Resumen de Modelos y herramientas de simulación de las primeras etapas del sistema visual
The focus of this thesis was on the implementation of software tools and biological models that simulate the neuronal behavior of the first stages of the visual system, retina, lateral geniculate nucleus (LGN) and primary visual cortex (V1), and can serve as a realistic experimentation framework within which several hypotheses of the neural coding of visual processing can be explored.
First, we developed an efficient software platform that facilitates the implementation of retina models at different abstraction levels, from single-cell to large-scale network levels. The platform provides a set of computational retinal microcircuits as basic building blocks that can be combined to form different retina architectures. To show the configurability and potential of the proposed framework, we constructed a series of different retina models that capture the properties of the retina response for some of the best-known phenomena observed in the retina: adaptation to the mean light intensity and temporal contrast, and differential motion sensitivity.
The next stage that was investigated was the LGN. A striking feature of the LGN circuit is that LGN cells, both relay cells (RCs) and interneurons (INs), not only receive feedforward input from retinal ganglion cells (GCs), but also a prominent feedback from cells in layer 6 of V1. We explored the spatial effects of cortical feedback on the relay-cell response by means of a biophysically detailed network model. We considered two different arrangements of synaptic feedback from the ON and OFF zones in V1 to the LGN, as well as different spatial extents of the corticothalamic projection pattern. Our simulation results are in agreement with the feedback-evoked increase in center-surround antagonism observed in experiments both for flashing spots and, even more prominently, for patch gratings.
Finally, we developed a comprehensive network model of the first stages in the primate parvocellular pathway, built upon two-dimensional grids of point neurons, which represent the retina, the LGN and a simplified version of the multilayered structure of V1. Special attention was given to ensuring that the morphological properties of the network (e.g., spatial extent of connections) were based strictly on experimental data of the primate visual system. We exhaustively benchmarked the model against well-established chromatic and achromatic visual stimuli, showing spatial and temporal responses of the model to light flashes of different shapes, spatially uniform squares and sine-wave gratings of varying spatial frequency. The model was used to validate a hypothesis that is under debate and concerns the spatial properties of the V1 response to surfaces of uniform color. According to this hypothesis, V1 population responses to chromatic and achromatic surfaces remain both edge-enhanced throughout the stimulus presentation but only achromatic surfaces elicit a neuronal “filling-in” response of the center.
In parallel, different optimization strategies based on genetic algorithm (GA) were investigated to fit parameters of some of the models proposed in this work.
