Document Actions

You are here: Home / News / Research News / How we learn to see. Active efficient coding explains the development of binocular vision

How we learn to see. Active efficient coding explains the development of binocular vision

Using a new model, scientists from the Frankfurt Institute for Advanced Studies (FIAS) and Goethe University can explain the development of binocular vision, which is the joint vision of the right and left eye. In doing so, they show how amblyopic visual impairment can occur.

Sehschwäche

A widespread early childhood developmental disorder is amblyopia (about 5% of all children are affected). It leads to visual impairment in one eye (in rare cases both eyes). However, the causes are still poorly understood. The standard therapy method is to temporarily close the suppressing eye. However, this very often does not lead to the desired success.

We view the world with our two eyes from slightly different perspectives. The small differences between the images projected onto the left and right retina are used by our brain to see spatially, i.e. to interpret the world in three dimensions. However, a widespread early childhood developmental disorder is amblyopia, which affects up to 5% of all children. It leads to visual impairment in one eye (or in rare cases both eyes). The brain does not learn to coordinate the two eyes correctly and to interpret the images of the two eyes in three dimensions. Instead, the eyes enter into a "competition" in which one eye "suppresses" the other.

Yet, the reasons for this are still poorly understood. The group around FIAS Senior Fellow Prof. Dr. Jochen Triesch has now proposed a new computer model of the development of binocular vision in collaboration with Prof. Dr. Bertram E. Shi from the Hong Kong University of Science and Technology. The model explains how precise binocular vision calibrates itself automatically under healthy conditions. Furthermore, it describes why amblyopia develops in case of refractive errors of one eye. Interestingly, the model also shows conditions that are necessary for successful treatment.

Newborns have difficulty focusing on objects and cannot yet align their eyes precisely to the same point. They are also unable to interpret the images of both eyes in three dimensions. How infants learn the information processing and control mechanisms required for this is currently unknown. In certain cases, it also happens that the calibration of eye movement and focus control is not learned correctly. For example, anisometropia describes a difference in the refractive power of the eyes. If this is not corrected early in development, it can cause amblyopia: A disorder of the developing visual system characterized by a difference in the visual acuity of both eyes that cannot be corrected immediately by correcting the refractive power, e.g. with glasses. Amblyopia can be associated with a loss of spatial vision and in severe cases can lead to unilateral blindness.

The now presented "Active Efficient Coding" theory tries to explain the learning mechanism for binocular vision. It suggests that the brain optimizes both eye movements and the subsequent processing of the images of both eyes simultaneously in order to represent the visual impressions as compactly as possible. This works similarly to a compression algorithm that reduces the size of a computer file, but by learning coordinated eye movements, the data itself is optimized. Under healthy conditions, the model calibrates itself to precisely coordinate eye movements and interpret the visual sensations in three dimensions. During this process, nerve cells develop which register small differences between the images of the two eyes, so-called disparities. The model uses this to align the eyes to the same point. However, in the case of anisometry, in which the refractive power of the two eyes is different, the model develops an amblyopia-like state in which the information of one eye is not processed but suppressed. This leads to the development of nerve cells that only process information from the healthy eye. Hardly any disparities can be registered. The visual sensory impressions can therefore no longer be interpreted in a normal spatial manner. In addition, the model shows that a cure for this condition is only possible if the nerve cells are still able to change their processing of sensory impressions, a process known as plasticity. Overall, the model offers a consistent explanation for the development of binocular vision and its failure in the case of amblyopia.

Further Information and Contact:

Contact: Prof. Dr. Jochen Triesch, Frankfurt Institute for Advanced Studies und Fachbereich Physik und Fachbereich Informatik und Mathematik, Goethe-Universität Frankfurt, Campus Riedberg, Tel. 069 798 47531, triesch@fias.uni-frankfurt.de.

Publication

Active efficient coding explains the development of binocular vision and its failure in amblyopia Samuel Eckmann, Lukas Klimmasch, Bertram E. Shi, Jochen Triesch

Proceedings of the National Academy of Sciences Mar 2020, 201908100; DOI:10.1073/pnas.1908100117