patterns can also be produced by an interocular delay, which can be expressed as a temporal disparity. Spatial and temporal disparity can be traded off against each other to produce a stimulus that contains spatial binocular disparity, but is perceived as flat.

Once we have an understanding of the normal system of depth perception it will be possible to compare the cortical responses of subjects who lack functional stereoscopic vision and, in the longer term, to evaluate corrective therapy of such deficits.