Neural networks and vision
Recording information from a few neurons is not enough to explain the part played by the vast networks of nerve cells involved in processing sensory information to arrive at an image. The CODING_IN_V1 (How visual information is represented by neuronal networks in the primary visual cortex) project used in vivo two-photon calcium microscopy to find out how neural networks encode visual information. They also looked at how these networks gained processing ability during postnatal development. The researchers recorded simultaneous activity from hundreds of neurons in the visual cortex while applying different visual stimuli to anaesthetised mice. Fluorescent proteins expressed in distinct cell types showed how excitatory and inhibitory neurons interact to shape an overall response during vision. CODING_IN_V1 developed a new method for measuring synaptic connections in cells whose role in the visual function network is known. For example, the scientists looked at neurons that are responsible for the edges in images and how being interconnected is translated into their input in the image. Overall, the project scientists uncovered some of the basic principles of how images are built up by nerve networks. Future extension of this line of research can be applied in other regions of the brain to understand the fundamental problems behind conditions with a neurological basis such as schizophrenia, autism and epilepsy.
Keywords
Neural network, vision, CODING_IN_V1, image, two-photon calcium microscopy