Abstract of Biology Master’s degree thesis “Artificial neural network modeling of visual system”

Humans and other animals process information with neural networks. Computer algorithms that mimic these biological structures are formally called artificial neural networks.

The drawback of nervous system is that the neurons lose the ability to divide and form new cells after birth (amitotic). Various accidents, diseases, aging reduce the number of neurons. In some cases the loss is so huge that entire organism function is lost (e.g. vision, hearing), however neurons cannot divide and thus regenerate the damaged area. Technology enables us to restore the lost organism function by artificial means, as for example mechanical heart valve, replacement of lens in eye, pacemakers, however the restoration of the damaged part of nervous system still presents the significant problem, because the needed theoretical background is only beginning to evolve.

This work investigates how to integrate the theory of artificial neural networks (ANN) with the circuitry the biological visual system.

The new ANN model of early visual system has been proposed that mimics the response of the biological visual system to the visual non moving stimulus. As the basis for ANN modeling were chosen the early steps of the visual pathway of the cat, that is, the processing of visual information from receptors, through the bipolar and retinal ganglion cells to the geniculate relay cells. The thesis deals only with nonlagged X-cells, however other types of cells can be modeled analogically. ANN model of early visual system simulates spatial receptive field organization of ganglion and geniculate relay cells.

ANN model of early visual system provides theoretical basis for eye prosthesis for the blind or the construction of the eye for the future cyborg.